Datasets:

TaylorAI

/

pubmed_commercial

like 12

PMC005xxxxxx/PMC5002955.txt

==== Front Endocr ConnectEndocr ConnectECEndocrine Connections2049-3614Bioscientifica Ltd Bristol 26830329EC15011410.1530/EC-15-0114ResearchCharacterization of human follicular thyroid cancer cell lines in preclinical mouse models Characterization of thyroid cancer cell linesA. N Reeb et al.Reeb Ashley N Ziegler Andrea Lin Reigh-Yi Department of Otolaryngology, Head and Neck SurgerySaint Louis University School of Medicine, Saint Louis, Missouri, USACorrespondence should be addressed to Reigh-Yi Lin; Emailrlin7@slu.edu3 2016 01 3 2016 5 2 47 54 25 1 2016 1 2 2016 © 2016 The authors2016The authors This work is licensed under a Creative Commons Attribution 4.0 International License.Follicular thyroid cancer (FTC) is the second most common type of thyroid cancers. In order to develop more effective personalized therapies, it is necessary to thoroughly evaluate patient-derived cell lines in in vivo preclinical models before using them to test new, targeted therapies. This study evaluates the tumorigenic and metastatic potential of a panel of three human FTC cell lines (WRO, FTC-238, and TT1609-CO2) with defined genetic mutations in two in vivo murine models: an orthotopic thyroid cancer model to study tumor progression and a tail vein injection model to study metastasis. All cell lines developed tumors in the orthotopic model, with take rates of 100%. Notably, WRO-derived tumors grew two to four times faster than tumors arising from the FTC-238 and TT2609-CO2 cell lines. These results mirrored those of a tail vein injection model for lung metastasis: one hundred percent of mice injected with WRO cells in the tail vein exhibited aggressive growth of bilateral lung metastases within 35 days. In contrast, tail vein injection of FTC-238 or TT2609-CO2 cells did not result in lung metastasis. Together, our work demonstrates that these human FTC cell lines display highly varied tumorigenic and metastatic potential in vivo with WRO being the most aggressive cell line in both orthotopic and lung metastasis models. This information will be valuable when selecting cell lines for preclinical drug testing. Keywords follicular thyroid cancerorthotopicthyroidmetastasismouse model ==== Body Introduction Follicular thyroid cancer (FTC) is the second most frequently diagnosed thyroid cancer after papillary thyroid cancer (PTC). About 10,000 cases of FTC are diagnosed annually in the United States, comprising about 10–15% of all thyroid cancer cases. The majority of FTC patients have excellent prognoses when their cancers are caught early. However, the relative 5-year survival rate drops dramatically to about 50% for patients with advanced metastatic cancer who does not respond to standard treatments. More research is needed to determine the cause of the disease; to study tumor initiation, progression, and drug resistance; and to eventually develop new and improved drugs for this group of patients. Human cell lines derived from tumors removed from thyroid cancer patients provide a useful model of thyroid disease mechanisms. Genetically defined cell lines with known driver mutations are particularly useful for clinicians and researchers striving to develop a more personalized approach to thyroid cancer therapy. Patient-derived thyroid cancer cell lines can also be used for in vitro toxicology studies, as well as in vivo studies in immunodeficient murine models to study cancer progression and drug responses in the tumor microenvironment. Previously, we and other laboratories developed an orthotopic thyroid cancer model that faithfully recapitulates the clinical characteristics of advanced thyroid cancer (1, 2, 3, 4). In addition, tail vein or intracardiac injection models have been developed to study the distant metastasis of human thyroid cancer cell lines (5). In the current study, we characterize the tumorigenic and metastatic potential of a panel of three human FTC cell lines harboring various genetic mutations. The FTC-238 and TT2609-CO2 cell lines each has a TP53 mutation commonly found in many cancers (6, 7). The WRO cell line has a BRAFV600E mutation. This mutation is the most common mutation found in PTC, but it is less common in FTC (8). We engineered these cell lines to express firefly luciferase to allow in vitro and in vivo tracking by bioluminescent imaging using an IVIS Spectrum. In vivo tumor growth was initiated by orthotopically injecting these cell lines into the thyroids of immunodeficient NSG (NOD/ScidIl2rg-/-) mice to initiate primary tumors. These cell lines were also injected into the lateral tail vein to evaluate their metastatic potential for lung colonization. Tumor progression and metastasis were evaluated by bioluminescent imaging weekly as well as postmortem histological analysis. Materials and methods Cell culture The WRO cell line, established from the metastases of a female FTC patient, and TT2609-CO2 cell line, established from a 57-year-old male patient with FTC, were kindly provided by Dr John Copland (Mayo Clinic) with permission from Dr Guy J F Juillard (UCLA) for the WRO cell line and Dr Albert A Geldof (VU University Medical Center, The Netherlands) for the TT2609-CO2 cell line. The FTC-238 cell line, established from a lung metastasis of a 42-year-old male FTC patient, was purchased from ATCC. All cell lines have been confirmed by short tandem repeat (STR) analysis. The WRO cell line was cultured in DMEM/F12 (1:1) with 10% fetal bovine serum (FBS). The TT2609-CO2 cell line was cultured in RPMI1640 with 10% FBS. The FTC-238 cell line was cultured in DMEM/ F12 (1:1) with 5% FBS. The cultures were maintained in a humidified chamber in a 5% CO2:air mixture at 37°C. Quantitative real-time PCR RNA was isolated from 1 × 106 cells with the RNeasy Kit (Qiagen) and treated with RNase-free DNase (Qiagen). Two micrograms of RNA were reverse transcribed into cDNA using the Thermoscript First-strand Synthesis System (Invitrogen). The oligonucleotide sequences of the primers for PAX8, TTF1 TSHR, TG, and TPO have been published elsewhere (9, 10, 11). The mRNA levels were quantified in triplicate by quantitative real-time PCR on a ViiA7 PCR System (Applied Biosystems). Human GAPDH was used as the housekeeping gene during the amplification. Generation of luciferase-expressing cell lines The cells were transfected with a pSIN-luc vector expressing the firefly luciferase gene (a gift from Dr Yasuhiro Ikeda of Mayo Clinic) to generate stable clones. A detailed description of the construction and transfection protocol has been published (12). Animals and ethics statement Eight-week-old female NSG mice strains NOD.Cg-Prkdcscid Il2rgtm1Wjl/SZJ were obtained from Charles River (Boston, MA, USA) and maintained under pathogen-free conditions at Saint Louis University Animal Facility. This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Institutional Animal Care and Use Committee (IACUC) of Saint Louis University. All procedures were in accordance with institutional animal welfare guidelines, and all efforts were made to minimize suffering. Mouse tumor models For orthotopic transplantation, 5 × 105 cells were suspended in 10 μL Matrigel/RPMI or DMEM at a 1:1 dilution and injected into the right thyroid gland of NSG mice (1). The mice were killed when they reached humane endpoints including morbidity, immobility, unresponsiveness, recumbence, failure to eat or drink, and loss of more than 20% body weight. Tumors and adjacent tissues were collected and analyzed by hematoxylin and eosin (H&E) staining. In separate experiments, lung metastasis was experimentally induced by injecting 5 × 105 cells into the lateral tail vein (13). The mice were imaged using IVIS weekly to monitor for metastatic tumor growth. At the time of killing the mice, lung tissues were harvested, fixed in 10% formalin, paraffin-embedded, and stained with H&E according to a standard protocol. Bioluminescent imaging The anesthetized mice were injected with d-luciferin (150 mg/kg) and imaged with an IVIS Spectrum as described previously (14). The light emitted from the bioluminescent tumors was detected, digitized, and displayed, and the regions of interest from the displayed images were quantified as total photon counts per second (photons/s) using Living Image software (Caliper Life Sciences). Results Real-time PCR analysis of thyroid-specific genes Three human FTC cell lines (WRO, FTC-238, and TT2609-CO2) were validated and obtained from the sources indicated in the Materials and methods section. The mutational characteristics of these human FTC cell lines, based on published reports (7, 8, 15), are summarized below. FTC-238 and TT2609-CO2 each has a TP53 mutation, and WRO has a BRAFV600E mutation. None of the cell lines have mutations in HRAS, NRAS, KRAS, PI3KCA, or PTEN. Phase-contrast images of the three FTC cell lines are depicted in Fig. 1A. Next, we used quantitative real-time PCR to examine the expression of thyroid-specific genes in these three cell lines. We found that FTC-238 and WRO express PAX8, a thyroid transcription factor, whereas TT2609-CO2 expresses thyroid transcription factor 1 (TTF1). However, unlike normal human thyroid, none of these cell lines had detectable levels of mRNA for the thyroid differentiation markers thyroglobulin (TG), thyroperoxidase (TPO), sodium-iodide symporter (NIS), or the receptor for thyroid stimulating hormone (TSHR) (Fig. 1B), confirming the dedifferentiated state of these human FTC cell lines. Figure 1. Morphologies and gene expression analysis of human FTC cell lines. (A) Phase-contrast microscopy images of WRO, FTC-238, and TT2609-CO2 cell lines. (B) Quantitative real-time PCR analysis of PAX8, TTF1, TSHR, TG, TPO, and NIS. Y-axis, relative mRNA expression level compared with normal human thyroid. Human GAPDH was used as the housekeeping gene during the amplification for the normalization of the above gene expression data. The results are expressed as mean ± s.e.m. of two independent experiments with three parallels. Generation of stable luciferase-expressing human FTC cell lines To generate luciferase-expressing cell lines suitable for in vitro and in vivo live imaging, WRO, FTC-238, and TT2609-CO2 cells were transfected with a pSIN-luc vector encoding a firefly luciferase gene, and three stable clones from each cell line were selected for further analysis. The levels of luciferase activity in serially diluted cell cultures were quantified in vitro using IVIS in complete media supplemented with luciferin. The results show that luciferase activity was proportional to the number of seeded cells (from 10 to 10,000) for all cell lines (Fig. 2). Figure 2. In vitro bioluminescence of WRO, FTC-238, and TT2609-CO2 cell lines. Stable clones of WRO, FTC-238, and TT2609-CO2 cells expressing luciferase were serially diluted. Luciferin substrate was added to each well 10 min before imaging and the plate was imaged to obtain total flux (p/s) per cell line using IVIS Spectrum. The wells with media but not cells, or cells alone without luciferin, were included as controls. Note that the level of luciferase expression was proportional to the number of cells seeded. The results are expressed as mean ± s.e.m. of two independent experiments with three parallels. Tumorigenesis analysis using an orthotopic mouse model of thyroid cancer in immunodeficient mice To explore whether the FTC cell lines could initiate tumors in vivo, we used an orthotopic thyroid cancer model with weekly live imaging in immunodeficient mice to assess tumor initiation and progression (1). In this model, 5 × 105 luciferase-expressing WRO, FTC-238, or TT2609-CO2 cells were orthotopically injected into the right thyroid of 8-week-old female NSG mice (n = 2 per cell line). Tumor initiation and growth were monitored weekly via bioluminescent live imaging using IVIS. Notably, WRO-derived tumors were more aggressive and invasive and grew two to four times faster than tumors arising from the FTC-238 and TT2609-CO2 cell lines. Tumor volume, assessed as total flux (photon counts per second), revealed that mice injected with WRO cells rapidly developed primary tumors within 7 days of injection. These mice survived a total of 14–18 days, whereas mice injected with FTC-238 cells survived a total of 27–29 days. In contrast, mice injected with TT2609-CO2 cells experienced significantly delayed onset of tumors and the tumors grew more slowly than those arising from WRO or FTC-238 cells. These mice survived a total of 57–70 days (Fig. 3). Figure 3. WRO-derived tumors grew faster than FTC-238 and TT2609-CO2-induced tumors in a murine orthotopic model. Luciferase-tagged WRO, FTC-238, or TT2609-CO2 cells were injected into the thyroids of NSG mice, and tumor initiation and growth were monitored via bioluminescent live imaging weekly using IVIS Spectrum. Bioluminescent tumor signals were quantified as total flux (p/s) using Living image software and plotted against the days since tumor cell injection. The results are expressed as mean ± s.e.m. (n = 2 per group). An important feature of the orthotopic thyroid cancer model, which recapitulates advanced stage thyroid cancers, is that cancer cells are directly injected into the thyroid. As a result, aggressive tumor invasion occurs in nearby vital organs such as the trachea and esophagus. The subsequent compression of these organs leads to significant weight loss: a phenomenon we confirmed here with all three FTC cell lines (Fig. 4). For humane reasons, animals were killed when they had lost up to 20% of their body weight, an endpoint that was often accompanied by signs of unresponsiveness and failure to eat or drink. According to these criteria, one of the WRO-injected mice was killed 14 days after tumor cell injection, while the other mouse survived a total of 18 days (Fig. 4). FTC-238-injected mice were killed at day 28, but mice injected with TT2609-CO2 cells experienced a delayed onset of tumors. These mice continued to gain weight until day 49, when they began losing weight. One of these animals experienced a rapid decline in body weight beginning at day 56, while the other mouse survived a total of 70 days (Fig. 4). Figure 4. Time course of body weight and in vivo bioluminescence. Upper panel shows body weights of mice injected with WRO, FTC-238, or TT2609-CO2 cell lines. Lower panel shows primary thyroid tumors resulting from injection of WRO, FTC-238, or TT2609-CO2 cells in the orthotopic models as visualized by bioluminescent IVIS imaging. All mice (n = 2 per group) were reimaged using the same setting of the IVIS throughout the study. Postmortem histological analysis showed that all mice receiving WRO, FTC-238, and TT2609-CO2 cells developed thyroid tumors, with subsequent invasion into the trachea, esophagus, and surrounding muscles. Notably, mice injected with WRO cells also exhibited tumor cell invasion in their lung tissue, but mice injected with FTC-238 or TT2609-CO2 cells did not (Fig. 5). These observations demonstrate that WRO-derived tumors grow two to four times faster than tumors arising from FTC-238 and TT2609-CO2 cell lines in our orthotopic model, and show that WRO is the only cell line examined that can invade the lung and develop micrometastases under these conditions. Figure 5. Representative histology images of thyroid tumors and lung metastases in a murine orthotopic model. H&E stained sections of all mice orthotopically injected with WRO, FTC-238, or TT2609-CO2 cell lines show the primary tumor with local invasion to trachea and esophagus (A,D,F). The middle panel shows higher magnification (200x) of tumors from each cell line (B,E,G). Note that mice injected with WRO also developed a lung micrometastasis whereas mice injected with FTC-238 or TT2609-CO2 cells did not ©. Scale bar, 100 mm. Lung metastasis of FTC cell lines in a tail vein injection model To ensure that our orthotopic model did not underestimate the metastatic potential of FTC cell lines, perhaps because the mice may have reached the point of humane killing before lung metastasis, we established an artificial metastasis model by directly injecting tumor cells intravenously into the mouse tail vein. This technique allows us to more clearly evaluate the metastatic potential of each cell line. In this model, 5 × 105 cells were injected into the lateral tail vein. IVIS imaging showed that all mice injected with WRO cells exhibited rapid and aggressive growth of bilateral lung metastases (n = 2). In contrast, all mice injected with FTC-238 or TT2609-CO2 cells remained free of tumors 35 days after injection (Fig. 6A). Additional H&E staining of the isolated lung tissues of the mice injected with WRO cells confirmed tumor cells metastasized to the alveoli of the lung, and that those mice injected with FTC-238 or TT2609-CO2 cells did not develop lung metastasis (Fig. 6B). Figure 6. WRO cells develop lung metastases in the tail vein injection model. Lung metastases were experimentally induced by tail vein injection of 5 × 105 cells. (A) Representative bioluminescent images of mice (n = 2 per group) injected with WRO, FTC-238, or TT2609-CO2 cells taken at day 35. (B) H&E stained sections revealed tumor cells metastasized to the alveoli of the lung in mice injected with WRO cell line. Left panel, 50x magnification; right panel, 200x magnification. Scale bar, 100 mm. Discussion The aim of this study was to assess a panel of three human FTC cell lines for their tumorigenic and metastatic potential in xenotransplantation assays in immunodeficient mice. Our data show that all three human FTC cell lines form tumors in immunodeficient NSG mice with a 100% take rate. In addition, the three cell lines display highly varied in vivo tumor growth trends; WRO was the most aggressive cell line in both our orthotopic and tail vein injection models. For many years, investigators have relied on immunodeficient mice such as NOD/SCID mice or athymic nude mice for basic and preclinical research because they do not reject tumor cells transplanted from humans or other species. Athymic mice have no functional T cells. NOD/SCID mice have both T-cell and B-cell deficiencies, but they retain cytotoxic activity due to NK cells and macrophages. As a result, they may not produce high take rates in xenograft studies. A recent report evaluating primary thyroid tumor growth efficiency following orthotopic implantation and intracardiac injection in athymic nude mice has shown a highly variable take rate, ranging from 0 to 100% in 11 thyroid cancer cell lines (4). Such variability raises a concern as to whether the athymic nude mice can be used to adequately evaluate the tumorigenic potential of a given thyroid cancer cell line and whether they are a good choice for preclinical drug testing. In this study, we used NSG mice to evaluate the tumorigenic potential of three human FTC cell lines. This NSG mouse strain lacks mature T cells and B cells and is also deficient for Il2, Il4, Il7, and Il12, which are required for NK cells and innate immunity responses (16). Therefore these mice are more severely impaired than the commonly used NOD/SCID or athymic nude mice and they represent a gold standard for cancer research due to their superior xenografting capability (17, 18). We have previously used the orthotopic and tail vein injection models in NSG mice to study the anaplastic thyroid cancer cell (ATC) lines THJ-11T and THJ-16T. In these studies, we achieved a 100% tumor take rate (11, 13). We have also used NSG mice to confirm the presence of a rare subpopulation of cancer stem cells in the ATC cell lines (14). In this study, all three human FTC cell lines developed tumors in the orthotopic model with a 100% take rate. Recently, Zhang and coworkers used the same mouse strain and achieved a 100% take rate for all eight thyroid cancer cell lines examined (19). Together, these studies emphasize the importance of mouse strain choice in evaluating tumorigenic potential of a given thyroid cancer cell line in order to increase the efficiency and reduce the costs of preclinical drug testing. In this study, we found that WRO cell line, which has a BRAFV600E mutation, is the most tumorigenic of the three human FTC cell lines tested. Activating mutations in BRAF, a 766-amino acid serine/threonine-specific protein kinase, have been detected in a wide range of human malignancies. The presence of BRAF mutations is significantly associated with advanced thyroid cancers with metastasis and poor prognoses. Therefore our data support the proposition that the BRAFV600E mutation renders the WRO cell line particularly aggressive. Because the BRAFV600E mutation, which is commonly associated with high-risk clinicopathologic characteristics of patients with PTC, is less common in FTC (8), the origin of the WRO cell line has been questioned. However, several investigators have demonstrated with STR profiling analyses that the WRO cell line is not contaminated (7, 15). In addition to driver mutations, a number of other mechanisms may also contribute to the aggressiveness of FTC cell lines. Schmutzler and coworkers reported an inhibitory effect of all-trans retinoic acid (RA) on the growth of FTC-133 and FTC-238 cell lines (20). Both FTC-133 and FTC-238 were derived from the same patient, a 42-year-old male with FTC. Specifically, FTC-133 was established from a primary tumor, whereas FTC-238 from a lung metastasis. In in vivo studies, pretreatment of FTC-133 and FTC-238 with all-trans RA resulted in a reduced tumor growth after xenotransplantation onto adult athymic nude rats (20). The authors suggested that tumor-specific RA receptors might contribute to the pathogenesis and tumor progression of human thyroid cancer cell lines. In another study, Hoffmann and coworkers demonstrated a major differential expression pattern of integrin receptor molecules (IRM) in differentiated (FTC and PTC) and undifferentiated (ATC) thyroid cancer cell lines. Through cell substratum adhesion assays, they noted that the metastatic thyroid cancer cell lines express high levels of integrins α1-6 and β1 and show strong attachment to extracellular matrix (ECM) proteins including collagen I, collagen IV, laminin, and fibronectin (21). It is conceivable that these integrins may contribute to the development of metastatic disease. In summary, our work demonstrates that these human FTC cell lines display highly varied tumorigenic potential, with WRO being the most aggressive cell line in both in vivo models. This new information will be useful for selecting cell line for future pre-clinical drug testing and for basic research aiming at delineating the molecular basis for thyroid cancer pathogenesis. Declaration of interest The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported. Funding This work was supported in part by the National Institutes of Health Grant DK068057 (to R.Y.L.). Acknowledgements The authors thank Dr John Copland for providing the WRO and TT2609-CO2 cell lines and Dr Yasuhiro Ikeda for providing the pSIN-luc vector. They also thank Dr Karoly Toth and the Saint Louis University Animal Imaging Core Facility for assistance with IVIS imaging. ==== Refs References 1 Sewell W Reeb A Lin RY An orthotopic mouse model of anaplastic thyroid carcinoma . Journal of Visualized Experiments 2013 74 e50097 10.3791/50097 2 Antonello ZA Nucera C Orthotopic mouse models for the preclinical and translational study of targeted therapies against metastatic human thyroid carcinoma with BRAF(V600E) or wild-type BRAF . Oncogene 33 5397 –5404 . 10.1038/onc.2013.544 24362526 3 Nucera C Nehs MA Mekel M Zhang X Hodin R Lawler J Nose V Parangi S A novel orthotopic mouse model of human anaplastic thyroid carcinoma . Thyroid 2009 19 1077 –1084 . 10.1089/thy.2009.0055 19772429 4 Morrison JA Pike LA Lund G Zhou Q Kessler BE Bauerle KT Sams SB Haugen BR Schweppe RE Characterization of thyroid cancer cell lines in murine orthotopic and intracardiac metastasis models . Hormones & Cancer 2015 6 87 –99 . 10.1007/s12672-015-0219-0 25800363 5 Hatakeyama S Yamamoto H Ohyama C Tumor formation assays . Methods in Enzymology 2010 479 397 –411 .20816179 6 Geldof AA Versteegh LRT van Mourik JC Rooimans MA Arwert F Hermsen MA Schadee-Eestermans IL van Dongen GA van der Valk P van der Clement EHP, Clonally related but phenotypically divergent human cancer cell lines derived from a single follicular thyroid cancer recurrence (TT2609) . Thyroid 2001 11 909 –917 . 10.1089/105072501753210966 11716037 7 Saiselet M Floor S Tarabichi M Dom G Hebrant A van Staveren WC Maenhaut C Thyroid cancer cell lines: an overview . Frontiers in Endocrinology 2012 3 133 .23162534 8 Namba H Nakashima M Hayashi T Hayashida N Maeda S Rogounovitch TI Ohtsuru A Saenko VA Kanematsu T Yamashita S Clinical implication of hot spot BRAF mutation, V599E, in papillary thyroid cancers . Journal of Clinical Endocrinology & Metabolism 2003 88 4393 –4397 . 10.1210/jc.2003-030305 12970315 9 Boldarine VT Maciel RM Guimaraes GS Nakabashi CC Camacho CP Andreoni DM Mamone Mda C Ikejiri ES Kasamatsu TS Crispim F Hojaij FC, Development of a sensitive and specific quantitative reverse transcription-polymerase chain reaction assay for blood thyroglobulin messenger ribonucleic acid in the follow-up of patients with differentiated thyroid carcinoma . Journal of Clinical Endocrinology and Metabolism 2010 95 1726 –1733 . 10.1210/jc.2009-1354 20173019 10 Presta I Arturi F Ferretti E Mattei T Scarpelli D Tosi E Scipioni A Celano M Gulino A Filetti S, Recovery of NIS expression in thyroid cancer cells by overexpression of Pax8 gene . BMC Cancer 2005 5 80 10.1186/1471-2407-5-80 16029487 11 Li W Reeb AN Sewell WA Elhomsy G Lin RY Phenotypic characterization of metastatic anaplastic thyroid cancer stem cells . PloS ONE 2013 8 e65095 10.1371/journal.pone.0065095 12 Hasegawa K Nakamura T Harvey M Ikeda Y Oberg A Figini M Canevari S Hartmann LC Peng KW The use of a tropism-modified measles virus in folate receptor-targeted virotherapy of ovarian cancer . Clinical Cancer Research 2006 12 6170 –6178 .17062694 13 Reeb AN Li W Sewell W Marlow LA Tun HW Smallridge RC Copland JA Spradling K Chernock R Lin RY S100A8 is a novel therapeutic target for anaplastic thyroid carcinoma . Journal of Clinical Endocrinology and Metabolism 2014 100 E232 –E242 . 10.1210/jc.2014-2988 25423568 14 Reeb AN Li W Lin RY Bioluminescent human thyrospheres allow noninvasive detection of anaplastic thyroid cancer growth and metastases in vivo . Thyroid : Official Journal of the American Thyroid Association 2014 24 1134 –1138 . 10.1089/thy.2014.0014 24684373 15 Schweppe RE Klopper JP Korch C Pugazhenthi U Benezra M Knauf JA Fagin JA Marlow LA Copland JA Smallridge RC, Deoxyribonucleic acid profiling analysis of 40 human thyroid cancer cell lines reveals cross-contamination resulting in cell line redundancy and misidentification . Journal of Clinical Endocrinology and Metabolism 2008 93 4331 –4341 . 10.1210/jc.2008-1102 18713817 16 Quintana E Shackleton M Sabel MS Fullen DR Johnson TM Morrison SJ Efficient tumour formation by single human melanoma cells . Nature 2008 456 593 –598 . 10.1038/nature07567 19052619 17 Quintana E Piskounova E Shackleton M Weinberg D Eskiocak U Fullen DR Johnson TM Morrison SJ Human melanoma metastasis in NSG mice correlates with clinical outcome in patients . Science Translational Medicine 2012 4 159ra149 10.1126/scitranslmed.3004599 18 Drake AC Khoury M Leskov I Iliopoulou BP Fragoso M Lodish H Chen J Human CD34 + CD133 + hematopoietic stem cells cultured with growth factors including Angptl5 efficiently engraft adult NOD-SCID Il2rgamma-/- (NSG) mice . PloS ONE 2011 6 e18382 10.1371/journal.pone.0018382 21559522 19 Zhang L Gaskins K Yu Z Xiong Y Merino MJ Kebebew E An in vivo mouse model of metastatic human thyroid cancer . Thyroid 2014 24 695 –704 . 10.1089/thy.2013.0149 24262022 20 Schmutzler C Hoang-Vu C Ruger B Kohrle J Human thyroid carcinoma cell lines show different retinoic acid receptor repertoires and retinoid responses . European Journal of Endocrinology 2004 150 547 –556 . 10.1530/eje.0.1500547 15080786 21 Hoffmann S Maschuw K Hassan I Reckzeh B Wunderlich A Lingelbach S Zielke A Differential pattern of integrin receptor expression in differentiated and anaplastic thyroid cancer cell lines . Thyroid 2005 15 1011 –1020 . 10.1089/thy.2005.15.1011 16187909

PMC005xxxxxx/PMC5002965.txt

==== Front Endocr ConnectEndocr ConnectECEndocrine Connections2049-3614Bioscientifica Ltd Bristol 26864459EC16000410.1530/EC-16-0004ResearchMild cold effects on hunger, food intake, satiety and skin temperature in humans The response to mild cold in humansM Langeveld and othersLangeveld M 1*Tan C Y 1*Soeters M R 1*Virtue S 1Ambler G K 2Watson L P E 13Murgatroyd P R 13Chatterjee V K 1Vidal-Puig A 11 University of Cambridge Metabolic Research LaboratoriesWellcome Trust-MRC, Institute of Metabolic Science, Addenbrookes Hospital, Cambridge, UK2 Cambridge Vascular UnitAddenbrookes Hospital, Hills Road, Cambridge, UK3 NIHR/Wellcome Trust Clinical Research FacilityAddenbrookes Hospital, Cambridge, UKCorrespondence should be addressed to M R Soeters; Email: mrsoeters@gmail.com* (M Langeveld, C Y Tan and M R Soeters contributed equally to this work) 3 2016 01 3 2016 5 2 65 73 4 2 2016 10 2 2016 © 2016 The authors2016The authors This work is licensed under a Creative Commons Attribution 4.0 International License.Background Mild cold exposure increases energy expenditure and can influence energy balance, but at the same time it does not increase appetite and energy intake. Objective To quantify dermal insulative cold response, we assessed thermal comfort and skin temperatures changes by infrared thermography. Methods We exposed healthy volunteers to either a single episode of environmental mild cold or thermoneutrality. We measured hunger sensation and actual free food intake. After a thermoneutral overnight stay, five males and five females were exposed to either 18°C (mild cold) or 24°C (thermoneutrality) for 2.5 h. Metabolic rate, vital signs, skin temperature, blood biochemistry, cold and hunger scores were measured at baseline and for every 30 min during the temperature intervention. This was followed by an ad libitum meal to obtain the actual desired energy intake after cold exposure. Results We could replicate the cold-induced increase in REE. But no differences were detected in hunger, food intake, or satiety after mild cold exposure compared with thermoneutrality. After long-term cold exposure, high cold sensation scores were reported, which were negatively correlated with thermogenesis. Skin temperature in the sternal area was tightly correlated with the increase in energy expenditure. Conclusions It is concluded that short-term mild cold exposure increases energy expenditure without changes in food intake. Mild cold exposure resulted in significant thermal discomfort, which was negatively correlated with the increase in energy expenditure. Moreover, there is a great between-subject variability in cold response. These data provide further insights on cold exposure as an anti-obesity measure. Keywords Coldthermogenesishunger ==== Body Introduction At first sight obesity may appear as a condition that is easy to treat by either decreasing energy intake and/or increasing the energy expenditure. In practice, long-term weight loss is very difficult to achieve. Since strategies that reduce energy intake fail in the majority of patients, increasing energy expenditure seems to be an attractive alternative. Energy dissipating drugs (e.g., thyroid hormone, ephedrine, dinitrophenol) have been used successfully to decrease body weight, but their use was discontinued because of unacceptable cardiovascular side effects (1, 2, 3). Exercise may be a healthier approach to increase energy expenditure, but the amount of exercise needed to significantly influence energy balance, as well as the accompanying increase in appetite, makes it an ineffective strategy for long-term body weight reduction (4). Exposure to cold increases energy expenditure and is partly mediated by the activation of brown adipose tissue (BAT). Non-shivering thermogenesis (NST) is the increase in energy expenditure resulting from exposure to temperatures below the thermoneutral zone, but above the temperature threshold for shivering. By definition, no physiological mechanisms for temperature regulation are active at thermoneutrality and therefore no energy is spent on temperature maintenance. For naked humans, the thermoneutral zone is 27 ± 2°C (5) and for lightly clothed humans it lies around 22–24°C, depending on the insulative properties of the clothing (6). As described above, cold exposure induces physiological changes. More importantly, mild cold exposure may have a better adherence compared with profound cold exposure when used as an anti-obesity strategy to increase the metabolic rate. The key question is whether increasing energy expenditure through mild cold exposure is accompanied by an increase in appetite and food intake. Cold exposure is known to increase food and energy intake in a wide range of animal species, e.g., piglets, rats and birds (7, 8, 9), but not in humans. Besides the increase in energy expenditure, cold exposure may also trigger an insulative response. Vasoconstriction, mediated via activation of alpha-adrenergic receptors, limits heat loss via the skin. Interestingly, the vasoconstrictive response is highly variable between individuals and was shown to correlate negatively with the magnitude of NST in one study (10). Skin temperature changes during cold exposure may thus be a predictor of the metabolic response to cold exposure. This may also be the case for changes in the temperature of the skin overlying the supraclavicular BAT depot, as suggested by two reports using infrared thermography to measure skin temperature in the supraclavicular region during a cold challenge (11, 12). Moreover, it is the reduction in skin temperature during cold that is mediated by vasoconstriction, which is perceived as uncomfortable. In this study, we investigated the response to mild cold in healthy humans for this may be an attractive weight management strategy. More importantly, we focussed on changes in energy expenditure, food intake including appetite and satiety, and dermal temperature. Subjects and methods Subjects Healthy volunteers were recruited through local advertisements in the East Anglian region of the United Kingdom. We recruited five lean males and five lean females, non-smokers, aged between 22 and 60 years, who had no known medical conditions and were not taking any medications or supplements. To minimise seasonal variation of NST, which is known to exist, subjects were studied between April 2012 and September 2012 (13). All subjects provided written informed consent and the study conformed to the standards set by the latest revision of the Declaration of Helsinki. The study received approval from the Cambridge Central East of England Research Ethics Committee. Study outline The outline of the study design is depicted in Fig. 1. The subjects were studied twice during two days, about two weeks apart, one of the days the subjects were tested under thermoneutrality and the other day under mild cold exposure, they were blinded to the setting and tests were performed in a random order. The subjects were asked to refrain from strenuous physical activity, alcohol and caffeine for 24 h before their visit. Each participant arrived at the Clinical Research Facility around 16:00 h on day 0 and remained until 14:00 h on day 1. Height, weight and body composition (DXA (GE Lunar Prodigy GE Healthcare, Madison, WI, USA; software version 12.2)) were measured. At 18:00 h, a standardised dinner was served. The energy content of the meal was 1/3 of a participant's daily requirements estimated from predicted resting metabolic rate, using the Schofield equation, multiplied by an activity factor of 1.35. Meal composition was 30–35% fat, 12–15% protein and 50–55% carbohydrate by energy. The participants retired to bed in the temperature controlled room (24°C) at 23:00 h and were provided with standardised light clothing and bedding. The temperature controlled room is a habitual room including a desk, television, computer, sink and toilet. The participant was woken the next morning at 07:00 h and stayed in bed in a semi-supine position (upper part of the bed at 45°C) without bedding. All participants were asked to remain awake and inactive. To enable thermal imaging, male subjects had a bare torso and women wore a boob tube for the remainder of the experiment. Baseline indirect calorimetry, vital signs, cold and hunger scores and blood tests were taken at 7:30 h. Next, the subjects either stayed in this room at 24°C or were moved to the mild cold room (18°C). Subsequently, thermal imaging, vital signs, indirect calorimetry, cold and hunger scores and blood tests were repeated every 30 min (Fig. 1) during 2.5 h. In between the measurements, the subjects were allowed to read or watch TV but did not leave the bed except for toileting. Blood samples were drawn via a large indwelling venous catheter without using a heated hand box or blanket to prevent local warming. Afterwards, the universal eating monitor was used to assess the speed of eating ad libitum meal and other parameters related to appetite and food intake. Figure 1 Study design. Black bars represent measurements including indirect calorimetry, cold and hunger scores, vital signs and blood test. UEM, universal eating monitor. Indirect calorimetry REE was measured by ventilated canopy respiratory gas exchange (GEM; GEMNutrition, Daresbury, UK) in a supine position. The measurements were recorded during 12-min intervals for every 30 min. Energy expenditure was calculated from the macronutrient respiratory quotients and energy equivalents of oxygen published by Elia and Livesey (14). Cold and hunger scores The participants were asked to rate the sensation of cold of the whole body and hands separately on a 1–10 scale, with ratings as follows: 1 was rated as ‘not at all cold’ and 10 was the ‘coldest one had ever felt’. Similarly for the degree of hunger, with ratings as: 1 for ‘not hungry at all’, and 10 was rated as ‘the most hungry one had ever felt’. Blood biochemistry Glucose was measured by Hexokinase method on a Siemens Dimension RXL AutoAnalyser. Reagents and Calibrators were purchased from Siemens. Free fatty acids were measured using Roche Free Fatty Acid Kit. This assay was modified to run in a microtitre plate format. Thyroid-stimulating hormone (TSH), free thyroxin (fT4) and iodothyronin (T3) were measured by time-resolved fluorescence immunoassay on an AutoDELFIA analyser (Perkin Elmer) using kits from Perkin Elmer. Cortisol level was measured by fluorescence immunoassay on the Siemens Centaur Autoanalyser. A minimum of two quality control samples were run in each assay. Universal Eating Monitors (UEMs) The UEM (The Sussex Meal Patterning System) was used. The subjects ate an homogenous test meal (e.g., pasta) containing normal energy percent ratios (∼30% carbohydrates, ∼30% protein and ∼40% fat). Intake of test meal was continuously monitored using the UEM equipment (15). Herein, food is served and eaten from a plate placed on weighing scales connected to a computer. The generated intake data contained the amount eaten and the seconds spent on eating. The monitors allow automated combinations of appetite ratings by the visual analogue scales and intake data. The VAS scales rate feelings of hunger, sickness, fullness and desire to eat on a 0–100 scale (16). Thermography Skin temperature images were obtained using a ThermaCam 3000, and images were analysed by ThermaCAM Researcher Pro 2.9 software (both FLIR systems, Boston, United States). The camera settings were temperature dependent: at 24°C; emissivity: 0.98, humidity: 45%, distance: 1.2 m, external and reflected temperature: 24°C, at 18°C, emissivity: 0.98, humidity: 40%, distance: 1.2 m, and external and reflected temperature: 18°C. Two skin regions were defined: first the supraclavicular area (bordered by the acromioclavicular joint, sternoclavicular joint, and the sternocleidomastoid trapezoid angle) and secondly, the sternal area (the top 10 cm above of the sternum). For recognition of these anatomical landmarks on the thermal images, we placed metal markers on the skin. At each time point, per area the average of three images was taken for analysis. Statistical analyses All analyses were performed using SPSS 21. Time series data were analysed using repeated measures ANOVA. Each ANOVA model was built using ‘time’ as within-subject effect, ‘temperature’ as independent factor, and ‘time*temperature’ as the interacting term. A significant ‘time*temperature’ effect is interpreted as a significant effect of mild cold exposure on the rate of change over time. Each term in the ANOVA model was analysed for sphericity (Mauchly’s Test) and if found to be violated, within-subject effects was determined by the Greenhouse-Geisser test. For all statistical test, a P value of <0.05 was considered to be significant. All paired data were analysed by Student's t-test. The correlations were assessed using Pearson's test. Data are presented as mean ± S.D. Results Metabolic response to mild cold exposure We studied 10 Caucasian healthy subjects, 5 males and 5 females, age ranging from 22 to 60 years old, BMI ranging from 20.8 to 24.9 kg/m2. The characteristics of the subjects are included in Table 1. Mild cold exposure significantly increased energy expenditure without visible shivering compared with thermoneutrality (repeated measurement ANOVA for cold effect P = 0.01). Over 150 min of exposure to 18°C, a total of 48 ± 14 kJ (range 13–127 kJ) was expended above the baseline energy expenditure at 24°C (Fig. 2A). Respiratory exchange ratio (RER) dropped during the experiment under both conditions (repeated measurement ANOVA for the effect of time P < 0.01). There was no significant effect of temperature on RER (Fig. 2B; repeated measurement ANOVA for the effect of temperature P = 0.195). Figure 2 (A) Cumulative energy expended above basal metabolic rate over 150 min of thermal challenge. *P < 0.05 compared with 24°C. (B) Change in respiratory exchange ratio (RER) compared with baseline. Table 1 Characteristics of subject Males Female Age (years) 44.7 ± 5.2 33.7 ± 6.9 BMR (J/min) 4445.5 ± 225.9 3808 ± 196.6 Height (m) 1.76 ± 0.03 1.65 ± 0.04 Weight (kg) 69.5 ± 3.3 62.6 ± 4.1 BMI (kg/m2) 22.4 ± 0.8 22.9 ± 0.9 Fat (kg) 12.8 ± 2.7 20.2 ± 2.4 Lean (kg) 53.1 ± 1.8 39 ± 2.2 FFM (kg) 55.8 ± 1.9 41.5 ± 2.4 BMR, basal metabolic rate; BMI, body mass index; FFM, fat free mass. Vital signs Heart rate remained stable at thermoneutrality at 56.7 ± 4.2 (T = 0) to 58.1 ± 4.2 (T = 150 min) beats per minute (bpm). Heart rate decreased in response to mild cold exposure from an average of 59.5 ± 4.4 (t = 0) to 56.7 ± 4.5 bpm (t = 150 min) (Supplementary Figure 1A, see section on supplementary data given at the end of this article; repeated measurement ANOVA for the effect of time*temperature P = 0.025). Systolic blood pressure remained stable between 109 ± 2 (T = 0) and 113 ± 2 mmHg (T = 150 min) at thermoneutrality and increased from 107 ± 2 (T = 0) to 120 ± 4 mmHg (T = 150 min) during mild cold exposure (Fig. 3; repeated measures ANOVA for the effect of time*temperature P = 0.007). Diastolic blood pressure remained stable during both the stay at thermoneutrality and at mild cold exposure (Fig. 3). Figure 3 (A) Average heart rate, (B) systolic blood pressure and (C) diastolic blood pressure over 150 min of thermal challenge. *P < 0.05 for 24°C vs 18°C. Biochemistry Plasma concentrations of glucose and fT4 increased similarly, and plasma cortisol, TSH and T3 concentrations decreased similarly under both thermal conditions (Fig. 4). The only biochemical parameter that responded to a difference in ambient temperature was the plasma non-esterified fatty acid (NEFA) levels (Fig. 4B). NEFA levels increased under both conditions, but this increase was more at mild cold exposure (211 ± 62 μmol/L) compared with thermoneutrality (132 ± 59 μmol/L) (repeated measures ANOVA for the effect of temperature P < 0.001). Figure 4 (A) Plasma glucose, (B) non-esterified free fatty acid (NEFA), (C) cortisol, (D) thyroid-stimulating hormone (TSH), (4) free thyroxine (FT4) (E) and free tri-iodothyronine (FT3) (F) over 150 min of exposure to 24°C and 18°C. Cold sensation At thermoneutrality, the score for whole body cold sensation remained stable between 2.1 ± 0.5 and 2.3 ± 0.5, whereas during mild cold exposure the score significantly increased from 3.0 ± 0.5 to 7.2 ± 0.3 (repeated measures ANOVA for the effect of temperature P < 0.001) (Fig. 5A). The same pattern was observed also for hand, at thermoneutrality no significant change, but during mild cold exposure the cold score increased significantly from 3.0 ± 0.6 to 7.5 ± 0.5 (Fig. 5B, repeated measures ANOVA for the effect of temperature P < 0.001). There was a significant negative correlation between the cold score for hands after 150 min of mild cold exposure and the cumulative increase in energy expenditure above baseline during this time (Fig. 5C, r2 = 0.481, Pearson P < 0.001). There was no correlation between the cold score for whole body at 150 min and the cumulative increase in energy expenditure in response to mild cold (r2 = 0.000). Figure 5 (A) Score for the perception of cold (cold score) for whole body and (B) hands. (C) Correlation between cold score for hands and EE over baseline after 150 min of mild cold exposure. Hunger, food intake and satiety Feelings of hunger increased during both situations over 150 min. A trend towards a higher hunger score during cold was observed (Fig. 6A, repeated measurement ANOVA for the effect of time P = 0.021, for the effect of temperature P = 0.064). Before the meal, feelings of fullness, hunger, sickness, and desire to eat were similar after exposure to both thermal conditions (Fig. 6B). During the meal, the same amount of food was consumed after thermoneutrality and mild cold exposure (2740 ± 567 vs 2878 ± 492 kJ paired t-test P = 0.69) (Fig. 6C). There were no differences in the time spent eating during both situations: thermoneutrality (714 ± 124 s) versus cold (778 ± 115 s) (paired t-test P = 0.14) (Fig. 6D). There was no correlation between the amount of food consumed and basal metabolic rate, and between the amount of food consumed and the increase in energy spent during 150 min of mild cold exposure (r2 = 0.076, P = 0.271 and r2 = 0.00, P = 0.962 Pearson test). Feelings of fullness, hunger, and desire to eat after the test meal were not different after either situation: cold versus thermoneutrality (Fig. 6B). Feeling of sickness was significantly greater after mild cold exposure (Fig. 6B). Figure 6 Score for perception of hunger over 150 min of thermal challenge. (A) Visual analogue scale for hunger and satiety before and after UEM test meal (B) Amount of food consumed (KJ) (C) and time spent eating (seconds) (D) during UEM test meal. *P < 0.05 for 24°C vs 18°C. Skin temperature changes assessed by thermography During both situations (thermoneutrality and cold exposure), the supraclavicular skin temperature was higher compared with the sternal skin area (repeated measurement for the effect of location ANOVA P < 0.01). At 24°C skin temperature in both areas remained stable (Fig. 7A). In response to mild cold exposure, skin temperature dropped in both areas during the first 30 min and stayed stable thereafter. This temperature drop was greater in the sternal area compared with the supraclavicular area (Fig. 7A). The temperature difference between both areas increased in response to mild cold exposure and remained stable at thermoneutrality (Fig. 7B; repeated measurement ANOVA for the effect of temperature*time interaction P < 0.01). There was a tight positive correlation between the skin temperature of the sternal area and the increase in energy expenditure (Fig. 7D; r2 = 0.787, P = 0.001, Pearson test). Supraclavicular temperature was not correlated with the increase in energy expenditure (Fig. 7E; r2 = 0.286, P = 0.103, Pearson test). Figure 7 Surface temperature over the supraclavicular and sternal area as measured by thermal imaging at 24°C and 18°C (A) over 150 min of thermal challenge. (B) Temperature difference between supraclavicular area and sternal area. (C) Correlation between cumulative energy expended (EE) over baseline after 150 min of mild cold exposure and the change in temperature difference between supraclavicular and sternal areas after 150 min. (D) Correlation between EE over baseline after 150 min of mild cold exposure and sternal temperature after 150 min. (E) Correlation between EE over baseline after 150 min of mild cold exposure and supraclavicular temperature after 150 min. Discussion In this paper, we investigated the response to mild cold in healthy humans with a focus on changes in energy expenditure, food intake, and dermal temperature. As described in previous studies, exposure to mild cold induced a small but significant increase in energy expenditure. We showed that, at least in short-term, energy intake does not increase since two and a half hours of mild cold exposure had no influence on the amount of food eaten and time spent eating. During cold exposure, feelings of hunger showed a trend to increase to a level above what was observed during thermoneutrality. Once out of the cold, the difference disappeared and no differences could be found in pre-meal feelings of hunger, desire to eat, fullness or sickness after exposure to the two different thermal conditions. We cannot exclude the possibility that excess energy expenditure in cold is compensated by increased food-intake later. Whether prolonged mild cold exposure, with meals consumed in the cold, would increase energy intake remains to be determined. Historical data obtained under harsher thermal conditions show a negative correlation between outdoor temperature (ranging from −30 to +35°C) and food intake (9). To our knowledge, there are no data on the effect of prolonged mild cold exposure on hunger and food intake. Skin temperature falls in response to a drop in ambient temperature. In the studies published so far, it was found that the supraclavicular temperature either increased (12, 17) or decreased to a lesser extend compared with other body areas (11, 18). A recently published study has shown a positive correlation between the skin temperature in the supraclavicular region and clavicular BAT volume and activity during cold exposure (17). Based on these data, we proposed a hypothesis that the change in the temperature difference between the supraclavicular region and the sternal area would be positively correlated with the increase in energy expenditure in response to mild cold exposure. This was not the case, nor was the supraclavicular temperature in itself related to the metabolic response to mild cold. We conclude that measurements of skin temperature in the areas overlying the superficial BAT depots are not helpful in predicting the metabolic responses to cold. Relative mild cold exposure (18°C) resulted in significant thermal discomfort but no visible shivering in our study (Fig. 4A and B). This thermal discomfort is also likely to be perceived in daily life, since the average living room temperature in the UK has increased from approximately 18 to 21°C over the last three decades (18). Mild cold exposure is uncomfortable due to the perceived reduction in skin temperature, which is mediated by vasoconstriction. We not only show that a negative correlation between the metabolic and the vasoconstrictive response to mild cold exists, but also that the metabolic response to mild cold is negatively correlated with the cold score for hands. Therefore, the lower the metabolic response to mild cold exposure, the colder one feels. At extreme ends, individuals could be classified as ‘vasoconstrictors’; those with a strong insulative response and a low metabolic response to mild cold exposure and ‘metabolisers’; those with a high metabolic response that allows a relatively high rates of peripheral heat loss by maintaining a higher skin temperature. Hypothetically, vasoconstrictors would be at greater risk to develop overweight or obesity, since they will take behavioural measures to avoid the negative sensation elicited by mild exposure and spent less energy on thermoregulation compared with metabolisers. As described earlier, mild cold exposure resulted in an increase in plasma free fatty acid (FFA) concentrations (e.g., 15, 19, 20). Studies on more extreme cold exposure have shown an increased rate of lipolysis in humans (21), making this the most likely source of increased FFAs during mild cold exposure. Mild cold exposure in humans increases systolic blood pressure (e.g., 15, 22, current study) and may increase LDL cholesterol concentrations (23, 24). Taken together, short-term mild cold exposure results in unfavourable changes in several cardiovascular risk factors. Lower temperatures are known to increase the incidence of cardiovascular events such as myocardial infarction and stroke, which have an increased incidence rate in winter, even in countries with milder climates (25). Whether the changes in cardiovascular risk factors that occur when lowering temperature in the range from around 22°C to around 16–18°C also results in a higher cardiovascular disease incidence remains to be established. We aimed to standardise the experimental circumstances as much as possible. Therefore, subjects were admitted the afternoon before the study and exposed to exact identical meals, temperatures and sleep time. However, some limitations remain. Due to the intensive study protocol we only included 10 subjects, which may have prevented us from finding smaller effects. This may also explain why we did not find differences between men and women (data not shown) although the latter both may have more BAT and insulative capacity (26, 27, 28). Also the time spent in the cold and the temperature in which the meal was consumed may have led to results that are not generalisable. In addition, we did not measure BAT FDG uptake, but our primary aim was to investigate skin temperature in relationship with energy expenditure. Finally, we included healthy non-obese subjects and our results may not apply to obese subjects who may have larger insulative capacity due to more subcutaneous adipose tissue. In conclusion, short-term mild cold exposure results in an increase in energy expenditure that is not directly compensated by an increase in energy intake and may thus be used to alter energy balance (29). The increase in energy expenditure is small, but if maintained throughout longer periods could be used to prevent weight gain or even promote modest weight loss. However, lower temperatures lead to thermal discomfort, especially in those with a low metabolic response to cold and may induce unfavourable cardiovascular and metabolic changes. Moreover, the response to cold is variable and not all subjects may show an increase in energy expenditure during cold exposure. Long-term studies measuring the effect of lowering ambient temperature on thermal comfort, body weight, adiposity and cardiovascular risk factors will have to establish whether mild cold exposure is an effective anti-obesity measure. Supplementary data This is linked to the online version of the paper at http://dx.doi.org/10.1530/EC-16-0004. Declaration of interest The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported. Funding The study was funded by NIHR, BRC Seed Fund, individual grants: ML and MS: Marie Curie Fellowship, CYT: Welcome Trust Fellowship, SV: MRC, BHF and BBSRC, AVP: BBSRC. Acknowledgements The authors thank Katie Bird, Liz Blower, Cathy Baker and many others of the Addenbrookes Clinical Research Facility Staff for their excellent assistance during the studies. ==== Refs References 1 Haller CA Benowitz NL. Adverse cardiovascular and central nervous system events associated with dietary supplements containing ephedra alkaloids . New England Journal of Medicine 2000 343 1833 –1838 . 10.1056/NEJM200012213432502 11117974 2 Yen M Burns EM. Toxicity of weight loss agents . Journal of Medical Toxicology 2012 8 145 –152 . 10.1007/s13181-012-0213-7 22351299 3 Comerma-Steffensen S Grann M Andersen CU Rungby J Simonsen U. Cardiovascular effects of current and future anti-obesity drugs . Current Vascular Pharmacology 2014 12 493 –504 . 10.2174/1570161112666140423223529 24846238 4 Thomas DM Bouchard C Church Slentz TC Kraus WE Redman LM Martin CK Silva AM Vossen M Westerterp K Heymsfield SB. Why do individuals not lose more weight from an exercise intervention at a defined dose? An energy balance analysis . Obesity Reviews 2012 13 835 –847 . 10.1111/j.1467-789X.2012.01012.x 22681398 5 Erikson H Andersen KL Scholander PF. The critical temperature in naked man . Acta physiologica Scandinavica 1956 37 35 –39 . 10.1111/j.1748-1716.1956.tb01339.x 13339450 6 Kingma B Frijns A van Marken Lichtenbelt W The thermoneutral zone: implications for metabolic studies . Frontiers in Bioscience 2012 4 1975 –1985 . 10.2741/E518 7 Macari M Dauncey MJ Ingram DL. Changes in food intake in response to alterations in the ambient temperature: modifications by previous thermal and nutritional experience . Pflugers Archiv 1983 396 231 –237 .6844126 8 Brobeck JR. Food intake as a mechanism of temperature regulation . Yale Journal of Biology & Medicine 1948 20 545 –552 .18872321 9 Blaxter K Energy metabolism in animals and man , p 204 –206 . Cambridge, UK : Cambridge University Press , 1989 . 10 Van Marken Lichtenbelt WD Schrauwen P van de Kerckhove S Westerterp-Plantenga MS. Individual variation in body temperature and energy expenditure in response to mild cold . American Journal of Physiology: Endocrinology & Metabolism 2002 282 E1077 –E1083 . 10.1152/ajpendo.00020.2001 11934673 11 Lee P Ho KKY Greenfield JR. Hot fat in a cool man: infrared thermography and brown adipose tissue . Diabetes, Obesity & Metabolism 2010 13 92 –93 . 10.1111/j.1463-1326.2010.01318.x 12 Symonds ME Henderson K Elvidge L Bosman C Sharkey D Perkins AC Budge H. Thermal imaging to assess age-related changes of skin temperature within the supraclavicular region co-locating with brown adipose tissue in healthy children . Journal of Pediatrics 2012 161 892 –898 . 10.1016/j.jpeds.2012.04.056 22677567 13 Nishimura T Motoi M Egashira Y Choi D Aoyagi K Watanuki S. Seasonal variation of non-shivering thermogenesis (NST) during mild cold exposure . Journal of Physiological Anthropology 2015 34 11 10.1186/s40101-015-0051-9 25858699 14 Elia M Livesey G. Energy expenditure and fuel selection in biological systems: the theory and practice of calculations based on indirect calorimetry and tracer methods . World Review of Nutrition & Dietetics 1992 70 68 –131 .1292242 15 Celi FS Brychta RJ Linderman JD Butler PW Alberobello AT Smith S Courville AB Lai EW Costello R Skarulis MC Minimal changes in environmental temperature result in a significant increase in energy expenditure and changes in the hormonal homeostasis in healthy adults . European Journal of Endocrinology 2010 163 863 –872 . 10.1530/EJE-10-0627 20826525 16 Dovey TM Clark-Carter D Boyland EJ Halford JCG. A guide to analysing Universal Eating Monitor data: assessing the impact of different analysis techniques . Physiology & Behavior 2009 96 78 –84 . 10.1016/j.physbeh.2008.08.016 18796309 17 Boon MR Leontine Bakker LEH van der Linden RAD Arias-Bouda LP Smit F Verberne HJ van Marken Lichtenbelt WD Jazet IM Rensen PCN. Supraclavicular skin temperature as a measure of 18F-FDG uptake by BAT in human subjects . PloS ONE 2014 9 e98822 10.1371/journal.pone.0098822 24922545 18 Yoneshiro T Aita S Matsushita M Kameya T Nakada K Kawai Saito M Brown adipose tissue, whole-body energy expenditure, and thermogenesis in healthy adult men . Obesity 2011 19 13 –16 . 10.1038/oby.2010.105 20448535 19 Blondin DP Labbé SM Tingelstad HC Noll C Kunach M Phoenix S Guérin B Turcotte EE Carpentier AC Richard D Increased brown adipose tissue oxidative capacity in cold-acclimated humans . Journal of Clinical Endocrinology & Metabolism 2014 99 E438 –E446 . 10.1210/jc.2013-3901 24423363 20 Ouellet V Labbé SM Blondin DP Phoenix S Guérin B Haman F Turcotte EE Richard D Carpentier AC. Brown adipose tissue oxidative metabolism contributes to energy expenditure during acute cold exposure in humans . Journal of Clinical Investigation 2012 122 545 –552 . 10.1172/JCI60433 22269323 21 Koska J Ksinantova L Sebokova E Kvetnansky R Klimes I Chrousos G Pacak K. Endocrine regulation of subcutaneous fat metabolism during cold exposure in humans . Annals of the New York Academy of Sciences 2002 967 500 –505 . 10.1111/j.1749-6632.2002.tb04308.x 12079880 22 Kingma BRM Frijns AJH Saris WHM Steenhoven AA van Marken Lichtenbelt WD Increased systolic blood pressure after mild cold and rewarming: relation to cold-induced thermogenesis and age . Acta Physiologica 2011 203 419 –427 . 10.1111/j.1748-1716.2011.02336.x 21707931 23 Hong YC Kim H Oh SY Lim YH Kim SY Yoon HY Park M. Association of cold ambient temperature and cardiovascular markers . Science of the Total Environment 2012 435–436 74 –79 . 10.1016/j.scitotenv.2012.02.070 24 Dong M Yang X Lim S Cao Z Honek J Lu H Zhang C Seki T Hosaka K Wahlberg E Cold exposure promotes atherosclerotic plaque growth and instability via UCP1-dependent lipolysis . Cell Metabolism 2013 18 118 –112 . 10.1016/j.cmet.2013.06.003 23823482 25 The Eurowinter Group Cold exposure and winter mortality from ischaemic heart disease, cerebrovascular disease, respiratory disease, and all causes in warm and cold regions of Europe . Lancet 1997 349 1341 –1346 . 10.1016/S0140-6736(96)12338-2 9149695 26 Cypess AM Lehman S Williams G Tal I Rodman D Goldfine AB Kuo FC Palmer EL Tseng YH Doria A Identification and importance of brown adipose tissue in adult humans . New England Journal of Medicine 2009 360 1509 –1517 . 10.1056/NEJMoa0810780 19357406 27 van den Beukel JC Grefhorst A Hoogduijn MJ Steenbergen J Mastroberardino PG Dor FJ Themmen AP Women have more potential to induce browning of perirenal adipose tissue than men . Obesity 2015 23 1671 –1679 . 10.1002/oby.21166 26179979 28 Solianik R Skurvydas A Vitkauskiene A Brazaitis M. Gender-specific cold responses induce a similar body-cooling rate but different neuroendocrine and immune responses . Cryobiology 2014 69 26 –33 . 10.1016/j.cryobiol.2014.04.015 24809633 29 Johnson F Mavroggiani A Ucci M Vidal-Puig A Wardle J Could increased time spent in a thermal comfort zone contribute to population increases in obesity? Obesity Reviews 2011 12 543 –551 . 10.1111/j.1467-789X.2010.00851.x 21261804

PMC005xxxxxx/PMC5002970.txt

==== Front BMC MedBMC MedBMC Medicine1741-7015BioMed Central London 65710.1186/s12916-016-0657-8Research ArticleThe epidemiology, healthcare and societal burden and costs of asthma in the UK and its member nations: analyses of standalone and linked national databases http://orcid.org/0000-0002-3083-436XMukherjee Mome +44 (0)131 650 3813mome.mukherjee@ed.ac.uk 12Stoddart Andrew 12Gupta Ramyani P. 3Nwaru Bright I. 14Farr Angela 5Heaven Martin 6Fitzsimmons Deborah 5Bandyopadhyay Amrita 6Aftab Chantelle 7Simpson Colin R. 1Lyons Ronan A. 6Fischbacher Colin 8Dibben Christopher 9Shields Michael D. 10Phillips Ceri J. 5Strachan David P. 3Davies Gwyneth A. 11McKinstry Brian 2Sheikh Aziz 11 Asthma UK Centre for Applied Research, Centre for Medical Informatics, Usher Institute of Population Health Sciences and Informatics, The University of Edinburgh, Edinburgh, EH8 9AG UK 2 Edinburgh Clinical Trials Unit, Centre for Medical Informatics, Usher Institute of Population Health Sciences and Informatics, The University of Edinburgh, Edinburgh, EH8 9AG UK 3 Population Health Research Institute, St George’s, University of London, Cranmer Terrace, London, SW17 0RE UK 4 School of Health Sciences, University of Tampere, 33014 Tampere, Finland 5 Swansea Centre for Health Economics (SCHE), College of Human and Health Science, Swansea University, Singleton Park, Swansea, SA2 8PP UK 6 Farr Institute, Swansea University Medical School, Singleton Park, Swansea, SA2 8PP UK 7 The University of Edinburgh & The Royal College of Surgeons of Edinburgh, Clinical Surgery, Royal Infirmary of Edinburgh, Edinburgh, EH16 4SA UK 8 Information Services Division (ISD), NHS National Services Scotland, Room 111, Gyle Square, 1 South Gyle Crescent, Edinburgh, EH12 9EB UK 9 School of Geosciences, The University of Edinburgh, Edinburgh, EH8 9AG UK 10 Centre for Experimental Medicine, School of Medicine, Dentistry and Biomedical Sciences, Wellcome Wolfson Institute Experimental Medicine, Queen’s University Belfast, 97 Lisburn Road, Belfast, BT9 7BL UK 11 Asthma & Allergy Group, Institute of Life Science, Swansea University Medical School, Swansea University, Singleton Park, Swansea, SA2 8PP UK 29 8 2016 29 8 2016 2016 14 1 11316 3 2016 15 7 2016 © The Author(s). 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Background There are a lack of reliable data on the epidemiology and associated burden and costs of asthma. We sought to provide the first UK-wide estimates of the epidemiology, healthcare utilisation and costs of asthma. Methods We obtained and analysed asthma-relevant data from 27 datasets: these comprised national health surveys for 2010–11, and routine administrative, health and social care datasets for 2011–12; 2011–12 costs were estimated in pounds sterling using economic modelling. Results The prevalence of asthma depended on the definition and data source used. The UK lifetime prevalence of patient-reported symptoms suggestive of asthma was 29.5 % (95 % CI, 27.7–31.3; n = 18.5 million (m) people) and 15.6 % (14.3–16.9, n = 9.8 m) for patient-reported clinician-diagnosed asthma. The annual prevalence of patient-reported clinician-diagnosed-and-treated asthma was 9.6 % (8.9–10.3, n = 6.0 m) and of clinician-reported, diagnosed-and-treated asthma 5.7 % (5.7–5.7; n = 3.6 m). Asthma resulted in at least 6.3 m primary care consultations, 93,000 hospital in-patient episodes, 1800 intensive-care unit episodes and 36,800 disability living allowance claims. The costs of asthma were estimated at least £1.1 billion: 74 % of these costs were for provision of primary care services (60 % prescribing, 14 % consultations), 13 % for disability claims, and 12 % for hospital care. There were 1160 asthma deaths. Conclusions Asthma is very common and is responsible for considerable morbidity, healthcare utilisation and financial costs to the UK public sector. Greater policy focus on primary care provision is needed to reduce the risk of asthma exacerbations, hospitalisations and deaths, and reduce costs. Electronic supplementary material The online version of this article (doi:10.1186/s12916-016-0657-8) contains supplementary material, which is available to authorized users. Keywords AsthmaEpidemiologyBurdenCostUKhttp://dx.doi.org/10.13039/501100000362Asthma UKAUK-PG-2012-178Sheikh Aziz Edinburgh Clinical Trials Unit (ECTU)The Farr Instituteissue-copyright-statement© The Author(s) 2016 ==== Body Background Asthma is now one of the most common long-term conditions in the world [1, 2]. Our previous related work, commissioned to inform United Kingdom (UK) and Scottish parliamentary reviews of allergy services, demonstrated that the UK had amongst the highest prevalence of allergy and asthma in the world [2–4]. This work was important in influencing a range of national policy developments, but is now dated. Given that asthma was highlighted as the major contributor to the estimated burden and costs, there is a particular need for a more up-to-date and detailed review of the burden, healthcare utilisation and costs, and outcomes of asthma [5]. This need was underscored by the recent National Review of Asthma Deaths, which concluded that “46 % of asthma deaths could have been avoided with better routine care” [6]. In undertaking the present study, we sought to overcome important limitations of our previous studies [2–4] by extending the scope from healthcare costs alone to include wider societal costs and by incorporating data from previously unavailable datasets; these included out-of-hours care, ambulance, accident and emergency (A&E), intensive care unit (ICU) utilisation, and disability living allowance (DLA) data. In this paper, we describe the overall epidemiology, healthcare utilisation and costs of asthma for the UK as a whole and, for the first time, its member countries (i.e. England, Scotland, Wales and Northern Ireland). Methods Overview of methods Our methods have been described in detail in our published protocol [7]. We undertook secondary analyses of national health surveys, primary and secondary National Health Service (NHS) datasets, and national administrative data. In instances where relevant data were unavailable from a single source, datasets were linked. Overall, we analysed data from 27 datasets, of which five were linked. The data sources used to measure the study outcomes for each country are shown in Table 1. We analysed and reported the findings for 2011–12, or where this was not available, for 2010–11. The base year used for all costs was the 2011–12 financial year, applying appropriate inflation indices where required [8].Table 1 Study outcomes with datasets used, demographic and time-trend information availability therein, by UK nation Blank cells indicate unavailability of data source for the needs and scope of our study aDue to nature of data collection, data could not be standardised for all ages Symbols represent: age, sex, socioeconomic status, time trend SAIL, Secure Anonymised Information Linkage Study population The denominator for each dataset was based on the total sample of people in the dataset or the mid-year population estimate of the country where the dataset covered the entire population. The mid-year UK population estimates were 62,759,456 in 2010–11 and 63,285,100 in 2011–12 [9, 10]. Study outcomes Depending on the dataset, patients with asthma were defined as follows:Being diagnosed with asthma in primary care based on relevant Read codes (Additional file 1: Appendix 1) [7] Respondents in national health surveys who reported symptoms or treatment suggestive of asthma or reported clinician-diagnosed asthma [7] Having received asthma medications prescribed by their general practitioner (GP) for asthma, where prescriptions were coded using British National Formulary (BNF) codes (Additional file 1: Appendix 2) [7, 11] Using NHS out-patient clinic, out-of-hours service, ambulance service or A&E for asthma Having a primary diagnosis of asthma with ICD-10 code of J45 for asthma or J46 for status asthmaticus at discharge from hospital [7, 12] Having a primary diagnosis of asthma at admission with Read codes (Additional file 1: Appendix 3) in paediatric ICU and ICD-10 codes J45 or J46 or Acute Physiology and Chronic Health Evaluation (APACHE) III diagnostic code for asthma in adult ICUs [7, 13] Having ICD-10 codes J45 or J46 for asthma as the main disabling condition for claiming DLA or as the underlying cause of death at registration [12, 14] Thus patients or events where the main reason for healthcare or societal care utilisation was asthma were all included. This criterion thus (1) includes patients with asthma who might have had comorbidities, but only their asthma was accounted for and not their comorbidities, (2) does not include patients where asthma was not the main diagnosis. Outcome measures and datasets used The data sources used in the respective countries to assess the outcome measures along with availability of demographic information used and time trend by UK nations is presented in Table 1. Incidence Our primary aim was to measure healthcare utilisation; therefore, our focus was to estimate asthma incident-spells that generated a contact with primary care (see description for England below). Secondarily, where possible, we also estimated the incidence of first occurrence of asthma (incident cases) (see description for Scotland and Wales below). However, considering that an asthma episode may present in secondary care and that most UK primary and secondary care data are not linked, it was challenging to identify with certainty if an asthma episode presenting in secondary care represented the first occurrence of the asthma case. Due to differences among data sources and reporting, we used two measures of incidence, namely (1) clinician-reported mean weekly incident spells of asthma and (2) clinician-reported onset of asthma. In England, weekly incidence of asthma episodes was estimated from averaging new weekly episodes recorded by the Weekly Returns Service (WRS) of the Royal College of General Practitioners [15]. WRS receives notifications of weekly episodes and numbers of consultations for asthma using ICD-9 code 493, from about 90 general practices covering over 800,000 people in England. WRS episodes are available by age-groups and sex for each quarter and year. In Scotland, Practice Team Information (PTI) was used to measure onset of asthma resulting in new GP consultation [16]. PTI is a GP-database comprising a sample of 60 general practices representing about 6 % of Scottish general practices and around 6 % of the Scottish patient population. It includes GP and nurse consultations and diagnoses using Read codes, along with demographic data. PTI was established in 2003–04 and we used this year as the starting point of follow-up for 5 years. Onset of asthma was defined as new GP consultation in patients who were consistently in PTI since 2003–04 and did not consult their GP for asthma for those 5 years, but consulted their GP for asthma after 2008–09. This assumes that patients who consulted their GP for asthma before 2003 would come to see their GP at least once in those 5 years. Following this method, only new consultations which had a Read code for asthma in 2011–12 were counted. In Wales, onset of asthma resulting in new GP consultation was estimated from the Secure Anonymised Information Linkage (SAIL) databank, which during the study period collected data from 42 % of the GP practices in Wales [17]. There were data available on demographics and diagnoses based on Read codes (Additional file 1: Appendix 1). Only patients who had not deregistered from the participating GP practices and did not consult a GP for asthma between 2006–07 and 2010–11 and had new consultations with Read code for asthma (Additional file 1: Appendix 1) in 2011–12 were counted. We could not identify any GP-database in Northern Ireland that could be used to estimate annual onset of asthma by new GP consultation within the available budget for this work. Prevalence We defined annual prevalence as the proportion of the population who experienced symptoms of asthma at least once during the study year and life prevalence as the proportion of the population who experienced symptoms of asthma for at least part of their lives at any time during their life course [18]. Besides using lifetime and annual prevalence, we distinguished between patient-reported and clinician-reported measures, when the data pertained to health surveys and primary care data recorded by GPs, respectively. Thus, we used seven measures of prevalence, of which (1) lifetime prevalence of patient-reported symptoms suggestive of asthma, (2) annual prevalence of patient-reported symptoms suggestive of asthma, (3) lifetime prevalence of patient-reported clinician-diagnosed asthma, (4) annual prevalence of patient-reported clinician-diagnosed symptomatic asthma, and (5) annual prevalence of patient-reported clinician-diagnosed-and-treated asthma were based on national health surveys and (6) annual prevalence of clinician-reported-and-diagnosed asthma and (7) annual prevalence of clinician-reported-diagnosed-and-treated asthma were based on primary care data. The health surveys used were the Health Survey for England (HSE) [19], Scottish Health Survey (SHeS) [20], Welsh Health Survey (WHS) [21] and Northern Ireland Health Survey [22]. These surveys were of randomly selected samples of people broadly representative of their respective general populations. They included information on self-reported health and utilisation of health services. While the questions for asthma in the national health surveys were similar in England and Scotland, in Wales, only one question was asked questioning whether the respondent had been treated for asthma in that year [23]. Thus, only annual prevalence of patient-reported clinician-diagnosed-and-treated asthma could be estimated for Wales using the national health survey. Northern Ireland Health Survey asthma data were mainly on adult respondents, since information on children of ages between 2 and 14 years were grouped together. We thus could not use this information on children on age standardisation and hence national estimates could not be ascertained for Northern Ireland using national health survey data [23]. The prevalence estimates from primary care databases came from WRS in England [15], PTI in Scotland [16] and SAIL-GP in Wales [17], all of which had individual-level data. Additional estimates came from the Quality and Outcomes Framework (QOF) data, which was the only data source available from all four UK nations [24–27]; this was, however, aggregated at GP practice level, so could not provide a breakdown by age and sex [7]. QOF data pertaining to asthma (which is one of the many indicators) are a count of all people of all ages with asthma registered with GP practices, excluding patients with asthma who were not prescribed asthma-related drugs in the last 12 months (Quality Improvement code Asthma 1) [28]. We were unable to identify any suitable primary care data source from Northern Ireland within our budget. The definitions for the prevalence measures used were:Lifetime prevalence of patient-reported symptoms suggestive of asthma – defined as the number of people who had responded yes to “Have you had wheezing or whistling in the chest at any time, either now/in the past?” in HSE for England or in SHeS for Scotland, divided by the number of respondents who had answered that question in HSE or SHeS for England and Scotland, respectively. Annual prevalence of patient-reported symptoms suggestive of asthma – defined as the number of people who had responded yes to “Have you had wheezing or whistling in the chest in the last 12 months?” in HSE for England or in SHeS for Scotland, divided by the number of respondents who had answered that question in HSE or SHeS, for England and Scotland, respectively. Lifetime prevalence of patient-reported clinician-diagnosed asthma – for England, defined as the number of people who had responded yes to “Did a doctor or nurse ever tell you that you had asthma?” in HSE divided by the number of respondents who had answered that question, and for Scotland “Did a doctor ever tell you that you had asthma?” in SHeS, divided by the number of respondents who had answered that question. Annual prevalence of patient-reported clinician-diagnosed symptomatic asthma – for England, defined as the number of people who had responded yes to both the questions “Have you had wheezing or whistling in the chest in the last 12 months?” and “Did a doctor or nurse ever tell you that you had asthma?” in HSE, divided by the number of respondents who had answered the former question. For Scotland, the number of people who had responded yes to both the questions “Have you had wheezing or whistling in the chest in the last 12 months?” and “Did a doctor ever tell you that you had asthma?” in SHeS, divided by the number of respondents who had answered the former question. Annual prevalence of patient-reported clinician-diagnosed-and-treated asthma – defined as the number of people who had responded yes to both the questions “Did a doctor or nurse ever tell you that you had asthma?” and “Over the last 12 months, have you used an inhaler/puffer/nebuliser prescribed by a doctor to treat your asthma/wheezing/whistling/difficulty in breathing?” in HSE, divided by the number of respondents who had answered the latter question for England. For Scotland, the number of people who had responded yes to both the questions “Did a doctor ever tell you that you had asthma?” and “Were you treated in the past 12 months for wheeze by GP/nurse at surgery/community/school/district nurse/hospital, consultant/specialist at hospital, consultant/specialist elsewhere, homeopath/acupuncturist/other alternative medicine professional” in SHeS, divided by the number of respondents who had answered the latter question. For Wales, the only available question used was “are you currently being treated for asthma” in WHS, with the numerator as the number of respondents who said yes to that and the denominator as the total number of respondents to that question. Annual prevalence of clinician-reported-and-diagnosed asthma – based on PTI’s Read code grouping ‘asthma’ (Additional file 1: Appendix 1) and obtained from 39 practices participating in PTI in 2011–12 who had submitted complete GP and practice-nurse data in Scotland and from the GP practices that participated in SAIL-GP in Wales in 2011–12. PTI data were broadly representative of the Scottish population, and so is the population covered by the GP practices representing the Welsh population, thus these estimates are generalizable for Scotland and Wales. Annual prevalence of clinician-reported-diagnosed-and-treated asthma – defined as the proportion of people of all ages who were prescribed asthma-related drugs by GPs for their symptoms of asthma in the last 12 months (Quality Improvement code Asthma 1) [28], compared to the population size the GP practice covered in that year (the list size). Since age and sex are not available in QOF, age standardised rate could not be reported and thus only crude rates are presented. Healthcare utilisation in primary care GP and nurse consultation For estimating GP and nurse consultations for asthma, WRS was used for England, PTI for Scotland and SAIL-GP for Wales. WRS had ICD-9 codes, PTI and SAIL-GP had Read codes (Additional file 1: Appendix 1). Prescriptions Community prescriptions for asthma were identified from the SAIL-GP database based on our list of BNF codes (Additional file 1: Appendix 2) [7]. Some medications indicated for use in asthma can also be prescribed to treat other conditions, primarily chronic obstructive pulmonary disease (COPD). We therefore confined our analysis of medications to people with a clinician-recorded diagnosis of asthma, but without a diagnosis of COPD [7]. To address this issue, medications of patients who had COPD and asthma were included so long as they received the list of asthma medication we had (Additional file 1: Appendix 2), but patients with COPD but no asthma were excluded. For comparison, the total cost of all medications used for asthma (excluding immunosuppressants, which are also used for other conditions) was also examined in Prescription Costs Analysis (PCA) data for each of the four countries [29–32]. It should be noted that, because PCA data did not include information on diagnosis, these data included medications with an indication for asthma, but which were prescribed for other conditions. In pre-specified sensitivity analyses, community prescribing costs for Scottish patients under the age of 40 (in whom COPD is rare), extracted from Scottish Prescribing Information Systems (PIS) data, were compared to the figures extrapolated from the SAIL-GP results to test the reliability of the extrapolation process. Out-of-hours Information on calls to an out-of-hours NHS service were only obtainable from NHS 24 Scotland, the national telephone triage and advice service [33]. Data are available from 2008 onwards. All calls where the nurse triaging the out-of-hours call selected an asthma-specific algorithm to support their decision-making were collected. In England, although an out-of-hours surveillance team exists, a breakdown by asthma was not available. The out-of-hours data in Wales were inconsistently collected across areas and hence were not used. We could get out-of-hours data on asthma for Northern Ireland. Healthcare utilisation in secondary care Out-patient attendances Although routine data on attendances in NHS out-patient clinics were available across the four nations, these data were, however, captured under the broader heading of ‘respiratory’ consultations and it was therefore not possible to estimate the proportions of these consultations that were particularly for asthma. This is noted as a major data gap. Ambulance services For Scotland, asthma data from the Scottish Ambulance Service, which had data from 2008–09 onwards, were used where the record had “Emergency call-asthma selected” [34]. Usage of ambulance service due to asthma could not reliably be estimated from the aggregated routinely collected data available in England, Wales and Northern Ireland; this is therefore identified as a data gap. A&E services In England, there were no accurate published data on A&E attendances for asthma. In Scotland, the A&E data mart was used for sites which reported (excluding Orkney and Tayside Health Boards, since they currently only submit high level diagnosis codes to ISD) patient-level information from their A&E departments since 2010–11 [35]. The data reported here are from ‘new’ and ‘unplanned return’ attendances at A&E, i.e. excluding those who were ‘recall’ or ‘planned return’. If the ‘disease code’ included the ICD-10 codes above or ‘R062’ (Family history of asthma) or if the ‘presenting complaint text’ or ‘diagnosis text’ referred to any of the terms asthma, wheezing, low saturation, chest tightness or shortness of breath, then those cases were selected. In Wales, the SAIL Emergency Department Dataset, which contains data since 2009, were used [36]. There was an audit data in Northern Ireland on A&E in one of its Trusts, namely Belfast Health and Social Care Trust, which collected data from 2007–08 onwards. Since that data from one urban Trust may not have been representative of the entire nation, we did not use these for our national estimates. Inpatient and day cases in hospitals We queried the Hospital Episode Statistics for England [37], General/Acute Inpatient and Day Case dataset for Scotland [38], SAIL Patient Episode Database for Wales [39], and Hospital Inpatient System dataset in Northern Ireland [40], for primary diagnosis of asthma with ICD-10 codes to identify all asthma episodes. However, hospital-based prescribing was not included in these datasets. ICUs The Paediatric Intensive Care Audit Network (PICA Net) is a national audit which collects data on all critically ill children admitted to paediatric ICUs across the UK [41]. It had data from England and Wales from 2002, from Scotland from 2007, and from Northern Ireland from 2008 onwards, recorded in Read version 3 (Additional file 1: Appendix 3). For adults in ICUs, for England, Wales and Northern Ireland, we used Intensive Care National Audit & Research Centre (ICNARC) data, which have been collected since 1996 and which uses ICD-10 codes [42]. For Scotland, the Scottish Intensive Care Society Audit Group (SICSAG) data were queried and APACHE III diagnoses for asthma was used [13, 43]. For all of these countries, data from stand-alone ICUs or from ICUs mixed with high dependency units were included. Data from stand-alone high dependency units were excluded. Wider societal impact To capture impact beyond the health services, we investigated absenteeism in school and at work, care-at-home, DLA and mortality. Absenteeism School and work absenteeism data were obtained from HSE 2010 for England, from the questions “Over the last 12 months, how many days has your asthma/wheezing/whistling in (your/his/her) chest caused (you/him/her) to be absent from school?” and “Over the last 12 months, how many days has your wheezing/whistling in your chest, shortness of breath or difficulty in breathing caused you to be absent from work?”, respectively, among asthma respondents. We could not identify any suitable data source to investigate school and work absenteeism in Scotland, Wales or Northern Ireland. Since HSE reports absence as a categorical variable: < 5 days, 5–9, etc., we used mid-points to estimate the number of days. The estimates produced were for the number of days of absence as a proportion of the sample population, so the rates were of number of days of absence per 1000 population. Care-at-home We were unable to identify any suitable data to estimate costs of care at home for asthma from any nation. DLA Aggregated data were available from the Department of Work and Pensions (DWP) [14], the government agency providing national benefits on number of people receiving DLA and DLA amount, with asthma as the main disabling condition in England, Scotland and Wales for 2011–12. For Northern Ireland, there were data available from the Department for Communities on number of people receiving DLA due to asthma as the main disabling condition and total amount by age-group, sex and SES from 2008 [44]. Premature retirement We could not identify a data source for this outcome and therefore identified this as a data gap. Mortality Mortality data with underlying cause of death as asthma from death certificate registrations, coded using ICD-10, are available from the Office of National Statistics for England and Wales [45], National Records Scotland [46], and Northern Ireland Statistics and Research Agency for Northern Ireland [47] were queried. Analyses Counts of events or people, as the case was, were obtained across all the age-groups (<5, 5–9, 10–14, … 70–74, > 75 years) (except QOF) for that year, along with the denominator. For comparison across nations, figures obtained across the datasets were age standardised using the European Standard Population (Version 2013) [48]. Age-standardised epidemiological, healthcare utilisation, school and work absenteeism and DLA estimates, accompanied by their respective 95 % confidence intervals (CIs) were reported based on the Poisson approximation [49]. UK-wide summaries of incidence and prevalence estimates and associated 95 % CIs were calculated by inverse variance, fixed effect meta-analyses in R (Version 3.1.0). Healthcare costs were estimated from a NHS perspective based on healthcare utilisation using NHS data, detailed in Additional file 1: Appendix 4. Where a given dataset did not include a direct measure of costs, standard UK price weights were applied to generate cost estimates for each form of healthcare [8, 11, 50]. Additional file 1: Appendix 4 summarises the costing method applied in each case. Primary care price weights were taken from the Personal and Social Services Research Unit. Community prescribing costs were based on net ingredient costs based on SAIL-GP prescribing for asthma medications in non-COPD patients [11, 17]. These data contained details of the type and date of medication prescribed, but not the number of items prescribed on a given date. Therefore, a conservative assumption that a single pack was prescribed at each time had to be applied, which underestimated the costs associated with prescriptions for multiple items. Inpatient care costs were based on NHS Reference Cost estimates based on their associated Healthcare Resource Grouping (Version 4) [50]. Societal costs were estimated from a wider societal perspective, including NHS costs as above and DLA. Though we originally aimed to include productivity costs, it was not possible to reliably estimate costs due to school and work absenteeism since the data were not asthma specific and excluded some key variables. Addressing data gaps for cost analysis Data gaps found were of three forms: (1) within country, where no single dataset in a given country held sufficient variables to provide an estimate, but linkages between datasets could overcome this; (2) between countries, where data on the variable of interest (after allowing for linkages) were available in one member country, but not in another; and (3) across countries, where no data (after allowing for linkages) from any member country were found for the variables of interest. For type (1) data gaps, linked data were used. For type (2) data gaps, estimates of a given variable from one country, where available (e.g. prescription costs in Wales), were mapped onto other countries, adjusting for population size, annual prevalence of clinician-reported-diagnosed-and-treated asthma (QOF), which was available across all the nations, and the age and sex distribution of patients who reported having asthma in the respective country’s health survey, or in the SAIL-GP database (Additional file 1: Appendix 5). For type (3) data gaps, a literature search was undertaken in an attempt to provide parameter estimates for modelling (although no usable data-fitting modelling requirements were found). Economic modelling An economic model of the costs of asthma in the UK and its member countries was built in Excel 2010 (Additional file 1: Appendix 6). In brief, data on resource use were taken directly from healthcare utilisation data based on the internal diagnostic coding (such as Read code or ICD-10) available in each dataset. For factors such as hospital episodes and DLA, whole population datasets were available and complete, thus not requiring any adjustment. The model applied the price weights detailed above to generate costs from this resource use [8, 11, 50]. Costs based on a sample within a country were extrapolated to population levels by rescaling per head of age-sex stratified population. Where results were extrapolated to another country due to data gaps, additional rescaling was undertaken based on each country’s relative QOF prevalence, in order to account for differences in prevalence rates between countries [25–28]. QOF was selected for this process due to its relationship to treated asthma, which provided the most appropriate data source for healthcare utilisation and was the only measure of prevalence measured in a uniform manner in all countries. The model was also used to sum the cost estimates into any required groupings; bootstrapping (with 10,000 replicates) was used to estimate 95 % CIs around the joint distributions of each total cost estimate using the percentile method (Additional file 1: Appendix 6) [51]. Following recommendations in standard modelling guidance [52], the uncertainty around prevalence estimates were simulated using a beta distribution and uncertainty around cost estimates were simulated using a gamma distribution or a normal distribution where sample sizes were large and central limit theorem was expected to hold (Additional file 1: Appendix 6). Results The data below refer to UK-wide estimates unless otherwise stated. Incidence We estimated that the annual age-standardised incidence of GP-diagnosed asthma was 3.8/1000 (95 % CI, 3.8–3.9), equivalent to approximately 240,000 people in the UK developing asthma in 2011–12. On average, there were 5600 weekly incident GP episodes of asthma (Table 2).Table 2 Incidence and prevalence of asthma in patients of all ages by UK nation Epidemiologic measures England Scotland Wales Northern Ireland UK estimate (inverse variance, fixed effect meta-analysis) n ASRi n ASRi n ASRi n ASR n ASR N (95 % CI) N (95 % CI) N (95 % CI) N (95 % CI) N (95 % CI) Incidence/1000 in 2011–12 Clinician-reported onset of asthma1,2,a,b,c 20,780 3.8 4779 3.7 240,483 3.8 5,511,732 (3.7–3.9) 1,108,024 (3.6–3.9) 63,285,100 3.8–3.9 Clinician-reported mean weekly incident-spells of asthma3,a 77 0.1 5,696 0.1 722,885 (0.1–0.1) 63,285,100 0.1–0.1 Prevalence/100 patient-reported in 2010–11 and clinician-reported in 2011–12 Lifetime prevalence of patient-reported symptoms suggestive of asthma4,5,d 4335 31.3 794 24.6 18,514,040 29.5 14,112 (30.2–32.4) 3256 (22.9–26.4) 62,759,456 27.7–31.3 Annual prevalence of patient-reported symptoms suggestive of asthma4,5,e 2465 18.0 489 14.8 10,731,867 17.1 14,112 (17.1–18.8) 3256 (13.4–16.2) 62,759,456 15.7–18.5 Lifetime prevalence of patient-reported clinician-diagnosed asthma4,5,f 2280 16.1 443 14.0 9,790,475 15.6 14,112 (15.3–16.9) 3256 (12.6–15.3) 62,759,456 14.3–16.9 Annual prevalence of patient-reported clinician-diagnosed symptomatic asthma4,5,g 1235 8.6 229 7.0 5,083,516 8.1 14,112 (8.0–9.1) 3256 (6.1–7.9) 62,759,456 7.2–9.1 Annual prevalence of patient-reported clinician-diagnosed-and-treated asthma4,5,6,h 1320 9.3 322 9.8 1901 9.8 6,024,908 9.6 14,112 (8.6–9.9) 3255 (8.7–11.0) 19,225 (9.4–10.4) 62,759,456 8.9-10.3 Annual prevalence of clinician-reported-and-diagnosed asthma1,2,a,b,c 310,050 5.7 63,873 5.7 3,600,861 5.7 5,511,732 (5.7–5.7) 1,119,368 (5.7–5.8) 63,285,145 5.7–5.7 Annual prevalence of clinician-reported-diagnosed-and-treated asthma7,ii Crude rate 3,295,944 5.9 319,091 6.0 218,243 6.9 113,518 6.0 4,303,390 6.8 55,525,732 (5.9–5.9) 5,299,097 (6.0–6.0) 3,185,538 (6.9–6.9) 1,898,678 (5.9–6.0) 63,285,145 6.8–6.8 Source: 1Practice Team Information (PTI), Scotland; 2Secure Anonymised Information Linkage-GP, Wales; Health Survey for England 2010; 3Weekly Returns Service, England; 4Health Survey for England, 5Scottish Health Survey 2010; 6Welsh Health Survey, 2010; 7Quality Outcomes Framework (QOF) iAge standardised rate (ASR) iiSince age and sex are not available in QOF, crude rate is presented Blank cells had no data availability aBased on ISD’s Read Code Grouping ‘Asthma’ bPTI estimates are based on 40, 43, 39 and 39 practices that submitted complete GP and practice-nurse data over a 6-year period ending 31 March 2009, 2010, 2011 and 2012, respectively. PTI data are broadly representative of the Scottish population cThe Welsh estimates apply to GP practice areas that participate in SAIL-GP. Population covered by these GP practices represent the Welsh population, thus these estimates are generalizable for Wales d,e,f,g,hPrevalence estimates were derived from questions in repeated population health surveys of the respective UK nations d“Have you ever had wheezing/whistling in the chest at any time, either now/in the past?” in England and Scotland e“Have you had wheezing or whistling in the chest in the last 12 months?” in England and Scotland fEngland – “Did a doctor or nurse ever tell you that you had asthma?”; Scotland – “Did a doctor ever tell you that you had asthma?”; Wales – there was no equivalent question asked in the survey from Wales gQuestions in e and f hQuestions in f AND “Over the last 12 months, have you used an inhaler/puffer/nebuliser prescribed by a doctor to treat your asthma/wheezing/whistling/difficulty in breathing?” for England, “Were you treated in the past 12 months for wheeze by GP/nurse at surgery/community/school/district nurse/hospital, consultant/specialist at hospital, consultant/specialist elsewhere, homeopath/acupuncturist/other alternative medicine professional?” for Scotland, and “Are you currently being treated for asthma?” for Wales Prevalence The lifetime prevalence of patient-reported symptoms suggestive of asthma was 29.5 % (95 % CI, 27.7–31.3), equivalent to 18.5 m people. The annual prevalence of patient-reported symptoms suggestive of asthma was 17.1 % (95 % CI, 15.7–18.5), equivalent to 10.7 m people (Table 2). The lifetime prevalence of patient-reported clinician-diagnosed asthma was 15.6 % (95 % CI, 14.3–16.9), equivalent to 9.8 m people; annual prevalence of patient-reported clinician-diagnosed symptomatic asthma was 8.1 % (95 % CI, 7.2–9.1), which equated to 5.1 m people; annual prevalence of patient-reported clinician-diagnosed-and-treated asthma was 9.6 % (95 % CI, 8.9-10.3), which equated to 6.0 m people; annual prevalence of clinician-reported-and-diagnosed asthma was 5.7 % (95 % CI, 5.7–5.7), which equated to 3.6 m people; clinician-reported-diagnosed-and-treated asthma was 6.8 % (95 % CI, 6.8–6.8), which equated to 4.3 m people. Healthcare utilisation in primary care There were an estimated 2.7 m (95 % CI, 2.6–3.0) GP consultations, 3.7 m (95 % CI, 3.6–4.1) nurse consultations and 54,000 (95 % CI, 53,000–60,000) out-of-hours calls for asthma (Table 3).Table 3 Healthcare utilisation in primary care for asthma across all ages in 2011–12 by UK nation Healthcare utilisation measure in primary care England Scotland Wales Northern Ireland UK estimate3 n ASRa n ASRa n ASRa n ASRa n (000 s) N (95 % CI) N (95 % CI) N (95 % CI) N (95 % CI) (95 % CI) Number of General Practitioner consultations1 215,610 39.1 2700 5,511,732 (39.0–39.3) (2600–3029) Number of nurse consultations1 289,120 53.4 3693 5,511,732 (53.2–53.6) (3577–4152) Out of hours calls2 4575 0.9 54.3 5,299,900 (0.8 to 0.9) (53–60) Source: 1Practice Team Information for Scotland; 2NHS 24 for Scotland; 3from cost modelling aAge standardised rate (ASR) per 1,000 people registered with GP practices in Wales and population for Scotland. Estimates were standardised using the 2013 European Standard Population Blank cells had no data availability Healthcare utilisation in secondary care There were an estimated 113,000 (95 % CI, 108,000–132,000) ambulance conveyances for asthma; 121,000 (95 % CI, 108,000–146,000) A&E attendances; 93,900 (95 % CI, 93,900–93,900) in-patient episodes; 6100 (95 % CI, 5900–6200) day-case episodes; and 1800 (95 % CI, 1700–1900) ICU episodes (Table 4). The total length of stay for inpatients and day-cases in UK relating to asthma was 195,000 days.Table 4 Healthcare utilisation in secondary care for asthma across all ages in 2011–12 by UK nation Healthcare utilisation measure in secondary care England Scotland Wales Northern Ireland UK estimate7 n ASRa n ASRa n ASRa n ASRa n (000 s) N (95 % CI) N (95 % CI) N (95 % CI) N (95 % CI) (95 % CI) Ambulance conveyance1 8263 1.6 112.9 5,299,900 (1.6–1.7) (107.6–131.8) Accident and emergency (A & E) attendances in hospital2,b 8457 1.7 2321 0.7 121.1 4,868,230 (1.6–1.7) 3,033,591 (0.7–0.8) (108–146) Inpatient episodes of hospital care (for asthma as the primary reason for care)3,c 76,319 1.4 7744 1.5 7887 2.6 1966 1.1 93,916 53,107,200 (1.4–1.4) 5,299,900 (1.4–1.5) 3,033,591 (2.5–2.7) 1,814,318 (1.0–1.1) (93,916–93,916) Day-case episodes of hospital care (for asthma as the primary reason)3,c 5066 9.4 142 2.7 768 25.7 144 7.0 6120 53,107,200 (9.1–9.7) 5,299,900 (2.2–3.1) 3,033,591 23.9–27.6 1,814,318 5.88–8.20 (5929–6248) Intensive care unit episodes for asthma as the primary reason for care4,5,6,d 1537 2.8 179 3.3 97 3.0 55 3.0 1868 53,107,200 (2.7–3.0) 5,299,900 (2.8–3.8) 11,931,062 (2.4–3.6) 1,704,245 (2.2–3.8) (1739–1932) Source: 1Scottish Ambulance Service (SAS); 2A&E data mart in Scotland (excluding Orkney and Tayside Health Boards) and SAIL-Emergency Department Dataset for Wales; 3Hospital Episode Statistics-England, General/Acute Inpatient and Day-Case-Scotland, SAIL-Patient Episode Database-Wales and Department of Health, Social Service and Public Safety in Northern Ireland; 4For children, Paediatric Intensive Care Audit Network (PICANet); 5For adults, Intensive Care National Audit & Research Centre (ICNARC)-England, Northern Ireland and Wales and 6Scottish Intensive Care Society Audit (SICSAG)-Scotland; 7From cost modelling aAge standardised rate (ASR), using the 2013 European Standard Population; per 1000 population of the country for ambulance, accident and emergency (A&E) and inpatients, and per 100,000 population for day-cases and intensive care bIncludes ‘New’ and ‘Unplanned Return’ attendances only at A&E, excludes those who are ‘Recall’ or ‘Planned Return’. For Scotland based on A&E sites which returned episode-level information for at least one of the following: ICD10 Diagnosis code (R098/R068/R062/R060/R05X/R05/J46X/J46/J459/J458/J451/J450/J45/R688/R69X/R69/Z825/J21/J210/J211/J218/J219/R06/R09/R092) OR Diagnosis free-text extracted from “Wheez”/“Asthma”/“Ashtma”/“low” AND “sats”(“chest” AND “tight”) AND (“SOB” OR (“short” AND “breath”)). However, most Health Boards use a pick list/disease code from ICD-10 codes, these are usually mapped from diagnosis text where a pick list has been used. NHS Tayside and NHS Orkney only submit high-level diagnosis codes (comprises about 6 % of total attendance), thus have been excluded here. Thus, figures presented here will be an underestimate of the true number of attendances to A&E for Scotland cICD-10 codes J45/J46 as primary reason for care. For Wales, the first non-R or Z code in day-cases were also used additionally. R codes refer to “symptoms” and Z codes to “factors influencing health status and contact with health services” dAsthma as primary reason for care with Read codes in PICANet, ICD-10 codes J45/J46 in ICNARC and APACHE III diagnostic codes in SICSAG Blank cells had no data availability Wider societal impact School absenteeism for asthma or asthma symptoms accounted for 252.4 days/1000 children (95 % CI, 241.3–263.5; n/N = 1267/5352), equivalent to 2.8 m (95 % CI, 2.6–3.0) absences. Work absenteeism for asthma symptoms accounted for 78.9 days/1000 adults (95 % CI, 72.6–85.3, n/N = 535/6978), equivalent to 4.1 m (95 % CI, 3.4–4.7) work-days lost. For asthma, DLA was claimed by an estimated 36,980 people; 24,100 people in England, 3600 people in Scotland, 3300 people in Wales and 5980 people in Northern Ireland. There were an estimated 1160 deaths (2.1/100,000; 95 % CI, 2.0–2.2) due to asthma; 982 deaths in England (2.1/100,000; 95 % CI, 2.0–2.2), 94 in Scotland (2.0/100,000; 95 % CI, 1.6–2.3), 58 in Wales (2.0/100,000; 95 % CI, 1.5–2.5), and 26 in Northern Ireland (1.9/100,000; 95 % CI, 1.2–2.7). Financial costs of asthma We estimated that asthma cost at least £1.1 bn with the majority of costs (74 %) arising in primary care, of which 81 % were for community prescribing. Table 5 provides a detailed breakdown of this estimate by member countries and cost elements. For comparison, the total cost of all medications listed in PCA data with an indication for use in asthma (irrespective of condition actually prescribed for) was also £1.1 bn in 2011 (2011–12 for Scotland), of which £821.2 m, £97.5 m, £66.1 m and £38.7 m were incurred in England, Scotland, Wales and Northern Ireland, respectively. It is important to note that the parity between the £1.1 bn total cost estimate from our model and the £1.1 bn total cost of medications with an indication for asthma from PCA data is entirely co-incidental and occurs due to PCA providing an overestimate of medication costs in this context, rather than being a component of the costs used in the model.Table 5 Breakdown of estimated costs for asthma in the UK by member country in 2011–12 England Scotland Wales Northern Ireland UK Cost (95 % CI) Cost (95 % CI) Cost (95 % CI) Cost (95 % CI) Cost (95 % CI) (£000 s) (£000 s) (£000 s) (£000 s) (£000 s) (£000 s) (£000 s) (£000 s) (£000 s) (£000 s) GP consultations 89,926 (86,614–101,526) 8624 (8138–9120) 6408 (6116–7411) 3029 (2906–3436) 107,987 (103,986–121,168) Practice nurse consultations 43,021 (41,614–48,745) 4048 (3876–4213) 3202 (3073–3706) 1431 (1379–1627) 51,702 (50,083–58,131) Community prescribing 552,514 (536,694–568,687) 54,514 (51,890–57,191) 40,572 (40,178–40,977) 18,845 (18,150–19,504) 666,445 (650,112–683,375) Calls to out-of-hours 1325 (1291–1485) 130 (130–130) 86 (84–98) – – 1541 (1507–1710) Ambulance Trips 27,511 (26,077–32,480) 2408 (2408–2408) 2378 (2238–2876) 876 (828–1033) 33,172 (31,624–38,649) Accident and emergency 10,907 (9553–13,357) 913 (913–913) 889 (759–1131) 392 (298–495) 13,101 (11,625–15,782) Hospital episodes (excluding intensive care units (ICU)) 69,162 (69,162–69,162) 6342 (6342–6342) 8128 (8087–8169) 2,064 (2064–2064) 85,696 (85,656–85,737) ICU episodes 4413 (4413–4413) 482 (482–482) 236 (236–236) 129 (129–129) 5260 (5260–5260) Total NHS cost 798,780 (780,199–824,168) 77,462 (74,296–79,704) 61,899 (61,141–63,650) 26,764 (25,975–27,772) 964,905 (945,648–991,409) Disability living allowance 95,500 (95,500–95,500) 14,800 (14,800–14,800) 12,800 (12,800–12,800) 23,832 (23,832–23,832) 146,932 (146,932–146,932) Total public sector costs 894,280 (880,112–924,082) 92,262 (89,579–94,986) 74,699 (74,177–76,686) 50,596 (49,935–51,732) 1,111,837 (1,097,840–1,143,601) Please see individual sections of this paper for full commentary and caveats. An important note on the derivation and interpretation of the confidence intervals detailed here is also available in our published protocol A sensitivity analysis comparing community prescribing costs for Scottish patients under the age of 40 years from PIS data to the costs estimated for the same group extrapolated from SAIL-GP data produced similar figures of £19.2 m and £18.8 m, respectively. A further sensitivity analysis for inpatient episodes costs with individual country results for England, Scotland and Northern Ireland produced similar results (Additional file 1: Appendix 7). However, due to the higher rates of inpatient episodes per head of population reported in Wales, sensitivity analysis which extrapolated all inpatient episode costs from the Welsh results raised the estimate to £147.0 m, i.e. approximately 70 % higher than the base case. Figures extrapolated from each of the other countries to Wales on the other hand ranged from £5.0 m to £5.5 m, or approximately 32–38 % lower (Additional file 1: Appendix 7). Discussion We found that the prevalence of asthma varied widely depending on the definition used, ranging from 29.5 % (18.5 m people) for lifetime symptoms suggestive of asthma to 5.7 % (3.6 m) for those with active, clinician-diagnosed-and-treated asthma. Considerable care therefore needs to be taken in defining the populations being discussed and consistent use of the seven different definitions proposed in this paper should help greatly in this respect. We also found that, even with conservative assumptions, there was considerable morbidity, healthcare utilisation and costs such that asthma now costs the UK public sector well in excess of £1.1 bn per annum. The overwhelming majority of these costs are incurred in relation to prescribing in primary care for preventive treatments, but despite this, there were almost 100,000 inpatient episodes for asthma and over 1000 asthma deaths. These data suggest that particular focus is warranted on primary care to assess whether the most effective and cost-effective treatment strategies are consistently being employed [53], and on the development of innovative strategies for the prevention and early detection of asthma attacks. We have produced the most comprehensive national work ever undertaken estimating the prevalence, care utilisation and financial costs of asthma in the UK. We scoped, obtained data from and interrogated 27 health and social care datasets from across the four nations of the UK, which either used well-defined sampling strategies (e.g. the national surveys) or covered large sections of the population (e.g. primary care databases) or indeed entire nations (e.g. hospital episode statistics and mortality data). We believe that our findings are therefore likely to be generalizable across the UK. Additional strengths come from the fact that we followed a pre-specified analysis plan and that we undertook a range of pre-specified sensitivity analyses to test our assumptions [7]. There are, however, a number of limitations that need to be considered. First, whilst we have undoubtedly made progress in addressing important data gaps previously identified (e.g. in relation to providing estimates for out-of-hours care, urgent care clinics, ambulance trips, A&E attendances and ICU admissions), some still exist, for example, in relation to out-patient clinic visits, presenteeism (i.e. attending work when unwell) [54], and absence from work to care for children [54]. Our results from A&E data marts may be underestimates because patients presenting with asthma exacerbations may not always have been coded with asthma (with terms such as shortness of breath or wheeze being used instead). We could not access reliable data on prescribing of Omalizumab, a biological agent used for severe, persistent asthma [55]. Given the above described limitations, our cost estimates should therefore be seen as minimum likely financial costs to the UK public sector. Second, whilst use of national surveys offers important insights into patient perspectives, these exclude the homeless, those living in institutional care and special populations (e.g. armed forces and prisoners). Third, many of the costs estimated in this study required extrapolation from one country to another (Additional file 1: Appendix 5). This process was undertaken by first rescaling based on differences in Office of National Statistics population estimates, then by annual prevalence of clinician-reported-diagnosed-and-treated (QOF) asthma between countries. While broader definitions of asthma would be expected to generate larger prevalence estimates within a country, we might expect their relative rates between countries to be similar and thus have minimally impacted this process. However, this assumption is impossible to test since no other definition of prevalence is uniformly measured within all four countries. It was not possible to account for differences in definitions of diagnosis of asthma (i.e. before uplift to national level estimates where necessary) as observed in differences between coding systems because basic resource use attributable to asthma is recorded using varying system coding based definitions (such as Read or ICD-10). The extrapolation process additionally accounted for differences in population by age and sex and prevalence in countries, but due to data limitations, could not account for other factors such as socioeconomic status/deprivation, ethnicity or disease severity profiles. It also makes the assumption that the rate of resource use per asthma patient is the same in all UK countries. We made efforts to cross-validate this assumption by comparing extrapolated results to known results, where possible. Although the majority of these exercises produced similar figures, extrapolating inpatient episode costs to and from Wales provided an exception due to the higher rate of inpatient episodes observed in Wales (Table 4). It is not possible to rule out similar issues where extrapolation could not be cross-validated. The age-standardised prevalence and burden of asthma reported in our study are not easily compared to other estimates of asthma prevalence because (1) of differences in the age-groups, time periods and geographical settings studied [1, 3, 56–59]; (2) in contrast to many previous studies, we generated a number of estimates of asthma ‘prevalence’; and (3) DALYs have been reported in some previous studies (which was neither within the scope of this work nor was it possible given the data gaps identified). Our estimate of the proportion of medication cost broadly agrees with a systematic review which found that medications accounted for 38 % to 89 % of the total cost of asthma [5]. Although we have captured more data sources and costs, the increases to total costs were small and partially offset by more conservative prescribing assumptions. This is likely to explain why our costs are similar to a previous study in England and Wales that estimated spending at £754.4 m in 2000–02 [2], (£994.9 m at 2011–12 prices). They are also lower in Scotland than a previous study which estimated £98.1 m in 2003–05 (£117.0 m in 2011–12 prices) [3], again likely due to methodological differences and the more conservative approach used here. For example, only the burden of asthma as the main problem of the patient had been taken into account. Thus, for a patient with asthma and comorbidities which might have resulted in higher health and societal care utilisation, additional costs of care for comorbidities were not accounted for. It is therefore important that these estimates are not confused with burden of asthma, and it is clarified that these estimates are for burden of asthma in patients who utilised health and societal care when asthma was their main problem. Despite these differences, there is broad agreement that the UK has one of the highest asthma burdens in the world [1, 58, 60]. We have created a profile of asthma for the UK based on available data. This information will be useful to national and regional policymakers and health planners both in the UK and internationally since it can be used as a template for similar mapping of asthma in other nations. It should also be of considerable interest to respiratory physicians, GPs and the public both to consider the current level of health and social care utilisation and by serving as a benchmark for national and regional improvement efforts. Our work also has important implications for the academic community, particularly in relation to considering approaches to harmonising definitions and data collection procedures across the four nations of the UK [23], and in terms of finding novel ways of filling the outstanding data gaps. Finally, this work will also offer a number of potentially transferable lessons for generating robust national estimates of the epidemiology, care utilisation and costs of other long-term conditions. Conclusions In summary, we have found that the UK continues to experience a very high disease and cost burden from asthma. Since much of the morbidity and mortality is considered potentially preventable, there is a need for ambitious national targets for reducing asthma exacerbations and the associated risks of hospitalisations and deaths. Additional file Additional file 1: Appendix 1. Read codes for asthma, version 2. Appendix 2. List of asthma medications by BNF codes. Appendix 3. Read codes for asthma, version 3. Appendix 4. Price weights and costings methodology. Appendix 5. Data sources and mapping used in economic modelling of the cost of asthma in theUK by member countries. Appendix 6. Cost modelling technical supplement. Appendix 7. Comparison of mapped inpatient episodes to known values for their respective countries. (DOCX 88 kb) Acknowledgements We are very grateful for the support we have received from Mrs Elisabeth Ehrlich, Patient and Public Involvement lead, Asthma UK Centre for Applied Research; Mohammad Al-Sallakh in Secure Anonymised Information Linkage (SAIL) for helping with some analysis for Wales towards the end of the project; Mary Scarlett, Katie McClelland, Kyle Hamilton, Kerri Burgess, Rachel Stewart, Suzanna McVeigh, Penny Murray, Jenny Gingles, John Maguire, Christine Kennedy and Jennifer Myers (Northern Ireland Department of Health); Carrie Doole (Northern Ireland Statistics and Research Agency); Carole Brunton, Claire Walsh, Neil McKeown (Northern Ireland Department for Regional Development); Alison Vitty (Northern Ireland Ambulance Service); Crown Copyright, National Centre for Social Research and UK Data Service for the health surveys; Jodie Batchelor, Phil McShane, Sarah Fleming, Roger Parslow, Allan Wardhaugh (Paediatric Intensive Care Audit Network); Sarah Power, Emily Robinson, Donna Hickford, Helen Dingle, Lucy Lloyd-Scott, David Harrison (Intensive Care National Audit and Research Centre); Paul Martin (Department for Work and Pensions); Chris Weir, Richard Parker, Joel Smith (Edinburgh Clinical Trials Unit); Andrew Malloy, Julie Ramsay (National Records of Scotland); Derek Milligan, Katy Barclay, Derek Ho (Scottish Ambulance Service); Malcolm Alexander, Gillian Burns, John McAnaw, Philippe Hourcastagné, Linda Ralph (NHS 24); Alistair Smith, Charmaine Walker, Donna Mikolajczak, Rachael Briggs, Leanne Hopkins, Lee Wilson, Murray Bell, Simon Quirk, Carole Morris, Maighread Simpson, Catriona Haddow, Sian Nowell, Darren Hair, Andrew Duffy, Ishbel Robertson, Anita Pritchard, Julie Peacock, Michael Fleming, Bradley Kirby, David Clark, Jackie Caldwell, Catriona Young, Rose Sisk, Sian Nowell (Information Services Division, NHS Scotland); Alex Bailey, Calum Melrose (NHS Lothian); Lynn Morrice, Anne Douglas, Catherine Bromley, Nazir Lone, Rebecca Campbell, Anna Wierzoch (The University of Edinburgh). Funding This study was funded by Asthma UK (reference: AUK-PG-2012-178), with staff support from Edinburgh Clinical Trials Unit and the Farr Institute. The funders had no influence on the analytical methods employed, interpretation of results or the decision to publish. The Farr Institute is supported by a 10-funder consortium: Arthritis Research UK, the British Heart Foundation, Cancer Research UK, the Economic and Social Research Council, the Engineering and Physical Sciences Research Council, the Medical Research Council, the National Institute of Health Research, the National Institute for Social Care and Health Research (Welsh Assembly Government), the Chief Scientist Office (Scottish Government Health Directorates), and the Wellcome Trust (MRC Grant No: MR/K006525/1). Availability of supporting data We had access to data which were a) publicly available online, e.g. national health surveys (Additional file 1: Appendix 8) b) aggregated by year, sex and age-group level (e.g. hospital discharge data), which are available only to named researchers and cannot be shared without the approval of relevant data custodians c) individual level data (e.g. primary care data), obtained after ethics approval, available only to named researchers and cannot be shared. Authors’ contributions AS conceived the study and was the PI. The grant holders were AS, DS, CP and GD. MM was the main researcher, study co-ordinator and wrote the first draft. ASt was the health economist for the cost analysis and reported on the cost sections. MM, ASt, RG, MH, AF, AB and BN analysed data. All co-authors contributed to the work and reviewed and commented on the drafts. All authors read and approved the final manuscript. Competing interests The authors declare that they have no competing interests. Consent for publication Not applicable. Ethics approval and consent to participate Ethics approval was obtained through The University of Edinburgh’s Centre for Population Health Sciences Research Ethics Committee and for patient level data access we obtained approvals from the respective Data Custodians. ==== Refs References 1. Mallol J Crane J von Mutius E Odhiambo J Keil U Stewart A Group tIPTS International Study of Asthma and Allergies in Childhood (ISAAC) Phase Three: A global synthesis Allergologia et Immunopathologia 2013 41 2 73 85 10.1016/j.aller.2012.03.001 22771150 2. Gupta R Sheikh A Strachan DP Anderson HR Burden of allergic disease in the UK: secondary analyses of national databases Clin Exp Allergy 2004 34 4 520 6 10.1111/j.1365-2222.2004.1935.x 15080802 3. Anandan C Gupta R Simpson C Fischbacher C Sheikh A Epidemiology and disease burden from allergic disease in Scotland: analyses of national databases JRSM 2009 102 10 431 42 10.1258/jrsm.2009.090027 4. Anandan C Nurmatov U van Schayck OC Sheikh A Is the prevalence of asthma declining? Systematic review of epidemiological studies Allergy 2010 65 2 152 67 10.1111/j.1398-9995.2009.02244.x 19912154 5. Bahadori K Doyle-Waters M Marra C Lynd L Alasaly K Swiston J FitzGerald JM Economic burden of asthma: a systematic review BMC Pulm Med. 2009 9 24 10.1186/1471-2466-9-24 19454036 6. Levy ML The national review of asthma deaths: what did we learn and what needs to change? Breathe 2015 11 1 14 24 10.1183/20734735.008914 26306100 7. Mukherjee M Gupta R Farr A Heaven M Stoddart A Nwaru BI Fitzsimmons D Chamberlain G Bandyopadhyay A Fischbacher C Estimating the incidence, prevalence and true cost of asthma in the UK: secondary analysis of national stand-alone and linked databases in England, Northern Ireland, Scotland and Wales—a study protocol BMJ Open 2014 4 11 e006647 10.1136/bmjopen-2014-006647 25371419 8. Curtis L, Personal Social Services Research Unit (PSSRU) University of Kent. Unit Costs of Health and Social Care 2012. 2012. http://www.pssru.ac.uk/project-pages/unit-costs/2012/. Accessed 7 March 2016. 9. Office of National Statistics. Population estimates for UK, England and Wales, Scotland and Northern Ireland, Mid-2001 to Mid-2010 Revised. 2013. http://www.ons.gov.uk/ons/rel/pop-estimate/population-estimates-for-uk--england-and-wales--scotland-and-northern-ireland/mid-2001-to-mid-2010-revised/index.html. Accessed 3 March 2016. 10. Office for National Statistics. Population estimates for UK, England and Wales, Scotland and Northern Ireland, Mid 2011. Estimates of the usual resident population for the UK as at 30 June of the reference year Provided by administrative area, single year of age and sex 2013. http://webarchive.nationalarchives.gov.uk/20160105160709/http://ons.gov.uk/ons/rel/popestimate/population-estimates-for-uk--england-and-wales--scotland-and-northern-ireland/mid-2011-and-mid-2012/index.html. Accessed 3 March 2016. 11. British National Formulary. 2012. www.bnf.org. Accessed 3 March 2016. 12. World Health Organization. International Classification of Disease-10. http://apps.who.int/classifications/icd10/browse/2010/en#/J40-J47. Accessed 3 March 2016. 13. Knaus WA Wagner DP Draper EA Zimmerman JE Bergner M Bastos PG Sirio CA Murphy DJ Lotring T Damiano A The APACHE III prognostic system. Risk prediction of hospital mortality for critically ill hospitalized adults Chest 1991 100 6 1619 36 10.1378/chest.100.6.1619 1959406 14. Department for Work & Pensions. https://www.gov.uk/government/organisations/department-for-work-pensions. Accessed 26 April 2016. 15. Royal College of General Practitioners. RCGP Research and Surveillance Centre. http://www.rcgp.org.uk/clinical-and-research/our-programmes/research-and-surveillance-centre.aspx. Accessed 13 May 2016. 16. Information Services Division. NHS National Services. GP Consultations/Practice Team Information (PTI) Statistics. http://www.isdscotland.org/Health-Topics/General-Practice/GP-consultations/. Accessed 13 May 2016. 17. Secure Anonymised Information Linkage (SAIL). http://www.saildatabank.com/. Accessed 3 March 2016. 18. Last J Porta M A Dictionary of Epidemiology 2008 5 New York Oxford University Press 19. UK Data Service. Health Survey for England, 2010. https://discover.ukdataservice.ac.uk/catalogue/?sn=6986. Accessed 13 May 2016. 20. UK Data Service. Scottish Health Survey, 2010. https://discover.ukdataservice.ac.uk/catalogue/?sn=6987. Accessed 13 May 2016. 21. Government of Wales. Welsh Health Survey, 2010. http://gov.wales/statistics-and-research/welsh-health-survey/?tab=previous&lang=en. Accessed 13 May 2016. 22. Department of Health Northern Ireland. Health Survey Northern Ireland, 2010. https://www.health-ni.gov.uk/articles/health-survey-northern-ireland. Accessed 13 May 2016. 23. Nwaru BI Mukherjee M Gupta RP Farr A Heaven M Stoddart A Bandyopadhyay A Fitzsimmons D Shields M Phillips C Challenges of harmonising data from UK national health surveys: a case study of attempts to estimate the UK prevalence of asthma J R Soc Med. 2015 108 433 9 10.1177/0141076815600909 26432811 24. Health and Social Care Information Centre. Quality and Outcomes Framework. http://www.hscic.gov.uk/qof. Accessed 7 May 2016. 25. Information Services Division NHS National Services. Quality & Outcomes Framework. http://www.isdscotland.org/Health-Topics/General-Practice/Quality-And-Outcomes-Framework/. Accessed 10 May 2016. 26. Department of Health Northern Ireland. Quality and Outcomes Framework Achievement Data 2011/12. https://www.health-ni.gov.uk/publications/quality-and-outcomes-framework-qof-achievement-data-201112. Accessed 13 May 2016. 27. Welsh Government. General Medical Services Contract: Quality and Outcomes Framework. 2011–12. http://gov.wales/statistics-and-research/general-medical-services-contract/?tab=previous&lang=en. Accessed 13 May 2015. 28. Health and Social Care Information Centre, UK Government. Indicators for Quality Improvement, Quality and Outcomes Framework, Outcomes framework domain classification of IQI indicators. 2011–12. http://www.hscic.gov.uk/media/10037/Outcomes-framework-domain-classification-of-IQI-indicators/xls/Copy_of_Outcomes_framework_domain_classification_of_IQI_indicators.xls. Accessed 27 Apr 2016. 29. HSC Business Services Organisation. Prescription Cost Analysis. 2011. http://www.hscbusiness.hscni.net/services/2266.htm. Accessed 13 May 2016. 30. Health and Social Care Information Centre, England. Prescription Cost Analysis – England, 2011, 2012 [NS]. 2011, 2012 http://www.hscic.gov.uk/searchcatalogue?productid=5461&q=title%3a%22prescription+cost+analysis%22&sort=Relevance&size=10&page=1#top; http://www.hscic.gov.uk/searchcatalogue?productid=11412&q=title%3a%22prescription+cost+analysis%22&sort=Relevance&size=10&page=1#top. Accessed 26 Apr 2016. 31. Information Services Division, NHS National Services Scotland. Prescription Cost Analysis. 2011, 2012. http://www.isdscotland.org/Health-Topics/Prescribing-and-medicines/Community-Dispensing/Prescription-Cost-Analysis/. Accessed 26 Apr 2016. 32. Welsh Government. Prescriptions dispensed in the community. Prescription Cost Analysis. 2011, 2012. http://gov.wales/statistics-and-research/prescriptions-dispensed-community/?tab=previous&lang=en. Accessed 26 Apr 2016. 33. NHS 24 Scotland’s National Telehealth and Telecare Organisation. NHS Inform. http://www.nhs24.com/. Accessed 13 May 2016. 34. Scottish Ambulance Service. http://www.scottishambulance.com/. Accessed 13 May 2016. 35. Information Services Division NHS National Services. Accident and Emergency (A&E) Data Mart. http://www.isdscotland.org/Health-Topics/Emergency-Care/Accident-and-Emergency-Data-Mart/. Accessed 13 May 2016. 36. Secure Anonymised Information Linkage. Emergency Department Data Set. http://www.saildatabank.com/data-dictionary/sail-datasets/emergency-department-data-set-%28edds%29.aspx. Accessed 3 March 2016. 37. Health and Social Care Information Centre. Hospital Episode Statistics, Admitted patient care. http://www.hscic.gov.uk/searchcatalogue?q=title%3A%22Hospital+Episode+Statistics%2C+Admitted+patient+care+-+England%22&area=&size=10&sort=Relevance. Accessed 26 Apr 2016. 38. Information Services Division Scotland. Publications-Hospital Care. http://www.isdscotland.org/Health-Topics/Hospital-Care/Publications/data-tables.asp?id=975#975. Accessed 26 Apr 2016. 39. Public Health Wales Observatory. Patient Episode Database for Wales. http://www.wales.nhs.uk/sitesplus/922/page/50308. Accessed 10 Sept 2014. 40. Department of Health, Northern Ireland. Hospital Statistics. www.health-ni.gov.uk/articles/inpatient-and-day-case-activity. Accessed 13 May 2016. 41. Paediatric Intensive Care Audit Network. Details of the data collected by PICANet are given in their annual reports. Paediatric Intensive Care Audit Network Annual Report 2012 - 2014 (published 2015): Universities of Leeds and Leicester. http://www.picanet.org.uk/Audit/Annual-Reporting/. Accessed 13 May 2016. 42. Intensive Care National Audit & Research Centre. https://www.icnarc.org/. Accessed 13 May 2016. 43. Scottish Intensive Care Society Audit Group. http://www.sicsag.scot.nhs.uk/. Accessed 13 May 2016. 44. Department for Communities in Northern Ireland. https://www.communities-ni.gov.uk/news/functions-new-department-communities. Accessed 13 May 2016. 45. Office for National Statistics. All data related to deaths. https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/datalist. Accessed 13 May 2016. 46. National Records Scotland (NRS). Deaths. 2011. http://www.nrscotland.gov.uk/statistics-and-data/statistics/statistics-by-theme/vital-events/deaths/deaths-time-series-data. Accessed 26 Apr 2016. 47. Northern Ireland Statistics and Research Agency website (NISRA). Deaths. http://www.nisra.gov.uk/demography/default.asp10.htm. Accessed 10 Sept 2014. 48. Eurostat - Statistical Office of the European Union. Revision of the European Standard Population. Report of Eurostat’s task force. Version 2013. http://ec.europa.eu/eurostat/documents/3859598/5926869/KS-RA-13-028-EN.PDF/e713fa79-1add-44e8-b23d-5e8fa09b3f8f. Accessed 3 March 2016. 49. Scottish Public Health Observatory. Standardisation-worked examples, 95 % C.I. for EASR (Poisson approximation). http://www.scotpho.org.uk/publications/other-key-resources/methodology. Accessed 3 March 2016. 50. Department of Health. NHS reference costs 2011–12. https://www.gov.uk/government/publications/nhs-reference-costs-financial-year-2011-to-2012. Accessed 3 March 2016. 51. Briggs AH Wonderling DE Mooney CZ Pulling cost-effectiveness analysis up by its bootstraps: a non-parametric approach to confidence interval estimation Health Econ. 1997 6 327 40 10.1002/(SICI)1099-1050(199707)6:4<327::AID-HEC282>3.0.CO;2-W 9285227 52. Briggs A Claxton K Sculpher M Decision modelling for health economic evaluation 2006 New York Oxford University Press 53. British Thoracic Society, Scottish Intercollegiate Guidelines Network. British Guideline on the Management of Asthma: A national clinical guideline. 2014. https://www.brit-thoracic.org.uk/document-library/clinical-information/asthma/btssign-asthma-guideline-2014/. Accessed 3 March 2016. 54. Gosselin E Lemyre L Corneil W Presenteeism and absenteeism: differentiated understanding of related phenomena J Occup Health Psychol 2013 18 1 75 86 10.1037/a0030932 23276197 55. National Institute for Health and Clinical Excellence. Technology Appraisal Guidance 278. Omalizumab for treating severe persistent allergic asthma (review of technology appraisal guidance 133 and 201). 2013 https://www.nice.org.uk/guidance/ta278/resources/costing-statement-425321245. Accessed 3 March 2016. 56. Global Asthma Network. The Global Asthma Report 2014. Auckland: GAN; 2014. 57. Torpy JM Campbell A Glass RM Chronic diseases of children JAMA 2010 303 7 682 10.1001/jama.303.7.682 20159879 58. Patel SP Jarvelin MR Little MP Systematic review of worldwide variations of the prevalence of wheezing symptoms in children Environ Health. 2008 10 7 57 59. Anderson HR Gupta R Strachan DP Limb E 50 years of asthma: UK trends from 1955 to 2004 Thorax 2007 62 1 85 90 10.1136/thx.2006.066407 17189533 60. Gupta R Anderson HR Strachan DP Maier W Watson L International trends in admissions and drug sales for asthma Int J Tuberc Lung Dis 2006 10 2 138 45 16499251

PMC005xxxxxx/PMC5002971.txt

==== Front Arch Public HealthArch Public HealthArchives of Public Health0778-73672049-3258BioMed Central London 14810.1186/s13690-016-0148-6ResearchLeisure time physical activity in Estonian population: adherence to physical activity recommendations and relationships with overweight http://orcid.org/0000-0002-2170-7780Tali Maie Maie.Tali@kliinikum.ee 12Lusmägi Peeter peeter@eok.ee 3Unt Eve Eve.Unt@kliinikum.ee 1241 Department of Sports Medicine and Rehabilitation, University of Tartu, 1a Puusepa St, 50406 Tartu, Estonia 2 Sports Medicine and Rehabilitation Clinic, Tartu University Hospital, 1a Puusepa St, 50406 Tartu, Estonia 3 Estonian Olympic Committee, 102c Pärnu mnt St, 11312 Tallinn, Estonia 4 Department of Cardiology, University of Tartu, 18 Ülikooli St, Tartu, Estonia 29 8 2016 29 8 2016 2016 74 1 365 2 2016 8 6 2016 © The Author(s). 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Background Sufficient physical activity (PA) is a key element for the prevention of non-communicable diseases. Considering leisure time physical activity (LTPA), the purpose of the survey was to provide descriptive data for LTPA, find the proportion of the study population meeting the recommended WHO PA criteria, and to detect the possible relationship between LTPA and overweight. Methods The National Physical Activity Survey was carried out in autumn 2015 in the Estonian population (n = 914) aged 15–69 years. For LTPA assessment, the LTPA domain of IPAQ-L interview version was used. LTPA was analysed in regard to fulfilment of the WHO PA recommendations and in association with BMI. Results Seventy-three percent of study participants reported any LTPA in the preceding 7 days. 22 % (26 % of men, 20 % of women) met WHO PA recommendations. 50 % of the study participants were considered overweight (48 % of men, 51 % of women) with BMI > 25.0 kg/m2, whereas 20 % of the total study population was obese (BMI ≥ 30.0 kg/m2). Lower adherence to WHO PA recommendations was associated with older age in men, and obesity in both men and women. Conclusions A strikingly low proportion of people met WHO PA recommendations and a relatively high proportion of overweight people were detected in the study group. Obesity had significant inverse associations with LTPA. Keywords Physical activityLeisure timeEstonian populationAdherence to physical activity recommendationsOverweightEstonian Ministry of Culture and Ministry of Education and ResearchIUT 02-7issue-copyright-statement© The Author(s) 2016 ==== Body Background According to WHO Global Health Risks report, high blood pressure, tobacco use, high blood glucose, sedentary lifestyle, and overweight are the five leading mortality risk factors in the world [1]. These risks are predisposing factors for the development of chronic non-communicable diseases, such as cardiovascular diseases (CVD), diabetes, and cancers [2]. On the other hand, there is augmentative evidence that a healthy lifestyle, including sufficient physical activity (PA), is a key element for the prevention of non-communicable diseases [3–5]. As in other countries in the world, CVD have been the most prevalent cause of early (before 65 years of age) disability and death in Estonia. In 2006, national guidelines for the prevention of CVD were renewed [6] and during the last decade the gradual decline in cardiovascular mortality was achieved [7]. Compared to Estonia’s closest neighbours, Latvia and Lithuania, and other former Soviet Union countries, lower CVD mortality rates have been attained [7]. Still, compared to other European countries, the CVD mortality rates are nearly twice as high or higher [7]. Hence, the modification of CVD risk factors, including lifestyle modification strategies is of utmost importance in achieving further improvements. The WHO global recommendations on PA for health suggest regular PA [3]. Unfortunately, the PA trends worldwide are inauspicious causing the term “pandemic of physical inactivity” to be introduced [8]. This declaration is in accordance with the latest special Eurobarometer survey on sports and PA, where 59 % of European citizens revealed never or seldom being engaged in PA or played sports [9]. In 2014, data of a health behaviour study, which is carried out every even year since 1990 among the Estonian adult population, indicated that 35.8 % of men and 37.4 % of women are engaged in leisure time PA (LTPA) (the questionnaire includes only one question regarding LTPA: “are you engaged in LTPA lasting at least 30 min per session, and at least 2–3 times per week”) [10]. According to the same study, 57.9 % of adult male population and 52.0 % of female population were overweight (BMI > 25.0 kg/m2) [10]. The trends in LTPA have been declining and the proportion of overweight people in the adult population has been increasing over the last decade in the Estonian population [10]. Since the previous data show the low proportions of people engaging in LTPA, there could be a potential reserve for applying interventions for health enhancement. Previous LTPA data of the Estonian population are based on a single question estimate [10], but more specific data are needed for planning interventions. The purpose of the present survey was to provide descriptive LTPA data and to find the proportion of the study population meeting the recommended criteria of WHO for PA considering LTPA. The secondary purpose was to detect the possible relationship between LTPA and overweight. Methods Study group The National Physical Activity Survey was ordered by the Estonian Ministry of Culture. The target population of the survey were permanent residents of Estonia aged 15 to 69 years. According to data of Statistics Estonia, 926,798 people of the required age group lived in Estonia on January 1, 2015 [11]. In the selection of the study group, the proportional model of the population was used, in which all have an equal opportunity of becoming a respondent. Afterwards, the socio-demographic structure of the sample in the breakdown of gender, age, ethnicity and place of residence with the corresponding statistical characteristics of the Estonian population was compared and adjusted if necessary. The initial sample consisted of 1768 respondents, who were visited in their homes by TNS Emor interviewers during September 14 and October 4, 2015. In total, 1004 subjects agreed to participate in the survey. The interviewers were educated to use a tablet, which was pre-programmed with the interview questionnaire. The face-to-face Tablet Assisted Personal Interviewing (TAPI) was conducted for data collection. In TAPI, the interview questions appeared on the tablet screen and the answers of a study respondent were inserted in the computer program by an interviewer immediately during the interview, which permits immediate surveillance of the interviewing process. Interviewers asked no guiding questions during the interview. Measures The study participants self-reported their age, body height and weight (for calculation of body mass index – BMI kg/m2); educational level (basic or primary, general secondary or vocational, higher); occupation (top manager/manager, professional/specialist/clerk, skilled worker/machine operator/driver; pupil/student; retired; other); monthly income per family member (low – below 300€, average – 301€–550€, over average – over 551€, no income/refused to answer), and place of residence (capital city, town > 35,000 inhabitants, town 3000-35,000 inhabitants, village/rural area). For the assessment of self-reported LTPA, the LTPA domain of the last seven days interview version of The International Physical Activity Questionnaire (IPAQ-L) was applied [12]. Since the IPAQ-L has not been used in Estonia before the present study, the English version of the questionnaire was translated into Estonian. In the translation process, relevant guidelines and recommendations were taken into account [12]. IPAQ-L was developed and validated at the beginning of the 2000s [13, 14] and is considered a national and international standard for PA surveillance in several countries. The IPAQ-L version provides information about the time spent in and intensity of PA for different domains (occupation, transportation, home, leisure time). Compared to other domains, IPAQ studies have shown the highest validity for LTPA [15–17]. In the present study, the LTPA domain (6 items) of IPAQ-L was used. Respondents were asked to report the frequency and duration of walking, moderate and vigorous intensity LTPA performed for at least 10 min per LTPA session during the last 7 days. Moderate activities were defined as activities with a temperate increase in respiration and heart rate (for example: brisk walking, Nordic walking, bicycling at a regular pace, swimming at a regular pace, doubles tennis). Vigorous activities were defined as activities with a significant increase in respiration, heart rate, and sweating (aerobics, running, fast bicycling, or fast swimming). Energy expenditure was expressed as metabolic equivalents of task (MET) multiplied by the time in minutes per week for walking, moderate and vigorous activity, and for the total LTPA (MET minutes per week). One MET corresponds to 3.5 ml O2 × kg−1 × min−1. Data processing For LTPA items (walking, moderate and vigorous), weekly and average time per day was calculated. 90 subjects were excluded due to confounding or missing data (reporting LTPA in more than 7 days a week, or refused). The final sample for LTPA prevalence analysis was 914 (373 men). Age was categorized into five groups (15–24 years, 25–34 years, 35–49 years, 50–64 years, and 65–69 years) for both genders. According to BMI, the study subjects were categorized as ‘underweight’– BMI < 18.5, ‘normal weight’ – BMI 18.5–24.9, ‘overweight’ – BMI 25.0–29.9, and ‘obese’ – BMI >30.0 [4]. Due to the unbalanced men/women ratio 1:17 in ‘underweight’, this group is not presented in the figures. For subjects aged 15–17 years, population-based age-adjusted criteria for BMI categorization were applied [18]. According to WHO recommendations, at least 150 min of moderate-intensity aerobic PA or 75 min of vigorous-intensity PA throughout the week is recommended for adults aged 18–64 years [3]. For adolescents, the WHO PA recommendation encompasses accumulation of at least 60 min of moderate-vigorous PA in every day of the week [3]. As there are no separate recommendations for LTPA and former research has shown that European adults are mostly physically active during leisure time [9], the WHO PA recommendations were used as a criterion for achieving a sufficient LTPA level. The latter approach has been implemented in earlier studies concerning LTPA [16, 19]. Regarding achievement of WHO PA recommendations, a dichotomized variable (meeting/not meeting) was created. To meet regularity and accumulation of moderate to vigorous LTPA in adults, the variable had to meet the following criteria: 1) 30 min of moderate LTPA per day on at least 5 days of the week; 2) accumulating 150 min of moderate LTPA per week with LTPA reported on at least 3 days of the week; 3) accumulating 25 min of vigorous LTPA on at least 3 days of the week; 4) a combination of moderate and vigorous LTPA on at least 3 days of the week with a total duration 150 min/week. In adolescents, accumulation of at least 60 min of moderate-vigorous LTPA in every day of the week was regarded as adherence to PA recommendations [3]. Statistical analysis Statistical analyses were carried out with the Statistical Package for the Social Sciences – SPSS, 22 (IBM, Armonk, NY, USA). To characterize the study sample, descriptive data – n, percentage (%), mean, SD, median, 25th and 75th percentile – were calculated. All LTPA distributions were skewed to the right. The unpaired t-test for parametric data and the Mann–Whitney U-test for nonparametric data were used to determine the differences between the groups. The chi-square (χ2-test) test was used to determine significant differences in proportions between the groups. Multinomial logistic regression analysis was conducted to reveal associations between achievement of PA recommendations and age groups, and BMI categories. In age groups the youngest group and in BMI categories the normal BMI category were set as reference groups. For all statistical analyses, the 0.05 level of significance was applied. Results The subjects’ mean age was 39.7 ± 15.3 years and 44.0 ± 15.9 years for men and women, respectively. In the final study group, there were more women than men and women were slightly older than men. Seventy-three percent of the study participants reported LTPA in the preceding 7 days (Table 1). Walking was the most frequently reported LTPA, while moderate-vigorous LTPA was reported at a considerably lower rate. Men reported significantly more vigorous LTPA than women (Table 1).Table 1 Leisure time physical activity duration and energy expenditure and the prevalence of people reporting respective leisure time physical activity according to gender. Leisure time physical activity survey of Estonian population in 2015 Men ( n = 373) Women (n = 541) Total study group (n = 914) Median (P 25th, P 75th) % Median (P 25th, P 75th) % Median (P 25th, P 75th) % LT Walking (min/day) 8.7 (0; 34.6) 58.2 17.1 (0; 60.0) 61.6 11.4 (0; 42.9) 61.1 Moderate LTPA (min/day) 0 (0; 15.0) 33.2 0 (0; 9.3) 33.3 0 (0: 12.8) 33.9 Vigorous LTPA (min/day) 0 (0; 17.1) 35.1 0 (0; 0) *** 23.3 0 (0; 5.9) 27.4 LT Walking (MET min/week) 198.0 (0; 800.3) 396.0 (0; 1097.2) 264.1 (0; 990.0) Moderate LTPA (MET min/week) 0 (0; 420.0) 0 (0; 260.8) 0 (0; 360.0) Vigorous LTPA (MET min/week) 0 (0; 960.0) 0 (0; 0) *** 0 (0; 320.0) Total LTPA (MET min/week) 891.0 (0; 2415.0) 70.2 760.0 (0; 2079.0) 74.3 792.0 (0; 2310.8) 72.6 Note: ***p < 0.001 women compared to men of the respective group P 25th, P 75th – 25 th and 75 th percentile LT leisure time LTPA leisure time physical activity The characteristics of the study group and the prevalence of people reporting LTPA, as well achievement of WHO PA recommendations in age groups and sociodemographic groups are presented in Table 2. The prevalence of reporting LTPA was higher in younger age groups. People with higher income and higher education reported participation in vigorous LTPA more often than people with low income and basic education. Among small-town male inhabitants, the higher prevalence of walking but lower prevalence of moderate LTPA was observed compared to capital residents; among women, the prevalence of walking was higher in village residents and the prevalence of moderate LTPA was lower in small-town residents. Compared to top managers, the prevalence of any LTPA was considerably lower in male skilled workers, whereas the prevalence of walking in LT was lower among female skilled workers. Considering WHO PA recommendations for weekly LTPA, 22 % of study participants met the criteria in total – 26 % of men and 20 % of women. The highest prevalence of people meeting WHO PA criteria was observed in men aged 15–24 years, whereas in other groups the prevalence was below 30 % with the lowest in men aged 65–69 years. Compared to the youngest age group of men, the WHO PA criteria were met by a considerably lower proportion of men older than 25 years, and by women of the youngest age group (p < 0.01). Among men, a significantly lower proportion of skilled workers and retired men compared to top managers met the WHO PA criteria. In contrast, the proportion of male students meeting the criteria was considerably higher compared to top managers.Table 2 The characteristics of the study group and the prevalence (%) of study group subjects reporting leisure time physical activities, and adherence to WHO physical activity recommendations. Leisure time physical activity survey of the Estonian population in 2015 Men (n = 373) Women (n = 541) % Walking % Moderate LTPA % Vigorous LTPA % Adherence to WHO PA recommendations % % Walking % Moderate LTPA % Vigorous LTPA % Adherence to WHO PA recommendations % Age groups 15–24 years 18.8 78.6 50.0 65.7 45.7 13.9 66.7 38.7 40.0 24.0 25–34 years 22.8 58.8** 37.6 34.1*** 27.1* 17.6 68.4 30.5 26.3 17.9 35–39 years 30.6 50.9*** 35.1 29.8*** 24.6** 27.9 60.3 30.5 23.8* 19.2 50–64 years 21.2 50.6*** 25.3** 12.7*** 15.2*** 29.1 53.2** 34.8 17.7*** 20.9 65–69 years 6.6 52.0* 12.0*** 20.0*** 12.0*** 11.5 29.0*** 33.9 11.3*** 14.5 Education Basic/primarya 11.3 45.5 27.3 12.1 18.2 7.3 27.3 33.7 10.0 13.3 Secondary/vocationala 64.2 38.3 29.8 28.7** 24.5 58.4 58.9 31.5 17.4 19.5 Higher 24.5 68.1 40.3 38.9** 23.6 34.3 64.8 37.3 28.9* 20.7 Place of residence Capital 34.3 28.9 37.5 37.5 31.3 31.1 29.2 35.7 26.2 18.8 Big town, >35,000 inhabitants 15.8 28.9 40.7 40.7 33.3 18.6 37.6 36.6 20.8 27.7 Small town, 3000–30,000 inhabitants 16.9 68.3*** 30.2** 28.6 23.9 19.2 34.6 28.8** 24.0 19.2 Village 33.0 42.3 28.2 20.5* 21.1 31.1 50.0*** 38.7 21.3 16.7 Income Low (<300 €) 14.7 49.1 27.3 16.4 21.8 18.1 44.9 18.6 11.2 18.8 Average (301–550 €) 23.9 49.4 23.6 33.7* 23.6 34.0 35.9 38.0 20.6* 19.7 High (>550 €) 35.4 38.6 45.5 37.9*** 30.3 22.4 37.2 31.4 29.8*** 16.5 No income/refused 26.0 36.1 35.1 36.1** 25.8 25.5 38.4 31.9 29.7** 23.2 Occupation Top manager 10.2 76.3 47.4 42.1 31.6 15.0 63.0 35.8 32.1 21.0 Professional/specialist 18.2 66.2 45.6 42.6 30.9 22.5 68.9 36.0 21.3 18.9 Skilled worker 28.2 48.6*** 20.0** 19.0*** 15.2*** 17.2 47.3* 34.4 19.4 20.4 Pupil/student 12.8 81.3 56.3 64.6* 45.8* 8.5 78.3 45.7* 54.3* 28.3 Retired 12.6 46.8 51.1 17.0* 21.3*** 18.1 58.2 18.4* 8.2*** 16.3 Other/Missing value 18.0 46.9*** 13.4 30.0 4.5*** 18.7 59.8 23.7 18.8 15.8 Note: a – pupils and students were excluded due to misclassification of education level *p < 0.05; **p < 0.01; ***p < 0.001 compared to the reference group (age group 15–24 years, basic/primary education, capital, low income, top manager) of the respective gender PA physical activity LTPA leisure time physical activity The mean BMI was 25.7 ± 4.7 kg/m2 and 26.2 ± 5.7 kg/m2 for men and women, respectively. 50 % of the study participants were considered overweight (48 % of men and 51 % of women), whereby 20 % of the total study population was obese (17 % of men and 22 % of women). The prevalence data of LTPA and adherence to WHO PA recommendations according to BMI categories are presented in Fig. 1. Compared to people with BMI in the normal range, a significantly lower proportion of overweight women as well obese men and women reported LT walking or vigorous LTPA and hence, there was a considerably lower prevalence of people in those groups who met the recommended WHO PA criteria. As an exception, in men with BMI in range 25.0–29.9 kg/m2, the proportion meeting WHO PA recommendations was the highest (32 %) compared to other BMI groups. In ‘underweight’ women (not presented in Fig. 1), adherence to PA recommendations did not significantly differ from the ‘normal weight’ group of the respective gender (28 % vs 24 %, respectively).Fig. 1 Proportion (%) of study group subjects participating in leisure-time physical activities and adherence to the WHO physical activity criteria, according to BMI groups. Note: *p < 0.05; **p < 0.01; ***p < 0.001 compared to the respondents in the normal body weight group (BMI 18.5–24.9 kg/m2 in adults and BMI in normal age adjusted range in adolescents) of respective gender. PA – physical activity Lower adherence to WHO PA recommendations was associated with older age in the total study group, which was mainly attributed to men, and obesity in both men and women (Table 3).Table 3 Associations between achievement of WHO physical activity recommendations and age, and body mass index [adjusted OR (95 % CI)]. Leisure time physical activity survey of the Estonian population in 2015 Achievement of WHO physical activity recommendations Men (n = 373) Women (n = 541) Total (n = 914) Age groups (OR adjusted for BMI, education level, occupation and place of residence) 15–24 years (reference group) 1.0 1.0 1.0 25–34 years 2.37 (1.05–5.36)* 1.59 (0.69–3.66) 2.01(1.14–3.53)* 35–49 years 2.49 (1.13–5.50)* 1.29 (0.57–2.91) 1.88 (1.08–3.27)* 50–64 years 4.30 (1.12–5.50)* 1.02 (0.45–2.29) 1.92 (1.08–3.43)* 65–69 years 5.23 (1.30–21.2)* 1.53 (0.56–4.19) 2.64 (1.21–5.73)* BMI categories (ORs are adjusted for age, education level, occupation and place of residence) 18.5–24.9 (reference group) 1.0 1.0 1.0 25.0–29.9 0.71 (0.41–1.20) 1.67 (0.97–2.86) 1.07 (0.74–1.54) >30.0 2.55(1.01–6.46)* 2.20 (1.15–4.21)* 2.14 (1.27–3.61)** <18.5 - 0.91 (0.28–3.00) - Note. *p < 0.05, **p < 0.01 compared to the reference group Data for underweight men’s group are not presented (includes one subject) Discussion In the present study, we focused on the description of LTPA among the Estonian population and adherence to WHO PA recommendations considering LTPA as well as on possible associations between LTPA and overweight. The study results revealed that 73 % of the study participants reported LTPA in the preceding 7 days, whereas 22 % (26 and 20 % of men and women, respectively) met WHO PA recommendations. Half of the study group participants had BMI > 25.0 kg/m2 and were thus considered overweight, and 20 % were obese. In the obesity group, considerably lower proportion of people reported LTPA and hence, the adherence to WHO PA recommendations was lower (in men as well as women) compared to the normal body weight group. The prevalence of study respondents reporting any LTPA in Estonia (73 %) is roughly similar with other European Union countries (73.1 %) [20] and the US (70.8 %) [21]. Although nearly three-fourths of the study participants’ reported LTPA, the amount of LTPA was rather modest, with median time per day for walking 11 min and moderate-vigorous LTPA 0 min. The latter means that at least half of subjects did not report any moderate-vigorous LTPA. Previous studies have demonstrated a wide variety in the LTPA amounts across countries. Calculations of energy expenditure for LTPA activities allow the comparisons between the countries. We acknowledge that the comparisons may be imprecise due to different methods used for LTPA assessment. Regardless, Estonians’ LTPA level (median 13.2 MET-hours/week) was, according to previously reported data, slightly higher than in southern European countries (<10 MET-h/week in Greece, Italy, Spain, and Portugal), in comparable range with the population of Germany (12.7 MET-h/week) and the United Kingdom (16.0 MET-h/week), but falling greatly behind the Netherlands (21.0 MET-h/week), Finland (21.5 MET-h/week), Austria (23.0 MET-h/week), and Sweden (24.0 MET-h/week) [20]. In previous studies, the north to south decline in LTPA has been described [20, 22], but despite the Nordic geographic location of Estonia, the LTPA is rather comparable to southern countries, implying that Estonian residents are physically more inactive than residents in neighbouring countries. The majority of the present study population reported LTPA. However, the amount, intensity and regularity sufficient to meet the PA recommendations for weekly PA was achieved only in a fifth of them, whereas compliance was somewhat higher in men compared to women. The finding in regard to the gender difference, of men being more active than women is in accordance with previous studies [16, 19, 20, 23]. The finding that higher proportions of those engaged in vigorous LTPA are mostly men and rather people in younger age groups is also compatible with previous results [20, 24]. Compared to Estonian data from 2014, whereby 35.8 % of men and 37.4 % of women reported participation in LTPA [10], the LTPA prevalence is lower in the current survey and more men than women were sufficiently active in LT. Suspected reasons of this discrepancy could be the difference in LTPA data collection on the one hand, and stricter criteria applied for PA recommendation adherence in the latter survey on the other hand. Many studies have reported the age-related decline in physical activity [19–21]. Among the present study population, the group of the youngest men differed distinctly with the highest prevalence of those meeting WHO PA recommendations, and among men, the gradual ageing-dependent decline in the prevalence of sufficiently active persons was demonstrated. In women, the same tendency was observed but since the prevalence of those meeting PA recommendations was low in the youngest age group, the decline did not seem as drastic as in men. In addition, considering the socio-demographic indices of the study population the vigorous LTPA was more often reported by people with higher education and higher income, whereas walking was more often reported by residents of smaller towns and villages. Regardless, these differences in LTPA patterns did not significantly affect the adherence to WHO PA recommendations in people with different levels of education and income or place of residence. However, in male skilled workers the prevalence of any LTPA and hence adherence to PA recommendations was considerably lower compared to the reference group of top managers. These findings are in concordance with previous studies [20, 23]. The probable reason for lower LTPA in men with a blue-collar occupation is the high level of work-related PA, which precludes the participation in LTPA. Hence, in subgroup of people with a blue-collar occupation the separately analysed LTPA may significantly underestimate the PA level and the assessment of work-related PA is essential. Concerning body weight, the inverse relation between LTPA and overweight is well documented [20, 23, 25, 26]. As an exception in the present study, men with BMI in the overweight range were equally active in LT compared to men with normal body weight. It could probably be explained by physically active men in this group having higher BMI due to larger muscle mass rather than a higher proportion of adipose tissue. According to our data, obesity was a factor that is associated with lower adherence to WHO PA recommendations. The current PA recommendations include only engagement in moderate and vigorous PA, but this could be hard to accomplish by obese and/or older people. Considering the LTPA pattern, walking was the most prevalent activity in the present study population, whereas moderate-vigorous LTPA was reported by a considerably lower proportion of respondents, particularly in overweight women, and obese men and women. The inverse correlation between BMI and vigorous PA (including LTPA) has also been shown in a Norwegian study [15]. According to Loprinzi et al. [27], there is growing evidence that light-intensity PA encompasses similar health enhancing effects as moderate-vigorous PA. Albeit the amount of time needed to achieve those effects is at least twice as much (≥300 min/week) [27] as the current PA recommendation, this could be more suitable for obese and older people. The strength of the present study is the stratified and randomly selected study population. In the final study group, there were more women than men and women were slightly older than men, but the study sample was representative of the Estonian population structure. Nevertheless, we acknowledge that the present study has several limitations. The BMI was calculated based on height and weight stated by the study respondents. The BMI calculated in this way may be overestimated although this effect is considered rather modest [28]. Typically, participants overestimate height and underestimate weight resulting in a BMI that is, on average, 1 kg/m2 lower than when height and weight are actually measured [29]. Thus, the proportion of overweight people could be even higher in the present study. Another limitation is the questionnaire-based evaluation of LTPA. Although the interview-based IPAQ-L version questionnaire has shown better validity than self-reported questionnaires [30], there is critical evidence in the literature about the overestimation of PA, especially in PA related to work and household domains [16, 17], which we also encountered in our data. Compared to other IPAQ-L domains, LTPA is considered the most valid measure in categorizing the population as physically active or sedentary [15, 17]. Recently published data in adults demonstrated a distinct discrepancy in PA level outcomes when different methods were used in PA evaluation, whereas self-reported PA showed systematic over-reporting of moderate-vigorous PA up to 10 times compared to accelerometer-based studies [31]. Hence, the over-reporting issue of self-reported PA with IPAQ-L cannot be overlooked. To overcome the over-report issues, stricter criteria for the calculation of the achievement of PA recommendations were applied in the present survey, where the accumulated time amount, intensity, as well the regularity of PA were taken into account. We acknowledge that the objectively measured PA using accelerometers would have given a more precise estimation of the PA level, but the main concerns in carrying out the present study were the cost and feasibility, and the objective PA measurement was not considered. Hence, before planning interventions for certain groups at the risk of physical inactivity, more precise PA monitoring than self-report should be taken into consideration. Conclusion In summary, the majority of the present study population reported LTPA. However, only fifth of them were sufficiently physically active to meet the WHO recommendations for weekly PA. Lower adherence to WHO PA recommendations was associated with older age in men, and obesity in both men and women. Although the purpose of the present descriptive study was not to explore the whole spectre of reasons influencing LTPA, the results of the present survey provide useful baseline data about LTPA and adherence to PA recommendations as well as associations between LTPA and overweight in the Estonian population. The strikingly low proportion of people in relation to sufficient LTPA demands future research in this area to promote increasing PA. Abbreviations BMI, body mass index; CVD, cardiovascular diseases; IPAQ-L, The International Physical Activity Questionnaire long form; LTPA, leisure time physical activity; MET, metabolic equivalent of task; PA, physical activity; TAPI, Tablet Assisted Personal Interviewing The authors wish to thank Helen Kaptein for excellent technical assistance. Funding This study was supported by Estonian Ministry of Culture and Ministry of Education and Research [institutional research funding No. IUT 02–7]. Availability of data and materials The dataset supporting the conclusions of this article is available in the Open Science Framework repository [https://osf.io/8wxfu/]. Authors’ contributions MT participated in data analysis and drafted the manuscript. PL participated in study design and coordination. EU performed the statistical analysis and helped to draft the manuscript. All authors read and approved the final manuscript. Competing interests The authors declare that they have no competing interests. Consent for publication Not applicable. Ethics approval and consent to participate Not applicable. The interviewing of study participants carried out by TNS EMOR was in accordance with ESOMAR/ICC international rules of conducting marketing and social surveys (http://www.esomar.org). According to the rules mentioned above, the ethics committee’s approval in not needed in population poll surveys in participants aged 15 years and older. ==== Refs References 1. World Health Organization Global Health Risks – mortality and burden of disease attributable to selected major risks 2009 2. World Health Organization Reducing risks, promoting healthy life World Health Report 2002 2002 Geneva World Health Organization 3. WHO: Global recommendations on physical activity for health. World Health Organization 2010, Geneva (Switzerland). http://whqlibdoc.who.int/publications/2010/9789241599979_eng.pdf. Accessed 19 Nov 2015. 4. Obesity: Preventing and managing the global epidemic. Report of a WHO consultation. Geneva, WHO 2000. http://www.who.int/nutrition/publications/obesity/WHO_TRS_894/en/. 5. Arem H Moore SC Patel A Hartage P de Gonzalez AB Visvanathan K Leisure time physical activity and mortality. A detailed pooled analysis of the dose–response relationship JAMA Intern Med 2015 175 959 67 10.1001/jamainternmed.2015.0533 25844730 6. Eesti südame – ja veresoonkonna haiguste preventsiooni juhised (in Estonian). 2006. http://www.ravijuhend.ee/downloader.php?file=guide-29-doctypetext27. 7. Townsend N Nichols M Scarborough P Rayner M Cardiovascular disease in Europe – epidemiological update 2015 Eur Heart J 2015 36 2698 705 10.1093/eurheartj/ehv428 8. Kohl HW 3rd Craig CL Lambert EV Inoue S Alkandari JR Leetongin G The pandemic of physical inactivity: global action for public health Lancet 2012 80 294 305 10.1016/S0140-6736(12)60898-8 22818941 9. European Commission, 2014. Special Eurobarometer 412. Sport and physical activity. European Commission, Directorate-General for Education and Culture and coordinated by Directorate-General for Communication, Brussels. http://ec.europa.eu/health/nutrition_physical_activity/docs/ebs_412_en.pdf. Accessed 19 Nov 201. 10. Tekkel M, Veideman T. Health behaviour among Estonian adult population, 2014. 2015. National Institute for Health Development, https://intra.tai.ee/images/prints/documents/14274488161_T2iskasvanud_rahvastiku_tervisek2itumise_uuring_2014.pdf. Accessed 15 Nov 2015. 11. Estonian Statistics. http://pub.stat.ee/px-web.2001/Dialog/varval.asp?ma=PO021&ti=POPULATION+BY+SEX+AND+AGE+GROUP%2C+1+JANUARY&path=../I_Databas/Population/01Population_indicators_and_composition/04Population_figure_and_composition/&lang=1. Accessed 12 Mar 2015. 12. International physical activity questionnaire. Guidelines for data processing and analysis of the international physical activity questionnaire (IPAQ) – short and long forms, http://www.ipaq.ki.se (accessed 16 November 2015). 13. Craig CL Marshall AL Sjöström M Bauman AE Booth ML Ainsworth BE International physical activity questionnaire: 12-country reliability and validity Med Sci Sports Exerc 2003 35 1381 95 10.1249/01.MSS.0000078924.61453.FB 12900694 14. Hagströmer M Oja P Sjöström M The International Physical Activity Questionnaire (IPAQ): a study of concurrent and construct validity Public Health Nutr 2006 9 755 62 10.1079/PHN2005898 16925881 15. Graff-Iversen S Anerssen SA Holme IM Jenum AK Raastad T An adapted version of the long International Physical Activity Questionnaire (IPAQ-L): construct validity in a low-income, multi-ethnic population study from Oslo, Norway Int J Behav Nutr Phys Act 2007 4 13 10.1186/1479-5868-4-13 17448225 16. Sebastião E Gobbi S Chdzko-Zaiko W Schwingel A Papini CB Nakamura PM The International Physical Activity Questionnaire-long form overestimates self-reported physical activity of Brazilian adults J Public Health 2012 126 967 75 10.1016/j.puhe.2012.07.004 17. Hallal PC Gomez LF Parra DC Lobelo F Mosquera J Florindo AA Lessons learned after 10 years of IPAQ use in Brazil and Columbia J Phys Act Health 2010 7 259 64 18. Grünberg H, Adojaan B, Thetloff M. Kasvamine ja kasvuhäired. Metoodiline juhend laste füüsilise arengu hindamiseks (in Estonian). Tartu 1998: Tartu Ülikooli Kirjastus. 19. Caban-Martinez AJ Lee DJ Fleming LE LeBlanc WG Arheart KL Chung-Bridges K Leisure-time physical activity levels of the US workforce Prev Med 2007 44 432 6 10.1016/j.ypmed.2006.12.017 17321584 20. Martinez-Gonzalez MA Varo JJ Santos JL De Irala J Gibney M Kearney J Prevalence of physical activity during leisure time in the European Union Med Sci Sports Exerc 2001 33 1142 6 10.1097/00005768-200107000-00011 11445761 21. Pratt M Macera CA Blanton C Levels of physical activity and inactivity in children and adults in the United States: current evidence and research issues Med Sci Sports Exerc 1999 31 526 33 10.1097/00005768-199911001-00007 22. Van Tuyckom C Macro-environmental factors associated with leisure-time physical activity: a cross-national analysis of EU countries Scand J Public Health 2011 39 419 26 10.1177/1403494810396553 21273227 23. Meseguer CM Galan I Herruzo R Zorrilla B Rodriguez-Artalejo F Leisure-time physical activity in a southern European Mediterranean country: adherence to recommendations and determining factors Rev Esp Cardiol 2009 62 1125 33 10.1016/S0300-8932(09)72381-4 19793518 24. Bauman A, Bull F, Chey T, Craig CL, Ainsworth BE, Sallis JF et al. The international prevalence study on physical activity: results from 20 countries. Int J Behav Nutr Phys Act. 2009;6. doi:10.1186/1479-5868-6-21 25. Tucker JM Welk GJ Beyler NK Physical Activity in U.S. Adults compliance with the physical activity guidelines for Americans Am J Prev Med 2011 40 454 61 10.1016/j.amepre.2010.12.016 21406280 26. Larsson I Lissner L Näslund I Lindroos AK Leisure and occupational physical activity in relation to body mass index in men and women Scand J Nutr 2004 48 165 72 10.1080/11026480410003785 27. Loprinzi PD Lee H Cardinal BJ Evidence to support including lifestyle light-intensity recommendations in physical activity guidelines for older adults Am J Health Promot 2015 29 277 84 10.4278/ajhp.130709-QUAN-354 24575724 28. de Berrington GA Hartge P Cerhan JR Flint AJ Hannan L MacInnis RJ Body-mass index and mortality among 1.46 million white adults N Engl J Med 2010 363 2211 9 10.1056/NEJMoa1000367 21121834 29. Physical status: the use and interpretation of anthropometry. Report of a WHO Expert Committee. World Health Organ Tech Rep Ser. 1995;854:1–452. http://www.who.int/childgrowth/publications/physical_status/en/. 30. Van Dyke D Gardon G Deforche B Bourdeaudhuij I IPAQ interview version: convergent validity with accelerometers and comparison of physical activity and sedentary time levels with the self-administered version J Sports Med Phys Fitness 2015 55 776 86 24921615 31. Van Holle V De Bourdeaudhuij I Deforche B Van Cauwenberg J Van Dyck D Assessment of physical activity in older Belgian adults: validity and reliability of an adapted interview version of the long International Physical Activity Questionnaire (IPAQ-L) BMC Public Health 2015 15 433 10.1186/s12889-015-1785-3 25928561

PMC005xxxxxx/PMC5002975.txt

==== Front Genes Dev Genes Dev genesdev genesdev GAD Genes & Development 0890-9369 1549-5477 Cold Spring Harbor Laboratory Press 27492367 8711660 10.1101/gad.282756.116 Research Communication The rate of protein synthesis in hematopoietic stem cells is limited partly by 4E-BPs Signer et al. 4E-BP hypophosphorylation in HSCs Signer Robert A.J. 12 Qi Le 1 Zhao Zhiyu 1 Thompson David 3 Sigova Alla A. 4 Fan Zi Peng 4 DeMartino George N. 3 Young Richard A. 45 Sonenberg Nahum 6 Morrison Sean J. 1 1 Howard Hughes Medical Institute, Children's Research Institute, Department of Pediatrics, University of Texas Southwestern Medical Center, Dallas, Texas 75390, USA; 2 Division of Regenerative Medicine, Department of Medicine, Moores Cancer Center, University of California at San Diego, La Jolla, California 92093, USA; 3 Department of Physiology, University of Texas Southwestern Medical Center, Dallas, Texas 75390, USA; 4 Whitehead Institute for Biomedical Research, Massachusetts Institute of Technology, Cambridge, Massachusetts 02142, USA; 5 Department of Biology, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA; 6 Department of Biochemistry, Goodman Cancer Centre, McGill University, Montreal, Quebec H3A 1A3, Canada Corresponding authors: sean.morrison@utsouthwestern.edu, rsigner@ucsd.edu 1 8 2016 1 8 2016 30 15 16981703 16 4 2016 18 7 2016 © 2016 Signer et al.; Published by Cold Spring Harbor Laboratory Press 2016 This article, published in Genes & Development, is available under a Creative Commons License (Attribution 4.0 International), as described at http://creativecommons.org/licenses/by/4.0/. Here, Signer et al. investigated the mechanism underlying the limited rate of protein synthesis in hematopoietic stem cells (HSCs). The authors found that adult HSCs had more hypophosphorylated eukaryotic initiation factor 4E-binding protein1 (4E-BP1) and 4E-BP2 as compared with most other hematopoietic progenitors, providing new insight into the mechanism by which HSCs attenuate protein synthesis. Adult stem cells must limit their rate of protein synthesis, but the underlying mechanisms remain largely unexplored. Differences in protein synthesis among hematopoietic stem cells (HSCs) and progenitor cells did not correlate with differences in proteasome activity, total RNA content, mRNA content, or cell division rate. However, adult HSCs had more hypophosphorylated eukaryotic initiation factor 4E-binding protein 1 (4E-BP1) and 4E-BP2 as compared with most other hematopoietic progenitors. Deficiency for 4E-BP1 and 4E-BP2 significantly increased global protein synthesis in HSCs, but not in other hematopoietic progenitors, and impaired their reconstituting activity, identifying a mechanism that promotes HSC maintenance by attenuating protein synthesis. 4E-BP protein synthesis stem cell Howard Hughes Medical Institute http://dx.doi.org/10.13039/100000011National Institutes of Health http://dx.doi.org/10.13039/100000002National Institute of Diabetes and Digestive and Kidney Diseases http://dx.doi.org/10.13039/100000062R01 DK100848 Cancer Prevention and Research Institute of Texas http://dx.doi.org/10.13039/100004917 ==== Body Adult hematopoietic stem cell (HSC) maintenance depends on these cells making less protein per hour as compared with other hematopoietic progenitors (Signer et al. 2014). This does not reflect merely HSC quiescence, as dividing HSCs also make less protein per hour as compared with other hematopoietic progenitors. Several other adult stem cells also exhibit lower rates of protein synthesis as compared with progenitors in the same tissues, including neural stem cells (Llorens-Bobadilla et al. 2015), skeletal muscle stem cells (Zismanov et al. 2016), hair follicle stem cells (Blanco et al. 2016), and Drosophila germline stem cells (Sanchez et al. 2016). Like HSCs, each of these stem cells is depleted by genetic changes that increase their rate of protein synthesis, suggesting that this is a widely shared property of adult stem cells. However, there is limited insight into the mechanisms that suppress protein synthesis in stem cells. Ribosome biogenesis limits protein synthesis in some stem cells. Quiescent neural stem cells express lower levels of ribosomal subunits and synthesize protein at a lower rate as compared with activated neural stem cells (Llorens-Bobadilla et al. 2015). Increased ribosome biogenesis in Drosophila germline stem cells increases protein synthesis, promoting differentiation and the loss of stem cells (Sanchez et al. 2016). Replication stress in HSCs from old mice transcriptionally silences ribosome genes, reducing ribosome biogenesis and perhaps impairing HSC function during aging (Flach et al. 2014). Runx1-deficient HSCs have impaired ribosome biogenesis, lower protein synthesis, and altered function (Cai et al. 2015). Defects in ribosome biogenesis impair HSC function (Le Bouteiller et al. 2013). Phosphorylated eukaryotic initiation factor 2α (eIF2α) inhibits translation initiation and is part of the mechanism that limits protein synthesis in some stem cells. eIF2α can be phosphorylated under steady-state circumstances, although phosphorylation is increased by a wide range of stresses, including unfolded protein responses (Wek et al. 2006). Protein synthesis in skeletal muscle stem cells is limited by high levels of phosphorylated eIF2α, which promotes quiescence and stem cell maintenance (Zismanov et al. 2016). However, eIF2α function may vary among stem cells, as HSCs express relatively low levels of phosphorylated eIF2α (Signer et al. 2014), and activation of the unfolded protein response can be deleterious to individual HSCs (van Galen et al. 2014). mTORC1 signaling promotes protein synthesis through phosphorylation and activation of ribosomal protein S6 kinase 1 (Brown et al. 1995) and phosphorylation and inhibition of eIF4E-binding proteins (4E-BPs) (Beretta et al. 1996; Brunn et al. 1997). Three genes (Eif4ebp1, Eif4ebp2, and Eif4ebp3) encode 4E-BPs that negatively regulate translation by binding the cap-binding protein eIF4E, inhibiting eIF4G binding, and preventing eIF4F complex assembly (eIF4E–eIF4G–eIF4A). Phosphorylation of 4E-BPs by mTORC1 weakens their binding to eIF4E, promoting eIF4F assembly and cap-dependent translation (Pause et al. 1994; Gingras et al. 1999). 4E-BP levels vary among tissues. Hematopoietic cells express abundant 4E-BP1 and 4E-BP2 but not 4E-BP3 (Tsukiyama-Kohara et al. 2001). 4E-BP-mediated translation inhibition does not generally affect global translation rates but particularly impairs the translation of certain subsets of mRNAs (Morita et al. 2013; Tahmasebi et al. 2014). Conditional Pten deletion increases mTORC1 signaling in HSCs (Lee et al. 2010), increasing 4E-BP phosphorylation (Magee et al. 2012), elevating protein synthesis, and promoting HSC depletion (Yilmaz et al. 2006; Signer et al. 2014). It is not clear whether 4E-BPs regulate HSC function, although knockdown of 4E-BP2 (Hartman et al. 2013) or eIF4E1 (Yang et al. 2014) in neural progenitors increases protein synthesis and promotes premature neuronal differentiation. The synthesis of O-propargyl-puromycin (OP-Puro) has facilitated the quantification of protein synthesis in individual cells in vivo (Liu et al. 2012). OP-Puro enters the acceptor site of ribosomes and is incorporated into nascent polypeptide chains. The amount of protein synthesis per hour in individual cells in vivo can then be quantified based on OP-Puro incorporation (Liu et al. 2012; Signer et al. 2014). In this study, we examined potential mechanisms that might influence differences in protein synthesis between HSCs and restricted hematopoietic progenitors in adult mouse bone marrow. Results and Discussion HSCs do not have high proteasome activity We administered OP-Puro to mice, killed them 1 h later, and compared OP-Puro incorporation by subpopulations of bone marrow cells by flow cytometry. Consistent with our prior study (Signer et al. 2014), we observed significantly less OP-Puro incorporation into CD150+CD48−Lineage− Sca-1+c-kit+ (CD150+CD48−LSK) HSCs (Kiel et al. 2005) and CD150−CD48−LSK multipotent progenitor cells (MPPs) (Oguro et al. 2013) as compared with restricted hematopoietic progenitors (Fig. 1A). Figure 1. Protein synthesis rates in HSCs and progenitor cells show little correlation with proteasome activity, frequency of cell division, or RNA content. (A) OP-Puro incorporation by HSCs and progenitor cells in vivo. n = 8 mice in eight experiments. (B) Proteasome activity in HSCs and progenitor cells normalized to unfractionated bone marrow cells. n = 3 experiments. (C) Mean protein synthesis rate (from A) plotted against relative proteasome activity (from B). (D) The frequency of cell division for each population based on the rate of 5-ethynyl-2′-deoxyuridine (EdU) incorporation in vivo (see Supplemental Fig. S1B–M for primary data). n = 5 mice in two experiments. (E) Mean protein synthesis (from A) plotted against the rate of cell division (from D). (F) Gene expression values for all mouse RefSeq genes in each cell population (in total reads per kilobase of exonic length per million mapped reads [RPKM]). Log2 fold change of gene expression in each sample is shown relative to the average values in HSCs. (G) Total RPKMs of all mouse RefSeq genes in each cell population. Two-sided paired Wilcoxon rank-sum test was used for pair-wise comparisons between HSCs and each cell type. (#) P < 2.2 × 10−16. (H) Mean protein synthesis (from A) plotted against median total RPKMs (from G). Data represent mean ± SD unless indicated otherwise. The statistical significance of differences relative to HSCs in A, B, and D were assessed using a repeated-measures one-way analysis of variance (ANOVA) followed by Dunnett's test for multiple comparisons. (*) P < 0.05; (**) P < 0.01; (***) P < 0.001. Regression analyses in C, E, and F were performed excluding HSCs, which were plotted independently. Ninety-five percent confidence intervals (dashed lines) and Pearson's correlation coefficients (R) are shown. To test whether HSCs have high proteasome activity like embryonic stem cells (Vilchez et al. 2013), we compared the proteasome activity of HSCs versus restricted progenitors. We sorted 30,000 unfractionated bone marrow cells, CD48−LSK cells (HSCs/MPPs), CD34+CD16/32lowCD127−Sca-1−LK common myeloid progenitors (CMPs) (Akashi et al. 2000), CD34+CD16/32highCD127− Sca-1−LK granulocyte macrophage progenitors (GMPs) (Akashi et al. 2000), Gr-1+ myeloid, IgM−CD43+B220+ pro-B, IgM−CD43−B220+ pre-B (Hardy et al. 1991), and B220+IgM+ B cells. Proteasome activity in HSCs/MPPs was similar to CMPs and pro-B cells but significantly lower than unfractionated bone marrow cells, GMPs, and Gr-1+ cells (Fig. 1B; Supplemental Fig. S1A). Overall, proteasome activity showed little correlation with OP-Puro incorporation (R = 0.17, P = 0.69) (Fig. 1C). Little correlation between protein synthesis and cell division or RNA content We administered the thymidine analog 5-ethynyl-2′-deoxyuridine (EdU) to mice and measured its incorporation by HSCs and restricted progenitors in the bone marrow after 2, 6, and 12–24 h to determine the frequency of cell division in each population (Fig. 1D). We observed a reasonably linear increase in the fraction of EdU+ cells over time in each population (Supplemental Fig. S1B–M). We observed a modest correlation between protein synthesis rates and cell division (R = 0.42, P = 0.23) (Fig. 1E). However, when we drove HSCs into cycle by treating them with cyclophosphamide and GCSF, we observed only a limited increase in protein synthesis (Fig. 1E). Therefore, the rate of protein synthesis in HSCs was not determined primarily by the rate at which they divide. To test whether protein synthesis rates reflected total RNA content (largely ribosomal RNA), we quantified total RNA by nanofluidic electrophoresis using a BioAnalyzer. HSCs had amounts of total RNA similar to those of CMPs and pro-B cells and significantly more total RNA per cell than Gr-1+, pre-B, and unfractionated bone marrow cells (Supplemental Fig. S1N), each of which had higher protein synthesis rates than HSCs (Fig. 1A). Therefore, we observed little correlation between protein synthesis rates and total RNA content per cell (R = 0.32, P = 0.44) (Supplemental Fig. S1O). We quantified the amount of mRNA per cell by RNA sequencing (RNA-seq) after adding RNA standards that allow normalization of transcript numbers to cell number (Fig. 1F; Loven et al. 2012). On average, HSCs had significantly more mRNA per cell as compared with pre-B cells, unfractionated bone marrow, and Gr-1+ cells but less mRNA per cell than CMPs, GMPs, and pro-B cells (Fig. 1G), all of which had higher protein synthesis rates than HSCs (Fig. 1A). Therefore, we observed little correlation between protein synthesis rates and mRNA content in hematopoietic cells (R = 0.48, P = 0.28) (Fig. 1H). HSCs have relatively high levels of hypophosphorylated 4E-BPs We assessed the levels of total, phosphorylated, and nonphosphorylated 4E-BP1 and 4E-BP2 in 30,000 sorted cells from each population (Fig. 2A). Control experiments using peripheral blood from 4E-BP1−/− (Tsukiyama-Kohara et al. 2001), 4E-BP2−/− (Banko et al. 2005), and 4E-BP1−/−; 4E-BP2−/− mice (Le Bacquer et al. 2007) showed that the upper band was 4E-BP1, and the lower band was 4E-BP2 (Supplemental Fig. S2A,B). HSCs/MPPs expressed relatively high levels of total 4E-BP1 and 4E-BP2 as compared with most hematopoietic progenitors—significantly higher than Gr-1+, pro-B, pre-B, and IgM+ B cells (Fig. 2A–C). Levels of nonphosphorylated 4E-BP1 and 4E-BP2 were also significantly higher in HSCs/MPPs as compared with Gr-1+, pro-B, pre-B, and IgM+ B cells (Fig. 2D,E). The ratio of phosphorylated to total 4E-BP1/2 tended to be lower in HSCs as compared with restricted progenitors (Supplemental Fig. S2C,D). Differences in β-actin levels among cell populations represented differences in the amount of β-actin per cell, as lysates from equal numbers of cells were used in these experiments (Fig. 2A). Nonetheless, even when levels of total and nonphosphorylated 4E-BP1 and 4E-BP2 were normalized to differences in β-actin, they tended to remain higher in HSCs/MPPs as compared with Gr-1+, pro-B, pre-B, and IgM+ B cells (although not all of the differences were statistically significant) (Supplemental Fig. S2E–H). Figure 2. HSCs tend to have high levels of hypophosphorylated 4E-BPs relative to most other hematopoietic progenitors. (A) Western blot of 3 × 104 cells from each cell population (one of four representative blots). Band intensity was normalized by cell number for each population. For phosphorylated and nonphosphorylated 4E-BP, the upper band is 4E-BP1, and the lower band is 4E-BP2 (see Supplemental Fig. S2A,B). (B–E) Quantification of band intensity for total 4E-BP1 (B), total 4E-BP2 (C), nonphosphorylated 4E-BP1 (D), and nonphosphorylated 4E-BP2 (E). (F,G) Mean intensity of phosphorylated S6 plotted against phosphorylated 4E-BP1 (F) or phosphorylated 4E-BP2 (G) levels for each cell population. (H,I) Mean intensity of phosphorylated Akt plotted against phosphorylated 4E-BP1 (H) or phosphorylated 4E-BP2 (I). (J,K) Mean protein synthesis (from Fig. 1A) plotted against phosphorylated 4E-BP1 (J) or 4E-BP2 (K) levels normalized to β-Actin. The regressions exclude HSCs/MPPs, which were plotted independently. Data in B–K represent mean ± SD from four experiments, including the one in A. The statistical significance of differences between HSCs/MPPs and other cell populations in B–E were assessed with one-way ANOVAs followed by Dunnett's test for multiple comparisons. (*) P < 0.05; (**) P < 0.01; (***) P < 0.001. For F–K, Pearson's correlation coefficient (R) and 95% confidence intervals (dashed lines) are shown. Levels of phosphorylated S6 strongly correlated with levels of phosphorylated 4E-BP1 (R = 0.88, P = 0.004) (Fig. 2F) or phosphorylated 4E-BP2 (R = 0.95, P = 0.0003) (Fig. 2G). However, correlations between phosphorylated Akt and phosphorylated 4E-BP1 (R = 0.56, P = 0.15) (Fig. 2H) or phosphorylated 4E-BP2 (R = 0.50, P = 0.21) (Fig. 2I) were weaker. These data suggest that differences in 4E-BP phosphorylation among HSCs and progenitor cells likely reflect differences in mTORC1 signaling. Levels of phosphorylated 4E-BP1 (R = 0.73, P = 0.06) (Fig. 2J) and phosphorylated 4E-BP2 (R = 0.57, P = 0.18) (Fig. 2K) normalized to β-actin exhibited stronger correlations with protein synthesis rates than any other parameter that we tested. Levels of phosphorylated 4E-BP1 (R = 0.40, P = 0.37) (Supplemental Fig. S2I) and phosphorylated 4E-BP2 (R = 0.24, P = 0.6) (Supplemental Fig. S2J) that were normalized based on cell numbers did not exhibit as strong a correlation with protein synthesis. To test whether 4E-BPs limit the rate of protein synthesis in HSCs, we examined OP-Puro incorporation by HSCs and progenitors in 4E-BP1−/−, 4E-BP2−/−, and 4E-BP1−/−; 4E-BP2−/− mice. 4E-BP1−/− or 4E-BP2−/− HSCs did not exhibit significant changes in OP-Puro incorporation compared with wild-type HSCs (Fig. 3A). However, 4E-BP1−/−; 4E-BP2−/− HSCs exhibited a significant increase in protein synthesis as compared with wild-type HSCs (23%, P < 0.05) (Fig. 3A). The magnitude of this increase was close to that observed after Pten deletion, which leads to HSC depletion (Signer et al. 2014). We did not detect a statistically significant effect of 4E-BP1 deficiency, 4E-BP2 deficiency, or combined 4E-BP1;4E-BP2 deficiency on the rate of protein synthesis in other hematopoietic progenitors (Fig. 3B). HSCs thus appear to be more sensitive to the effects of 4E-BP1/2 on global protein synthesis. Figure 3. 4E-BP1/2-deficient mice have HSCs with increased protein synthesis but largely normal hematopoiesis. (A) Mean OP-Puro incorporation by HSCs from 4E-BP1−/−, 4E-BP2−/−, 4E-BP1−/−;4E-BP2−/−, or wild-type mice. n = 9 mice per genotype in nine experiments. The value relative to wild type is shown above each bar. (B) Mean OP-Puro incorporation into each restricted progenitor population from the same mice as in A. (C–E) Number of white blood cells (WBC), red blood cells (RBC), and platelets (PLT) in the peripheral blood. (F) Cell number in the bone marrow (BM; one femur and one tibia), spleen (SPL), and thymus (Thy). (G–I) Frequencies of MPPs, CMPs, GMPs, and megakaryocyte erythroid-restricted progenitors (MEPs) (G); Gr-1+ myeloid and Ter119+ erythroid cells (H); and pro-B, pre-B, and IgM+ B cells (I) in the bone marrow. (J) Frequencies of CD4+ and CD8+ T cells in the spleen. (K) Frequency of HSCs in the bone marrow. n = 6 mice per genotype in three experiments in C–K. (L) Frequency of BrdU+ HSCs after a 72-h pulse of BrdU in vivo. (M) Frequency of annexin V+ HSCs. n = 5–6 mice per genotype in four experiments in L and M. All data represent mean ± SD. The statistical significance of differences relative to wild type was assessed using one-way ANOVAs (repeated-measures in A and B) followed by Dunnett's test for multiple comparisons. (*) P < 0.05; (**) P < 0.01; (***) P < 0.001. 4E-BP1−/−;4E-BP2−/− HSCs have impaired regenerative activity 4E-BP1−/−, 4E-BP2−/−, and 4E-BP1−/−;4E-BP2−/− compound mutant mice all exhibited relatively normal hematopoiesis (Fig. 3C–J). HSC frequency appeared to be higher in 4E-BP2−/− (65% ± 32%, P < 0.05) (Fig. 3K) and 4E-BP1−/−; 4E-BP2−/− (80% ± 53%, P < 0.01) mutant mice (Fig. 3K) as compared with controls. This modest increase in HSC frequency was not associated with significant changes in the frequency of dividing (Fig. 3L) or dying (Fig. 3M) HSCs. This suggests that the increase in HSC frequency may have been caused by a lower rate of differentiation by 4E-BP1/2-deficient HSCs. We also observed modest but significant declines in the frequencies of CMPs, Ter119+ erythroid progenitors, and pro-B cells in the bone marrow of some 4E-BP-deficient mice as compared with controls (Fig. 3G–I). To test the effect of 4E-BP deficiency on HSC function, we performed competitive reconstitution assays in irradiated mice. Bone marrow cells (5 × 105 cells) from 4E-BP1−/−, 4E-BP2−/−, 4E-BP1−/−;4E-BP2−/−, or wild-type mice (all CD45.2+) were transplanted with equal numbers of wild-type recipient bone marrow cells (CD45.1+) into irradiated mice (CD45.1+). Recipients of 4E-BP-deficient bone marrow cells had significantly higher levels of donor-derived hematopoietic cells in their blood 4–16 wk after transplantation (Fig. 4A–D). To test whether this increased reconstitution reflected increased HSC frequency in the bone marrow of 4E-BP-deficient donors (Fig. 3K) or increased function of 4E-BP-deficient HSCs, we competitively transplanted 10 CD150+CD48−LSK HSCs from 4E-BP1−/−;4E-BP2−/− or wild-type donors into irradiated recipient mice along with recipient bone marrow cells. 4E-BP-deficient HSCs gave normal levels of donor cell reconstitution (Fig. 4E–H). The higher level of donor cell reconstitution by 4E-BP-deficient bone marrow cells thus reflects higher HSC frequency in the donor bone marrow rather than increased reconstituting activity by 4E-BP-deficient HSCs. Figure 4. 4E-BP1−/−;4E-BP2−/− HSCs have impaired reconstituting potential upon serial transplantation. (A–D) Donor cell engraftment when 5 × 105 bone marrow cells from mice of the indicated genotypes were transplanted along with 5 × 105 recipient bone marrow cells into irradiated mice. Donor cell engraftment in the bone marrow is shown in Supplemental Figure S3A. (E–H) Donor cell engraftment when 10 HSCs from mice of the indicated genotypes were transplanted with 3 × 105 recipient bone marrow cells into irradiated mice. (I–L) Donor cell engraftment after serial transplantation of 3 × 106 bone marrow cells from primary recipients in A–D into secondary recipient mice. n = 8–11 recipients per genotype from two primary donors from wild type and three primary donors from each of the 4E-BP-deficient genotypes. Data represent mean ± SD. The statistical significance of differences relative to wild-type were assessed with one-way ANOVAs followed by Dunnett's test for multiple comparisons in A–D and I–L and a two-tailed Student's t-test in E–H. (*) P < 0.05; (**) P < 0.01; (***) P < 0.001. Although 4E-BP2−/− and 4E-BP1−/−;4E-BP2−/− mutant mice had increased HSC frequency in their bone marrow as compared with controls (Fig. 3K), the frequency of donor HSCs in primary recipient mice after bone marrow transplantation did not significantly differ between recipients of 4E-BP-deficient cells as compared with control cells (Supplemental Fig. S3B). These data raised the possibility that 4E-BP deficiency may impair HSC self-renewal after transplantation. To assess HSC self-renewal potential, we performed secondary transplants of 3 × 106 bone marrow cells from primary recipient mice with levels of donor cell reconstitution nearest the median values in each treatment. Deficiency for only 4E-BP1 or 4E-BP2 did not significantly affect donor cell reconstitution in secondary recipient mice as compared with control cells. In contrast, secondary recipients of 4E-BP1−/−;4E-BP2−/− cells had significantly less donor cell reconstitution in all lineages 16 wk after transplantation as compared with secondary recipients of wild-type donor cells (Fig. 4I–L). These data suggest that 4E-BP1−/−;4E-BP2−/− HSCs have reduced self-renewal potential as compared with control HSCs. 4E-BP1 and 4E-BP2 likely regulate the translation of a subset of mRNAs in HSCs. Increased expression of these mRNAs likely increases HSC frequency (Fig. 3K) and reconstituting capacity in the bone marrow of 4E-BP1/2−/− mice (Fig. 4A–D) while reducing HSC self-renewal upon serial transplantation (Fig. 4I–L). Nonetheless, other mechanisms must also limit protein synthesis in HSCs, as protein synthesis rates in 4E-BP1/2−/− HSCs remained significantly lower than in 4E-BP1/2−/− restricted hematopoietic progenitors. Moreover, both GMPs and megakaryocyte erythroid-restricted progenitors (MEPs) had higher levels of hypophosphorylated 4E-BPs relative to HSCs (Fig. 2D,E) and yet still had higher levels of global protein synthesis (Fig. 1A), perhaps because global protein synthesis in these cells is not as sensitive to 4E-BP1/2 as in HSCs (Fig. 3B). Materials and methods Mice Eif4ebp1−/− (Tsukiyama-Kohara et al. 2001), Eif4ebp2−/− (Banko et al. 2005), and Eif4ebp1−/−;Eif4ebp2−/− (Le Bacquer et al. 2007) mice have been described previously. These mice were backcrossed for at least 10 generations onto a C57BL background. C57BL/Ka-Thy-1.1 (CD45.2) mice were used as wild type throughout this study. C57BL/Ka-Thy-1.2 (CD45.1) mice were used as transplant recipients. Male and female mice between 8 and 14 wk old were used in all studies. Cyclophosphamide and GCSF were administered as described (Signer et al. 2014). All mice were housed in the Animal Resource Center at the University of Texas Southwestern Medical Center, and all protocols were approved by the University of Texas Southwestern Medical Center Institutional Animal Care and Use Committee. Proteasome activity assay Proteasome activity was determined by measuring the rate of hydrolysis of Suc–Leu–Leu–Val–Tyr–7-amino-4-methylcoumarin (AMC). This substrate is specific for the proteasome under the assay conditions used here and is hydrolyzed by all proteasome holoenzymes. The activity reported by this assay represents the total enzymatic capacity of the cellular proteasome and is typically dominated by the 26S proteasome (Sasaki et al. 1984; Demartino and Gillette 2007). From each population, 3 × 104 cells were double-sorted into 96-well black plates. One-hundred microliters of buffer containing 20 mM Tris-HCl (pH 7.6), 1 mM 2-mercaptoethanol, 5 mM MgCl2, 1 mM ATP, and 0.4% NP-40 was added to each well followed immediately by 50 µL of 200 µM Suc–Leu–Leu–Val–Tyr–AMC (Bachem). The fluorescence of free AMC generated by proteasomal hydrolysis of the substrate was monitored continuously (one read per minute for 180 min at 37°C) at 360ex/460em using a BioTek Synergy II plate reader. Controls included reactions conducted in the absence of cells, in the presence of 10 mM MG132 (UBPBio), or using purified bovine 26S proteasome instead of cells. Data are expressed as arbitrary fluorescent units per well. The rate of cell division Two milligrams of EdU (Thermo Scientific) in PBS was injected intraperitoneally per mouse every 6 h. Mice were analyzed 2, 6, 12, and 24 h after EdU administration. These short labeling times minimized the extent to which results were affected by the entry of labeled cells into each population through the maturation of upstream progenitors, the exit of labeled cells through differentiation into downstream progeny, or saturation of labels in rapidly dividing cell populations. Bone marrow cells (3 × 106 cells) were stained with antibodies as described in the Supplemental Material. Cells were fixed and permeabilized, and the azide–alkyne cycloaddition was performed as with OP-Puro detection and analyzed by flow cytometry. For cyclophosphamide- and GCSF-treated mice, BrdU was administered in place of EdU and was detected with the BrdU flow kit. Supplementary Material Supplemental Material Acknowledgments We thank K. Correll, N. Loof, and the Moody Foundation Flow Cytometry Facility. S.J.M. is a Howard Hughes Medical Institute Investigator, the Mary McDermott Cook Chair in Pediatric Genetics, the Kathryn and Gene Bishop Distinguished Chair in Pediatric Research, and the director of the Hamon Laboratory for Stem Cells and Cancer. L.Q. is supported by a fellowship from the Howard Hughes Medical Institute. This work was supported by the National Institutes of Health (National Institute of Diabetes and Digestive and Kidney Diseases R01 DK100848) and the Cancer Prevention and Research Institute of Texas. Supplemental material is available for this article. Article published online ahead of print. Article and publication date are online at http://www.genesdev.org/cgi/doi/10.1101/gad.282756.116. Freely available online through the Genes & Development Open Access option. ==== Refs References Akashi K, Traver D, Miyamoto T, Weissman IL. 2000 A clonogenic common myeloid progenitor that gives rise to all myeloid lineages. Nature 404 : 193–197.10724173 Banko JL, Poulin F, Hou L, DeMaria CT, Sonenberg N, Klann E. 2005 The translation repressor 4E-BP2 is critical for eIF4F complex formation, synaptic plasticity, and memory in the hippocampus. J Neurosci 25 : 9581–9590.16237163 Beretta L, Gingras AC, Svitkin YV, Hall MN, Sonenberg N. 1996 Rapamycin blocks the phosphorylation of 4E-BP1 and inhibits cap-dependent initiation of translation. EMBO J 15 : 658–664.8599949 Blanco S, Bandiera R, Popis M, Hussain S, Lombard P, Aleksic J, Sajini A, Tanna H, Cortes-Garrido R, Gkatza N, 2016 Stem cell function and stress response are controlled by protein synthesis. Nature 534 : 335–340.27306184 Brown EJ, Beal PA, Keith CT, Chen J, Shin TB, Schreiber SL. 1995 Control of p70 s6 kinase by kinase activity of FRAP in vivo. Nature 377 : 441–446.7566123 Brunn GJ, Hudson CC, Sekulic A, Williams JM, Hosoi H, Houghton PJ, Lawrence JCJr, Abraham RT. 1997 Phosphorylation of the translational repressor PHAS-I by the mammalian target of rapamycin. Science 277 : 99–101.9204908 Cai X, Gao L, Teng L, Ge J, Oo ZM, Kumar AR, Gilliland DG, Mason PJ, Tan K, Speck NA. 2015 Runx1 deficiency decreases ribosome biogenesis and confers stress resistance to hematopoietic stem and progenitor cells. Cell Stem Cell 17 : 165–177.26165925 Demartino GN, Gillette TG. 2007 Proteasomes: machines for all reasons. Cell 129 : 659–662.17512401 Flach J, Bakker ST, Mohrin M, Conroy PC, Pietras EM, Reynaud D, Alvarez S, Diolaiti ME, Ugarte F, Forsberg EC, 2014 Replication stress is a potent driver of functional decline in ageing haematopoietic stem cells. Nature 512 : 198–202.25079315 Gingras AC, Gygi SP, Raught B, Polakiewicz RD, Abraham RT, Hoekstra MF, Aebersold R, Sonenberg N. 1999 Regulation of 4E-BP1 phosphorylation: a novel two-step mechanism. Genes Dev 13 : 1422–1437.10364159 Hardy RR, Carmack CE, Shinton SA, Kemp JD, Hayakawa K. 1991 Resolution and characterization of pro-B and pre-pro-B cell stages in normal mouse bone marrow. J Exp Med 173 : 1213–1225.1827140 Hartman NW, Lin TV, Zhang L, Paquelet GE, Feliciano DM, Bordey A. 2013 mTORC1 targets the translational repressor 4E-BP2, but not S6 kinase 1/2, to regulate neural stem cell self-renewal in vivo. Cell Rep 5 : 433–444.24139800 Kiel MJ, Yilmaz OH, Iwashita T, Yilmaz OH, Terhorst C, Morrison SJ. 2005 SLAM family receptors distinguish hematopoietic stem and progenitor cells and reveal endothelial niches for stem cells. Cell 121 : 1109–1121.15989959 Le Bacquer O, Petroulakis E, Paglialunga S, Poulin F, Richard D, Cianflone K, Sonenberg N. 2007 Elevated sensitivity to diet-induced obesity and insulin resistance in mice lacking 4E-BP1 and 4E-BP2. J Clin Invest 117 : 387–396.17273556 Le Bouteiller M, Souilhol C, Beck-Cormier S, Stedman A, Burlen-Defranoux O, Vandormael-Pournin S, Bernex F, Cumano A, Cohen-Tannoudji M. 2013 Notchless-dependent ribosome synthesis is required for the maintenance of adult hematopoietic stem cells. J Exp Med 210 : 2351–2369.24062412 Lee JY, Nakada D, Yilmaz OH, Tothova Z, Joseph NM, Lim MS, Gilliland DG, Morrison SJ. 2010 mTOR activation induces tumor suppressors that inhibit leukemogenesis and deplete hematopoietic stem cells after Pten deletion. Cell Stem Cell 7 : 593–605.21040901 Liu J, Xu Y, Stoleru D, Salic A. 2012 Imaging protein synthesis in cells and tissues with an alkyne analog of puromycin. Proc Natl Acad Sci 109 : 413–418.22160674 Llorens-Bobadilla E, Zhao S, Baser A, Saiz-Castro G, Zwadlo K, Martin-Villalba A. 2015 Single-cell transcriptomics reveals a population of dormant neural stem cells that become activated upon brain injury. Cell Stem Cell 17 : 329–340.26235341 Loven J, Orlando DA, Sigova AA, Lin CY, Rahl PB, Burge CB, Levens DL, Lee TI, Young RA. 2012 Revisiting global gene expression analysis. Cell 151 : 476–482.23101621 Magee JA, Ikenoue T, Nakada D, Lee JY, Guan KL, Morrison SJ. 2012 Temporal changes in PTEN and mTORC2 regulation of hematopoietic stem cell self-renewal and leukemia suppression. Cell Stem Cell 11 : 415–428.22958933 Morita M, Gravel SP, Chenard V, Sikstrom K, Zheng L, Alain T, Gandin V, Avizonis D, Arguello M, Zakaria C, 2013 mTORC1 controls mitochondrial activity and biogenesis through 4E-BP-dependent translational regulation. Cell Metab 18 : 698–711.24206664 Oguro H, Ding L, Morrison SJ. 2013 SLAM family markers resolve functionally distinct subpopulations of hematopoietic stem cells and multipotent progenitors. Cell Stem Cell 13 : 102–116.23827712 Pause A, Belsham GJ, Gingras AC, Donze O, Lin TA, Lawrence JCJr, Sonenberg N. 1994 Insulin-dependent stimulation of protein synthesis by phosphorylation of a regulator of 5′-cap function. Nature 371 : 762–767.7935836 Sanchez CG, Teixeira FK, Czech B, Preall JB, Zamparini AL, Seifert JR, Malone CD, Hannon GJ, Lehmann R. 2016 Regulation of ribosome biogenesis and protein synthesis controls germline stem cell differentiation. Cell Stem Cell 18 : 276–290.26669894 Sasaki T, Kikuchi T, Yumoto N, Yoshimura N, Murachi T. 1984 Comparative specificity and kinetic studies on porcine calpain I and calpain II with naturally occurring peptides and synthetic fluorogenic substrates. J Biol Chem 259 : 12489–12494.6092335 Signer RA, Magee JA, Salic A, Morrison SJ. 2014 Haematopoietic stem cells require a highly regulated protein synthesis rate. Nature 509 : 49–54.24670665 Tahmasebi S, Alain T, Rajasekhar VK, Zhang JP, Prager-Khoutorsky M, Khoutorsky A, Dogan Y, Gkogkas CG, Petroulakis E, Sylvestre A, 2014 Multifaceted regulation of somatic cell reprogramming by mRNA translational control. Cell Stem Cell 14 : 606–616.24630793 Tsukiyama-Kohara K, Poulin F, Kohara M, DeMaria CT, Cheng A, Wu Z, Gingras AC, Katsume A, Elchebly M, Spiegelman BM, 2001 Adipose tissue reduction in mice lacking the translational inhibitor 4E-BP1. Nat Med 7 : 1128–1132.11590436 van Galen P, Kreso A, Mbong N, Kent DG, Fitzmaurice T, Chambers JE, Xie S, Laurenti E, Hermans K, Eppert K, 2014 The unfolded protein response governs integrity of the haematopoietic stem-cell pool during stress. Nature 510 : 268–272.24776803 Vilchez D, Boyer L, Lutz M, Merkwirth C, Morantte I, Tse C, Spencer B, Page L, Masliah E, Berggren WT, 2013 FOXO4 is necessary for neural differentiation of human embryonic stem cells. Aging Cell 12 : 518–522.23551888 Wek RC, Jiang HY, Anthony TG. 2006 Coping with stress: eIF2 kinases and translational control. Biochem Soc Trans 34 : 7–11.16246168 Yang G, Smibert CA, Kaplan DR, Miller FD. 2014 An eIF4E1/4E-T complex determines the genesis of neurons from precursors by translationally repressing a proneurogenic transcription program. Neuron 84 : 723–739.25456498 Yilmaz OH, Valdez R, Theisen BK, Guo W, Ferguson DO, Wu H, Morrison SJ. 2006 Pten dependence distinguishes haematopoietic stem cells from leukaemia-initiating cells. Nature 441 : 475–482.16598206 Zismanov V, Chichkov V, Colangelo V, Jamet S, Wang S, Syme A, Koromilas AE, Crist C. 2016 Phosphorylation of eIF2α is a translational control mechanism regulating muscle stem cell quiescence and self-renewal. Cell Stem Cell 18 : 79–90.26549106

PMC005xxxxxx/PMC5002978.txt

==== Front Genes Dev Genes Dev genesdev genesdev GAD Genes & Development 0890-9369 1549-5477 Cold Spring Harbor Laboratory Press 27492368 8711660 10.1101/gad.284430.116 Research Paper Transcription factors GAF and HSF act at distinct regulatory steps to modulate stress-induced gene activation Duarte et al. Genome-wide heat-shock regulation in flies http://orcid.org/0000-0002-1261-4929 Duarte Fabiana M. 1 Fuda Nicholas J. 13 Mahat Dig B. 1 Core Leighton J. 14 Guertin Michael J. 2 http://orcid.org/0000-0003-1201-9406 Lis John T. 1 1 Department of Molecular Biology and Genetics, Cornell University, Ithaca, New York 14835, USA; 2 Department of Biochemistry and Molecular Genetics, University of Virginia, Charlottesville, Virginia 22908, USA Present addresses: 3Department of Molecular and Cell Biology, University of California at Berkeley, Berkeley, CA 94720, USA; 4Department of Molecular and Cell Biology, University of Connecticut, Storrs, CT 06269, USA. Corresponding authors: johnlis@cornell.edu, mjg7y@virginia.edu 1 8 2016 1 8 2016 30 15 17311746 24 5 2016 11 7 2016 © 2016 Duarte et al.; Published by Cold Spring Harbor Laboratory Press 2016 This article, published in Genes & Development, is available under a Creative Commons License (Attribution 4.0 International), as described at http://creativecommons.org/licenses/by/4.0/. Here, Duarte et al. investigated the genome-wide heat shock (HS) response in Drosophila using precision run-on sequencing (PRO-seq) to assay the genome-wide distribution of transcriptionally engaged Pol II before and after HS in Drosophila S2 cells. This study comprehensively characterizes the transcriptional HS response and identifies the primary HS response genes and the rate-limiting steps in the transcription cycle that GAGA-associated factor (GAF) and HS factor (HSF) regulate. The coordinated regulation of gene expression at the transcriptional level is fundamental to development and homeostasis. Inducible systems are invaluable when studying transcription because the regulatory process can be triggered instantaneously, allowing the tracking of ordered mechanistic events. Here, we use precision run-on sequencing (PRO-seq) to examine the genome-wide heat shock (HS) response in Drosophila and the function of two key transcription factors on the immediate transcription activation or repression of all genes regulated by HS. We identify the primary HS response genes and the rate-limiting steps in the transcription cycle that GAGA-associated factor (GAF) and HS factor (HSF) regulate. We demonstrate that GAF acts upstream of promoter-proximally paused RNA polymerase II (Pol II) formation (likely at the step of chromatin opening) and that GAF-facilitated Pol II pausing is critical for HS activation. In contrast, HSF is dispensable for establishing or maintaining Pol II pausing but is critical for the release of paused Pol II into the gene body at a subset of highly activated genes. Additionally, HSF has no detectable role in the rapid HS repression of thousands of genes. heat shock transcription regulation transcription factors Pol II pausing National Institute of General Medical Sciences http://dx.doi.org/10.13039/100000057R01GM025232 ==== Body The heat-shock (HS) response in Drosophila melanogaster has been an effective model system to discover and study mechanisms of transcription and its regulation (Guertin et al. 2010). This highly conserved protective mechanism (Lindquist and Craig 1988) is regulated at the transcriptional level by the HS transcription factor (HSF) (Wu 1995). When activated by stress, HSF potently activates expression of HS genes, resulting in the accumulation of molecular chaperones, the HS proteins (HSPs), which helps the cell cope with stress-induced protein aggregation and misfolding (Lindquist and Craig 1988). The transcriptional HS response has been studied largely using Hsp70 as a model gene (Guertin et al. 2010). Hsp70 maintains a promoter-proximally paused RNA polymerase II (Pol II) molecule 20–40 base pairs (bp) downstream from the transcription start site (TSS) that is released to transcribe the gene at a low level during normal nonstress conditions (Rougvie and Lis 1988; Rasmussen and Lis 1993). The transcription factor GAGA-associated factor (GAF) is bound to the promoter of Hsp70 prior to HS, and GAF is important for the establishment and stability of paused Pol II (Lee et al. 1992, 2008; Kwak et al. 2013). GAF has a key role in keeping the promoter region open and free of nucleosomes (Tsukiyama et al. 1994; Fuda et al. 2015), which allows the recruitment of general transcription factors and the initiation of transcription by Pol II. Upon HS induction, HSF trimerizes and is rapidly recruited to the promoter, where it binds to its cognate HS DNA elements (HSEs) (Xiao and Lis 1988). After binding, HSF directly and indirectly recruits coactivators and other factors (Lis et al. 2000; Saunders et al. 2003; Ardehali et al. 2009) that affect the chromatin structure and composition and promote the release of Pol II from the paused complex into productive elongation. This transition from the paused state into productive elongation depends critically on the positive elongation factor P-TEFb and has been shown to be a very general step that is essential for the regulation of virtually all genes across different species (Rahl et al. 2010; Jonkers et al. 2014). The net result of this molecular cascade is an increase in transcription levels that can be ∼200-fold for some of the HS-regulated genes (Lis et al. 1981). Although the independent mechanisms of promoter-proximal pausing and escape to productive elongation have been well studied in the context of HS activation of Hsp70, we lack a comprehensive characterization of the genome-wide changes in transcription that result from HS in Drosophila. A thorough characterization of the affected genes is necessary to determine the generality and diversity of the roles of transcription factors such as GAF and HSF in the HS response and provide the statistical power to assess mechanisms of transcription regulation. Previous studies have mapped HSF-binding sites during normal growth conditions and after HS and observed that HSF recruitment to a promoter is not necessary or sufficient to direct HS gene activation (Trinklein et al. 2004; Guertin and Lis 2010; Gonsalves et al. 2011). Nonetheless, the rules governing the specificity of activation and repression across the Drosophila genome remain incomplete. Transcriptional changes after HS have also been measured in Drosophila and other organisms (Leemans et al. 2000; Guhathakurta et al. 2002; Murray et al. 2004; Trinklein et al. 2004; Sørensen et al. 2005; Gonsalves et al. 2011; Vihervaara et al. 2013); however, these studies were limited in resolution both temporally and spatially by measuring steady-state levels of mature mRNA. Furthermore, measurement of mRNAs cannot distinguish the effects on mRNA stability (Lindquist and Petersen 1990) and pre-mRNA processing (Yost and Lindquist 1986; Shalgi et al. 2014) from transcription or primary from secondary effects of the HS response. To overcome these limitations, we queried the genome-wide distribution of transcriptionally engaged RNA polymerases before and after HS induction using the precision nuclear run-on and sequencing (PRO-seq) assay and quantified differentially expressed genes. PRO-seq has high sensitivity and high spatial and temporal resolution, providing an unprecedented comprehensive view of the transcriptional profiles of cell populations. We show that the HS response is rapid and pervasive, with thousands of genes being repressed after 20 min of HS, and hundreds of genes being activated; moreover, the activated genes are not limited to the classical HSP genes. Promoter-proximal pausing is highly prevalent among the activated genes prior to HS, and here we demonstrate that its establishment on a subset of genes is dependent on GAF binding upstream and proximal to the TSS. Moreover, GAF depletion abrogates pausing and consequently impairs HS activation, indicating that this step in early transcription elongation is essential for gene activation. We also show that the recently identified transcription factor motif 1-binding protein (M1BP) (Li and Gilmour 2013) has a role in pausing and HS activation of a subset of genes that exhibit GAF-independent pausing. Furthermore, we demonstrate that only a relatively small fraction of HS-activated genes is regulated by HSF, and HS activation of these HSF-dependent genes is regulated at the level of pausing release. This study provides a genome-wide view of HS-induced transcriptional regulation and an understanding of how promoter context affects this process. Results Drosophila transcriptional HS response is rapid and pervasive We measured nascent transcription levels by PRO-seq in Drosophila S2 cells prior to HS (non-HS [NHS]) and 20 min after an instantaneous and continuous HS stress (Fig. 1A; Supplemental Table S1). PRO-seq maps the active sites of transcriptionally engaged RNA polymerase complexes by affinity purification and sequencing of nascent RNAs after a terminating biotin-NTP is incorporated during a nuclear run-on experiment (Kwak et al. 2013). The density of sequencing reads is proportional to the number of transcriptionally engaged polymerase molecules present at each position when the nuclei were isolated. PRO-seq has base-pair resolution, is strand-specific, and is not affected by the background levels of accumulated RNAs (Kwak et al. 2013). Biological replicates were highly correlated for both promoter and gene body PRO-seq reads (Spearman's coefficient ranged between 0.96 and 0.99) (Supplemental Fig. S1A,B, left panels). The expected genome-wide changes in transcription that occur during HS made it unfeasible to use total number of reads to normalize our data sets between conditions. Therefore, we normalized our libraries using a set of genes previously shown to have the same Pol II ChIP-seq (chromatin immunoprecipitation [ChIP] combined with high-throughput sequencing) signals in NHS and HS Drosophila S2 cells (Supplemental Fig. S2), where consistent backgrounds of ChIP-seq provide a basis of normalization (see the Materials and Methods; Supplemental Fig. S3 for the normalization method and our validation tests; Teves and Henikoff 2011). Figure 1. Drosophila transcriptional HS response is rapid and pervasive. (A) DESeq2 analysis of PRO-seq gene body reads between 20-min HS-treated and NHS cells displayed as an MA plot. Significantly changed genes were defined using a false discovery rate (FDR) of 0.001. Activated genes that passed our upstream transcription filter (see the Materials and Methods) are labeled in magenta, and repressed genes are labeled in blue. The number of genes in each class is shown in the plot. (fc) Fold change. (B) Representative view of a HS-repressed gene in the University of California at Santa Cruz (UCSC) genome browser (Kent et al. 2002). PRO-seq normalized reads for the plus strand are shown in red, and those for the minus strand are shown in blue. RNA sequencing (RNA-seq) reads for the plus strand are shown in green, and those for the minus strand are shown in orange. Gene annotations are shown at the bottom. (C) Gene ontology terms enriched in the HS-activated and HS-repressed classes. (D) Representative view of a HS-activated gene in the UCSC genome browser (Kent et al. 2002). Axes are the same as in B. We used DESeq2 (Love et al. 2014) to identify genes whose gene body reads significantly change after HS using a false discovery rate (FDR) of 0.001 (Supplemental Table S2). Due to the compactness of the Drosophila genome, some of the genes identified as differentially expressed by DESeq2 appear to be false positives caused by changes in run-through transcription originating at the upstream gene. To minimize the number of false positives, we implemented a filter to exclude from our analyses genes that have high levels of transcription in the region immediately upstream of the TSS (see the Materials and Methods; Supplemental Fig. S4 for a description of the implemented filter and validation tests). The genes that passed the filter were classified as activated or repressed (Supplemental Table S2). We observed a widespread shutdown of transcription, with 2300 genes being significantly repressed after HS (Fig. 1A [blue points], B [an example of a repressed gene, Vps39]). This finding is in agreement with low-resolution studies in Drosophila salivary gland polytene chromosomes that have shown that total Pol II levels and transcription decrease in response to HS (Spradling et al. 1975; Jamrich et al. 1977). A previous Pol II ChIP-seq study in Drosophila S2 cells has also observed a genome-wide decrease of Pol II levels in gene bodies (Teves and Henikoff 2011). Not surprisingly, measurements of steady-state mRNA levels before and after HS, including microarray studies and our own RNA sequencing (RNA-seq) data (Supplemental Fig. S5; Supplemental Table S3), were unable to detect a genome-wide shutdown of transcription despite having the sensitivity to detect a decrease in mRNA levels for some genes (Fig. 1B). Measurements of mRNA do not detect genome-wide transcriptional repression because the reduction of mRNA levels are obscured by steady-state levels of mRNAs already present in the cells; these mRNAs have much longer half-lives than the short HS time points examined here. Overall, our results greatly expand on these previous studies, identifying and quantifying the individual genes whose transcription is repressed after HS using a base-pair resolution method that specifically maps transcriptionally engaged RNA polymerase molecules. Gene ontology (GO) analysis reveals that the HS-repressed class is enriched for genes involved in basic cellular processes, such as cell cycle, RNA processing, protein transport and localization, and translation (Fig. 1C). This is consistent with previous findings in mammalian cells (Murray et al. 2004; Trinklein et al. 2004) and is expected as cells enter into a defensive nongrowth condition triggered by HS stress. Although not as abundant as the repressed class, hundreds of genes are activated by HS, many very highly (Fig. 1A [magenta points], D [an example of an activated gene, Hsromega]). Notably, we found that all seven classical HSP genes in our gene list show strong inductions after 20 min of HS and are among the top 10 genes with the highest HS induction (Supplemental Table S2), with fold changes ranging from 44-fold to 384-fold. Consistent with this result, GO analysis reveals that the HS-activated class is enriched for genes involved in the response to temperature and abiotic stimuli (Fig. 1C). Besides the classical HSP genes, our data reveal the activation of many genes that were not previously associated with the HS response and provide a comprehensive characterization and quantification of genes whose transcription is directly activated by HS. We measured nascent transcription levels as a function of time after HS induction to determine how fast activated and repressed genes respond to HS (Supplemental Fig. S6; Supplemental Table S1). Biological replicates produced high correlations for PRO-seq reads within either the promoter or gene body regions (Spearman's coefficient ranged between 0.9 and 0.99) (Supplemental Fig. S1C,D). The sequential HS time points displayed a progressive increase in the number of genes that were significantly activated (Supplemental Fig. S6A [magenta points], B [an example of an activated gene, CG13321]) and repressed (Supplemental Fig. S6A [blue points], C [an example of a repressed gene, CG14005]) by HS. No genes are significantly different after 30 sec, and only a small number of genes are significantly different after 2 min of HS. We observed a substantial genome-wide response to HS as early as 5 min after HS; the response is even more pervasive at later time points. Previous studies have shown that classical HSP genes are activated very rapidly (O'Brien and Lis 1993), but here we demonstrate that many other genes have a rapid response for both activation and repression. The number of significantly activated and repressed genes further increases after 10 and 20 min of HS (Supplemental Fig. S6A). Overall, our results demonstrate that the HS response produces an immediate and primary change in the transcription levels of ∼27% (∼24% repressed and ∼3% activated) of the unambiguously mappable mRNA-encoding genes (Core et al. 2012), with the repression of thousands of genes and the activation of hundreds of genes. Activated genes are highly paused prior to HS During normal cell growth, classical HSP genes have a paused, transcriptionally engaged polymerase between 20 and 50 bp downstream from the TSS (Rougvie and Lis 1988; Rasmussen and Lis 1993). Furthermore, promoter-proximal Pol II pausing is the major regulatory step for the HS activation of the Hsp70 gene, where it maintains the promoter region open and accessible to transcription factors (Lee et al. 1992; Shopland et al. 1995). We used our PRO-seq data to determine whether promoter-proximal pausing is a common feature of HS-activated genes. The average PRO-seq read intensity profile across HS-activated genes reveals a strong peak in the promoter-proximal region, which is substantially higher than repressed or unchanged gene classes (Fig. 2A). DNase I hypersensitivity data (Kharchenko et al. 2011) indicate that the promoter region of HS-activated genes is more accessible than the other two classes under basal uninduced conditions (Supplemental Fig. S7A). These data are consistent with the notion that promoter-proximal pausing is important to maintain an open chromatin environment around the TSS. Figure 2. GAGA factor is highly enriched in the promoter region of HS-activated genes prior to HS. (A) PRO-seq read density between −100 and +200 bp to the TSS (in 5-bp bins) for the LacZ-RNAi NHS data set of HS-activated (n = 249), HS-repressed (n = 2300), and unchanged (n = 517) genes. The shaded area represents the 75% confidence interval. (B) Heat map showing the GAF ChIP-seq signal in 20-bp windows from ±1 kb to the TSS of HS-activated (n = 249) and HS-repressed (n = 2300) genes. For each class, genes were ordered by the distance between the highest-intensity window and the TSS. (C) GAF ChIP-seq read density between −750 and +750 bp to the TSS (in 20-bp bins) of HS-activated (n = 249), HS-repressed (n = 2300), and unchanged (n = 517) genes. The shaded area represents the 75% confidence interval. (D) Cumulative distribution plots of the distance between the closest GAF ChIP-seq peak and the TSS of each gene in the HS-activated, HS-repressed, and unchanged classes. We calculated the pausing index (PI)—which is the ratio of read density in the promoter-proximal region relative to the gene body—for each individual gene (Supplemental Table S4; Core et al. 2008). The vast majority of HS-activated genes (∼90%) was classified as paused (Fisher's exact, P-value ≤ 0.01) (Supplemental Fig. S7B; Core et al. 2008). The PI is significantly higher for activated genes compared with the repressed (Mann-Whitney U-test, P-value < 2.2 × 10−16) and unchanged (Mann-Whitney U-test, P-value < 2.2 × 10−16) classes (Supplemental Fig. S7C). Although the pausing levels of HS-repressed genes are not as high as the activated class (Fig. 2A), a considerable percentage of repressed genes were also classified as paused (∼80%) (Supplemental Fig. S7B). Overall, our results indicate that high levels of promoter-proximal pausing are a general feature of HS-induced genes prior to HS and may play an important role in poising these genes for HS activation by transcription factors. GAGA factor is highly enriched in the promoter region of HS-activated genes To identify candidate factors that play a role in allowing genes to be HS-activated, we screened modENCODE and other publically available genomic transcription factor-binding data (Celniker et al. 2009; Li and Gilmour 2013; Fuda et al. 2015) for factors that are differentially enriched in HS-activated relative to HS-repressed or unchanged genes prior to HS (Supplemental Table S5). The most significant differential enrichment was observed for GAF (Fig. 2B; Supplemental Table S5; GAF ChIP-seq data from Fuda et al. 2015). As seen in Figure 2B, when compared with the repressed class, HS-activated genes show enriched GAF binding immediately upstream of the TSS, which is also evidenced by a peak in the average ChIP-seq intensity profile (Fig. 2C). Furthermore, de novo motif analysis identified the DNA sequence bound by GAF, the GAGA element (Omichinski et al. 1997; Wilkins 1998), as the most significantly overrepresented motif in the promoter region of HS-activated genes (Supplemental Fig. S8). We then identified the closest GAF ChIP-seq peak to the TSS of each gene and plotted the cumulative distribution of these distances for our three gene classes (activated, repressed, and unchanged) (Fig. 2D). GAF binds significantly closer to the activated genes than the repressed (Kolmogorov-Smirnov test, P-value < 2.2 × 10−16) (Fig. 2D) and unchanged (Kolmogorov-Smirnov test, P-value < 2.2 × 10−16) (Fig. 2D) classes, and >70% of activated genes are bound by GAF within ±1 kb of the TSS. These results suggest that GAF binding close to the TSS prior to HS is important for the activation of HS-induced genes. GAF is critical for HS activation when bound immediately upstream of the core promoter To investigate whether GAF binding is essential for HS activation, we performed PRO-seq in biological replicates of GAF-RNAi-treated cells prior to HS and after 20 min of HS (Spearman's coefficient ranged between 0.96 and 0.99) (Fig. 3A,B; Supplemental Table S1; Supplemental Fig. S1A,B, right panels). The decrease in GAF protein levels after the RNAi treatment produced similar numbers of genes that were significantly activated or repressed by HS (cf. Supplemental Fig. S9A and Fig. 1A); however, a comparison of the HS gene body reads in the GAF-RNAi and LacZ-RNAi control identified many genes that were significantly affected by the knockdown. The HS PRO-seq levels of 20% of activated genes were affected by GAF-RNAi and nearly all were reduced (Fig. 3C, left panel), while <1% of the repressed class were affected (Fig. 3C, right panel), demonstrating that GAF is important for HS activation but not repression. Greater than 90% of the genes that have reduced HS induction after GAF knockdown have GAF binding within ±1 kb of the TSS (Fig. 3C, left panel). Taken together, these results indicate that promoter-bound GAF is indispensable for the proper activation of many HS-activated genes. Figure 3. GAF's role in HS activation correlates with its function in establishing promoter-proximal pausing prior to HS. (A) Experimental setup. Drosophila S2 cells were treated with either GAF-RNAi or LacZ-RNAi for 5 d. Nuclei were then isolated for PRO-seq after cells were incubated at room temperature (NHS) or heat-shocked for 20 min (HS). (B) Western blot of whole-cell extracts from LacZ-RNAi-treated and GAF-RNAi-treated cells using antibodies detecting GAF and TFIIS (loading control). One-hundred percent is equivalent to 1.5 × 106 cells. (C) DESeq2 analysis to determine the effect of GAF-RNAi treatment on the PRO-seq gene body reads after HS for the HS-activated (n = 249) and HS-repressed (n = 2300) classes. DESeq2 was used to identify significantly changed genes between GAF-RNAi HS and LacZ-RNAi HS cells, and the results are displayed as MA plots. Significantly changed genes were defined using an FDR of 0.001. GAF-bound genes are labeled in purple, significantly changed genes (according to DESeq2) are labeled in green, and genes that are both GAF-bound and significantly changed are labeled in orange. (fc) Fold change. (D) PRO-seq read density between −100 and +200 bp to the TSS (in 5-bp bins) for the LacZ-RNAi NHS and GAF-RNAi NHS treatments of genes with GAF-dependent (n = 44) or GAF-independent (n = 199) HS activation (HS ↑). (E) Box plot showing the GAF-/LacZ-RNAi pausing region fold change prior to HS (NHS) for genes with GAF-dependent or GAF-independent HS activation. Mann-Whitney U-test, P-value < 2.2 × 10−16. Over 70% of the genes with GAF-dependent HS activation have significantly reduced pausing upon GAF depletion prior to HS, while only 15% of the GAF-independent genes were significantly affected. (F) Representative view in the UCSC genome browser (Kent et al. 2002) of a gene with GAF-dependent pausing prior to HS whose activation is inhibited by GAF-RNAi treatment. PRO-seq normalized reads for the different RNAi treatments (LacZ and GAF) before and after HS for the plus strand are shown in red, and those for the minus strand are shown in blue. Gene annotations are shown at the bottom. GAF is critical for the HS activation of many genes; however, the induction of >70% of the genes that are bound by GAF prior to HS is not affected by GAF knockdown. These two classes of GAF-bound genes, which respond differentially to GAF knockdown, cannot simply be explained by differences in the response to the RNAi treatment, since the GAF ChIP-seq signal for both classes is similarly reduced by the knockdown (Supplemental Fig. S9B). However, GAF-bound genes with GAF-dependent HS activation have significantly higher GAF ChIP-seq intensities when compared with the GAF-bound genes with GAF-independent HS activation (Mann-Whitney U-test, P-value = 8.96 × 10−10) (Supplemental Fig. S9C). The class of GAF-bound, HS-activated genes whose induction is dependent on GAF has a strong preference for GAF binding immediately upstream of the TSS, between −100 and −50 bp (Supplemental Fig. S9D, right panel). Taken together, these results suggest that higher binding levels and positioning upstream and proximal to the TSS are essential for GAF's role in HS activation. GAF's role in HS activation correlates with its function in establishing promoter-proximal pausing prior to HS GAF has been shown to have a role in the establishment of promoter-proximal pausing and consequent HS activation of two classical HSP genes (Glaser et al. 1990; Lee et al. 1992; Lu et al. 1993; O'Brien et al. 1995), and a recent study has demonstrated that pausing was significantly reduced on a large subset of GAF-bound genes upon GAF depletion (Fuda et al. 2015). However, the role of GAF-mediated pausing in gene activation has not yet been studied in a comprehensive genome-wide manner. We hypothesized that GAF's role in HS activation is connected to its ability to create promoter-proximal pausing prior to HS. To test this hypothesis, we compared the NHS promoter-proximal PRO-seq reads for the LacZ-RNAi control and GAF-RNAi treatment between the subset of GAF-bound genes whose HS induction is dependent on GAF (GAF-dependent HS activation) and the HS-activated genes whose induction is unaffected by GAF depletion (GAF-independent HS activation). As observed in Figure 3D, there is a substantial reduction in the NHS pausing levels after GAF knockdown for genes with GAF-dependent HS activation, while the NHS pausing levels of the GAF-independent class are largely unaffected. To quantify this effect, we compared the LacZ-RNAi and GAF-RNAi NHS reads in the pausing region for genes with GAF-dependent or GAF-independent HS activation. As observed in Figure 3E, most genes with GAF-dependent HS activation have a reduced number of reads (fold change <1) in the pausing region upon GAF knockdown prior to HS. In contrast, the distribution of fold changes for the GAF-independent class is centered around one, indicating that GAF binding prior to HS is not essential to establish pausing at these genes (Mann-Whitney U-test, P-value < 2.2 × 10−16). Figure 3F shows an example of a HS-activated gene that displays GAF-dependent pausing prior to HS whose induction is inhibited by GAF knockdown. Taken together, these results indicate that GAF's role in HS activation strongly correlates with its function in establishing promoter-proximal pausing prior to HS. Figure 3D also shows that GAF-dependent genes have higher levels of promoter-proximal pausing prior to HS than the GAF-independent ones. Interestingly, they also have a higher average HS/NHS induction (Mann-Whitney U-test, P-value = 0.0367) (Supplemental Fig. S9E); however, the distribution of HS/NHS fold changes for these two classes mostly overlap. Additionally, there was no preferential enrichment for classical HSP genes in either class. Insulator proteins and M1BP are enriched in the promoter region of HS-activated genes with GAF-independent induction While nearly all activated genes display promoter-proximal pausing prior to HS, we showed that GAF is essential for pausing establishment and HS activation on a subset of these genes. To identify factors that can contribute to the establishment of pausing on GAF-independent genes, we screened modENCODE and other publically available chromatin factor ChIP-seq or ChIP–chip data sets for factors that are differentially enriched in the promoter regions of GAF-independent relative to GAF-dependent genes prior to HS (Supplemental Table S6). Among the factors with the most significant differential enrichment were the transcription factor M1BP (ChIP-seq data from Li and Gilmour 2013), the insulator protein BEAF-32 (ChIP–chip data from Schwartz et al. 2012), and the chromodomain-containing protein Chromator (ChIP–chip data from Kharchenko et al. 2011) (Fig. 4A–C). M1BP is a recently discovered zinc finger transcription factor that has been shown to orchestrate promoter-proximal pausing in a GAF-independent manner (Li and Gilmour 2013). BEAF-32 is one of the insulator-associated proteins identified in Drosophila (Zhao et al. 1995), Chromator was initially identified as a mitotic spindle protein and later implicated in the regulation of chromosome structure through partial cooperation with BEAF-32 (Rath et al. 2006; Gan et al. 2011), and both of these proteins are enriched at the boundaries of physical chromosomal domains (Hou et al. 2012; Sexton et al. 2012). The factor with the highest differential enrichment for promoter-bound genes in our screen was the tandem kinase JIL-1 (Supplemental Table S6), which has been shown previously to interact with Chromator (Rath et al. 2006). However, a comparison between the JIL-1 ChIP–chip intensities of genes with GAF-dependent and GAF-independent HS activation did not show the same striking differences that were observed for M1BP, BEAF-32, and Chromator (Supplemental Fig. S10). Remarkably, almost no overlap exists between genes with GAF-dependent HS activation and genes bound by M1BP or insulator proteins within ±1 kb of the TSS (Fig. 4D). The mutually exclusive distributions of GAF and M1BP in promoter-proximal pausing has been reported previously (Li and Gilmour 2013), and our results suggest a possible role for M1BP in pausing and HS activation. Similarly to M1BP, the mutually exclusive distribution of insulator proteins and the GAF-dependent subset suggests that BEAF-32 and/or Chromator may have a role in generating promoter-proximal pausing when bound proximally to the TSS. GAF has also been classified as an insulator protein with enhancer-blocking activity (Ohtsuki and Levine 1998; Schweinsberg et al. 2004), which suggests a possible overlap between insulator function and a role in maintaining an open chromatin environment that enables promoter-proximal pausing and opens the possibility for a novel role of BEAF-32 and Chromator as pausing factors. Another possible explanation is that these insulator proteins reside between GAF and the TSS, therefore blocking any activity of GAF on the promoter, which could explain why pausing is not affected by GAF depletion at insulator-bound promoters. Figure 4. Insulator proteins and M1BP are enriched in the promoter region of HS-activated genes with GAF-independent induction. (A–C) M1BP ChIP-seq (A), BEAF-32 ChIP–chip (BEAF-HB antibody) (B), and Chromator ChIP–chip (BR antibody) (C) signal between −750 and +750 bp to the TSS (in 20-bp bins) of genes with GAF-dependent (n = 44) or GAF-independent (n = 199) HS activation (HS ↑). (D) Venn diagram showing the overlap between HS-activated genes with GAF-dependent activation and genes bound by M1BP or both insulator proteins (BEAF-32 and Chromator; only genes with insulator binding detected by both antibodies for these two proteins were considered) within ±1 kb of the TSS. M1BP is important for promoter-proximal pausing and HS activation of a subset of M1BP-bound HS-activated genes To investigate whether M1BP has a role in pausing and HS activation of M1BP-bound genes with GAF-independent HS activation, we performed PRO-seq in biological replicates of M1BP-RNAi-treated cells prior to HS and after 20 min of HS (Spearman's coefficient ranged between 0.97 and 0.99) (Fig. 5A,B; Supplemental Table S1; Supplemental Fig. S11A,B, right panels). As seen in Figure 5C, M1BP depletion by RNAi has a small effect on the HS activation of M1BP-bound, HS-activated genes when compared with the LacZ-RNAi control; however, this effect was not statistically significant. To assess whether, like GAF, M1BP's role in HS activation is associated with its role in establishing promoter-proximal pausing, we then focused on the subset of HS-activated, M1BP-bound genes that display M1BP-dependent pausing prior to HS. As seen in Figure 5D, M1BP knockdown has a significant effect on the HS activation of this subset of genes (Mann-Whitney U-test, P-value = 0.05), indicating that M1BP has a role in pausing establishment and HS activation of at least a subset of M1BP-bound genes. Figure 5E shows an example of a gene that displays M1BP-dependent pausing prior to HS whose HS induction is affected by M1BP knockdown. Thus, M1BP, like GAF, is important for pausing and HS activation of a subset of genes, supporting the hypothesis that pausing is a prerequisite for HS activation. Figure 5. M1BP is important for pausing and HS activation of a subset of M1BP-bound genes with GAF-independent induction. (A) Experimental setup. Drosophila S2 cells were treated with either M1BP-RNAi or LacZ-RNAi for 5 d. Nuclei were then isolated for PRO-seq after cells were incubated at room temperature (NHS) or heat-shocked for 20 min (HS). (B) Western blot of whole-cell extracts from LacZ-RNAi-treated and M1BP-RNAi-treated cells using antibodies detecting M1BP and TFIIS (loading control). One-hundred percent is equivalent to 1.5 × 106 cells. (C) Box plot showing the HS/NHS fold change of M1BP-RNAi or LacZ-RNAi control cells for all M1BP-bound, HS-activated genes (HS ↑). (D) Box plot showing the HS/NHS fold change of M1BP-RNAi or LacZ-RNAi control cells for M1BP-bound, HS-activated genes with M1BP-dependent pausing. Mann-Whitney U-test, P-value = 0.05. (E) Representative view in the UCSC genome browser (Kent et al. 2002) of a gene with M1BP-dependent pausing prior to HS whose activation is decreased by M1BP-RNAi treatment. PRO-seq normalized reads for the different RNAi treatments (LacZ and M1BP) before and after HS for the plus strand are shown in red, and those for the minus strand are shown in blue. Gene annotations are shown at the bottom. HSF is essential for the induction of only a small minority of HS-activated genes HSF is the evolutionarily conserved master regulator of the HS response and is essential for the activation of classical HSP genes (Wu 1995). Inducible HSF binding at those genes is critical for the recruitment of the positive elongation factor P-TEFb (Lis et al. 2000), which modulates the release of Pol II into productive elongation. We used our previously published HSF ChIP-seq data sets performed before and after 20 min of HS induction (Guertin and Lis 2010) to determine whether HSF also preferentially binds to noncanonical HS-activated genes. HSF ChIP-seq peaks are closer to the TSS in the HS-activated class when compared with the HS-repressed (Kolmogorov-Smirnov test, P-value = 4.04 × 10−12) (Fig. 6A) and unchanged (Kolmogorov-Smirnov test, P-value = 1.6 × 10−7) gene classes (Fig. 6A). Surprisingly, even though HSF is enriched in the proximity of activated genes, <20% of those genes have an HSF ChIP-seq peak within ±1 kb of the TSS. The existence of HSF-independent genes has been demonstrated previously in Drosophila (Gonsalves et al. 2011). However, our study substantially expands the number of identified genes and offers a more comprehensive view of HSF-independent regulation due to the considerably higher resolution and sensitivity afforded by our binding and nascent transcription assays. These results indicate that HSF can activate genes when bound to distal enhancer sites or that there are other factors dictating the induction of HS-activated genes. Figure 6. HSF is essential for the induction of only a small minority of HS-activated genes and activates genes by stimulating the release of paused Pol II. (A) Cumulative distribution plots of the distance between the closest HSF ChIP-seq peak and the TSS of each gene in the HS-activated (n = 249), HS-repressed (n = 2300), and unchanged (n = 517) classes. (B) Experimental setup. Drosophila S2 cells were treated with either HSF-RNAi or LacZ-RNAi for 5 d. Nuclei were then isolated for PRO-seq after cells were incubated at room temperature (NHS) or heat-shocked for 20 min (HS). (C) Western blot of whole-cell extracts from LacZ-RNAi-treated and HSF-RNAi-treated cells using antibodies detecting HSF and TFIIS (loading control). One-hundred percent is equivalent to 1.5 × 106 cells. (D) DESeq2 analysis to determine the effect of HSF-RNAi treatment on the PRO-seq gene body reads after HS for the HS-activated and HS-repressed classes. We used DESeq2 to identify significantly changed genes between HSF-RNAi HS and LacZ-RNAi HS cells, and the results are displayed as MA plots. Significantly changed genes were defined using an FDR of 0.001. HSF-bound genes are labeled in purple, significantly changed genes (according to DESeq2) are labeled in green, and genes that are both HSF-bound and significantly changed are labeled in orange. (fc) Fold change. (E) PRO-seq read density between −200 and +1000 bp to the TSS (in 5-bp bins) of genes with HSF-dependent (n = 44) or HSF-independent (n = 197) HS activation (HS ↑) for the LacZ-RNAi and HSF-RNAi data sets before (NHS) and after HS. HSF activates genes by stimulating the release of paused Pol II To investigate HSF's roles during the HS-induced transcriptional response, we performed PRO-seq in biological replicates of HSF-RNAi-treated cells prior to HS and after 20 min of HS (Spearman's coefficient ranged between 0.96 and 0.99) (Fig. 6B,C; Supplemental Table S1; Supplemental Fig. S1A,B, middle panels). Comparison of the HS gene body reads in the LacZ-RNAi control and HSF-RNAi for activated and repressed genes shows that the HS PRO-seq levels of 20% of activated genes were affected by HSF-RNAi, while a significant change was observed for only <1% of the repressed class, demonstrating that HSF is important for HS activation but not repression (Fig. 6D). Most of the activated genes that have HSF binding within ±1 kb of the TSS have reduced HS induction after HSF knockdown (Fig. 6D, left panel, orange points). HSF-bound genes with compromised induction upon HSF knockdown are enriched for HSF binding immediately upstream (within 200 bp) of the TSS, while the unaffected class displays a random distribution of distances (Supplemental Fig. S12A). Furthermore, HSF-bound genes with reduced HS induction have significantly higher HSF ChIP-seq binding intensity when compared with the unaffected class (Mann-Whitney U-test, P-value = 2.5 × 10−3) (Supplemental Fig. S12B), indicating that higher HSF binding levels and positioning upstream and proximal to the TSS are important for the induction of HSF's target genes. Additionally, a comparison of all induced, HSF-dependent genes with the remainder of HS-activated genes (HSF-independent HS activation) showed that genes depending on HSF have stronger HS induction (Fig. 6E). As expected, HSF is essential for the HS activation of all seven classical HSP genes in our gene list, which strongly contributes to the higher HS induction of HSF-dependent genes relative to genes with HSF-independent HS activation (Fig. 6E). HSF depletion does not affect the induction of most genes that are not bound by HSF within ±1 kb of the TSS (Fig. 6D, left panel, gray points). Nonetheless, the presence of significantly changed genes with no proximal HSF binding (Fig. 6D, left panel, green points) indicates that HSF may be able to mediate activation at distal enhancer sites on a small subset of genes. The enhancer activity of HSF has been shown previously in a focused study of Hsp70 to be weak and require large arrays of HSF-binding sites (Bienz and Pelham 1986). This early study and the rarity with which we found HSF acting at a distance might be explained if such long-range interactions required specialized binding sites and chromatin architecture. Clearly, the preferred mode of HSF action is close to promoters. Composite profiles show that the average pausing levels of genes with HSF-dependent and HSF-independent HS activation are not affected by HS in both LacZ-RNAi and HSF-RNAi conditions (Fig. 6E), indicating that HSF depletion and HS do not have much of an effect on pausing. Quantification of the effect of HSF knockdown on pausing levels in both NHS and HS conditions for all HS-activated genes revealed that the pausing level of only one gene was significantly affected by the knockdown (Supplemental Fig. S12C). Taken together, these results suggest that HSF acts mainly at the release of paused Pol II into productive elongation, which is consistent with the critical role of HSF in the recruitment of the pause release factor P-TEFb to Hsp70 (Lis et al. 2000). HS transcriptional repression results in a decrease of promoter-proximally paused Pol II Our data revealed that HS induction causes a vast transcriptional shutdown, with thousands of genes being repressed after 20 min of increased temperatures (Fig. 1A). To elucidate the mechanisms involved in this repression, we observed the PRO-seq profile for all repressed genes plotted as heat maps before and after HS (Fig. 7A). This analysis indicates the presence of enriched PRO-seq reads in the region immediately downstream from the TSS, representing promoter-proximally paused polymerases. Both gene body reads and reads in this promoter-proximal region are reduced after HS, which is also evidenced by the overall blue color of the fold change heat map (Fig. 7A, right panel). The overall NHS and HS distributions and the HS/NHS fold change are very similar after HSF depletion by knockdown (Fig. 7B), indicating that HSF does not play a role in gene repression by HS. Figure 7. HS transcriptional repression is HSF-independent and results in a decrease of promoter-proximally paused Pol II. Heat maps showing the NHS PRO-seq levels (left panel), the HS PRO-seq levels (middle panel), and the fold change between the two conditions (right panel) between −50 and +1000 bp to the TSS (in 5-bp bins) of HS-repressed genes (n = 2300) for the LacZ-RNAi (A) and HSF-RNAi (B) treatments. Genes in both A and B are sorted by the HS/NHS PRO-seq fold change in the LacZ-RNAi condition (highest to lowest). Discussion In this study, we used PRO-seq to comprehensively characterize the direct changes in Pol II distribution that occur in Drosophila S2 cells in the minutes following HS. We show that the HS response is more general than previously appreciated, with thousands of genes being repressed and hundreds activated by heat. This latter class is not limited to the group of cellular chaperones that are known to be activated by stress (Lindquist and Craig 1988) and includes hundreds of other genes with various cellular functions (Supplemental Table S2). Surprisingly, only 20% of the activated genes are regulated by HSF, which was previously believed to be the major orchestrator of the response. Moreover, we show that promoter-proximal pausing is highly pronounced and prevalent among activated genes prior to HS. GAF, which has been shown to be important for establishing pausing, is highly enriched at the promoters of HS-activated genes, and our results suggest that GAF-mediated pausing in a subset of these genes is essential for HS activation. Furthermore, our results indicate that HS activation of HSF-dependent genes is regulated at the level of pausing release, whereas HS repression of thousands of genes is regulated at the step of transcription initiation in Drosophila, and this process is independent of HSF. Very recently, we showed in mice that HS activation is similarly regulated by HSF at the level of pause release; however, in contrast to Drosophila, HS repression of genes is also mediated at pause release (Mahat et al. 2016). In both mammals and Drosophila, the widespread transcriptional repression is independent of HSF. Overall, by measuring how transcription changes after HS, our results provide insights into mechanisms of transcription activation and repression, the key regulating factors, and the steps in the transcription cycle that are modulated. GAF-mediated promoter-proximal pausing is essential for the HS activation of a subset of genes Classical HSP genes accumulate paused Pol II molecules between 20 and 50 bp downstream from the TSS prior to HS (Rougvie and Lis 1988; Rasmussen and Lis 1993). Our results and analyses greatly expand on these previous findings and indicate that pausing is a common feature among HS-activated genes and is not specific to the highly induced class of molecular chaperones. Previous studies have shown that the paused Pol II complex on Hsp70 and many other genes is remarkably stable (Henriques et al. 2013; Buckley et al. 2014; Jonkers et al. 2014), and this stably paused molecule can help to maintain an open chromatin environment that is accessible to transcription factors that will promote the release of Pol II into productive elongation, mediating a rapid response to HS. The open chromatin state of our newly identified HS-activated genes is confirmed with the higher DNase I hypersensitivity signal observed in the promoter region relative to repressed and unchanged genes (Supplemental Fig. S7A). GAF has been shown previously to be important for establishing pausing and is highly enriched in the promoter region of activated genes prior to HS. GAF is essential for HS activation of a subset of GAF-bound genes that have high levels of GAF binding in the region immediately upstream of the core promoter, indicating that GAF's positioning and levels are important for its role in the HS response. We also observed that the pausing levels of genes with GAF-dependent HS activation are dramatically reduced upon GAF depletion prior to HS. Transgenic studies of the model Hsp70 gene have demonstrated that the presence of the GAF-binding element is essential for generating pausing at this gene and that pausing level changes created by mutating the core promoter strongly correlate with the promoter's potential to induce transcription upon HS induction (Lee et al. 1992). Our results expand on these studies and demonstrate that GAF depletion prior to HS in the native chromatin environment of a subset of HS-activated genes abrogates Pol II pausing levels and the consequent induction of these genes by HS. Importantly, other GAF-bound genes that maintain pausing upon GAF knockdown due to the activity of other pausing factors like M1BP and possibly the insulator proteins BEAF-32 and Chromator remain fully HS-inducible. We propose a model in which GAF-mediated pausing is essential to maintain an open chromatin environment at the promoter region prior to HS (Supplemental Fig. S13A). When GAF is depleted by knockdown and pausing is not properly established, the promoter loses its potential to induce transcription after HS (Supplemental Fig. S13A). HSF acts at the step of promoter-proximal pausing release HSF depletion has almost no effect on pausing reads both before and after HS (Supplemental Fig. S12C), and the average pausing levels are largely unaffected by HS and HSF knockdown (Fig. 6E). The amount of pausing is determined by the transcription recruitment/initiation rate and the rate of escape into productive elongation. If HSF was acting at the step of Pol II initiation, we would expect the pausing levels to be reduced upon HSF depletion, which is not observed. Therefore, we propose a model in which, after being recruited to the promoter region upon HS, HSF promotes the release of Pol II into the gene body (Supplemental Fig. S13B), likely through the indirect recruitment of P-TEFb, which has been shown to be the case for classical HSP genes (Lis et al. 2000). Pausing also maintains an open chromatin environment that is accessible to transcription activators such as HSF. In this model, the activity of factors that are important for establishing pausing prior to HS, such as GAF, is crucial for HSF-dependent HS activation (Supplemental Fig. S13B), and failure to generate pausing prevents the induction of HSF target genes. Less than 20% of the genes with HSF-dependent HS activation are also dependent on GAF for activation, indicating that the action of other factors such as M1BP and possibly BEAF-32 and Chromator is important for pausing and consequent HS activation of these HSF-dependent genes. HS causes a rapid and broad reduction in transcription, which is regulated at the transcription initiation step and is independent of HSF Early low-resolution studies in Drosophila polytene chromosomes have shown that HS causes a genome-wide down-regulation of transcription (Spradling et al. 1975; Jamrich et al. 1977), presumably to reduce the accumulation of misfolded protein aggregates. Although this has been a paradigm in the HS field, higher-resolution genome-wide studies have failed to identify all of the primary genes that are repressed by heat, mostly due to the limitations of measuring steady-state levels of mRNA, which requires that the mRNAs already present in the cells have shorter half-lives than the HS time points used in the experiment. Our results provide definitive evidence to support the widespread shutdown of transcription caused by HS. We identified and quantified the genes with significantly reduced transcription and demonstrated that the HS-repressive response is very rapid, with more than a thousand genes being repressed after only 5 min of HS (Supplemental Fig. S6A). Furthermore, the Pol II density in the promoter-proximal region, which represents the paused Pol II molecules, is also significantly reduced across all HS-repressed genes (Fig. 7). The accumulation of Pol II in the pausing region depends on both the transcription initiation rate and the rate of escape into productive elongation. The reduction in pausing levels thus indicates that the recruitment and initiation of Pol II are affected by HS (Supplemental Fig. S13C). The HS-induced binding of HSF is not essential for the genome-wide transcriptional repression (Fig. 7), and, given the magnitude of this repressive response, we believe that it is unlikely that one single transcription repressor is responsible for inhibiting transcription initiation in all HS-repressed genes. We considered three possible mechanisms, which are not mutually exclusive, that could be responsible for HS-mediated repression. (1) The activity of a general transcription factor that is involved in recruitment of Pol II to the promoter could be modulated by heat stimulus. (2) Changes in nucleosomal composition or positioning induced by heat could generate an unfavorable chromatin environment that would prevent transcription initiation and elongation. A previous study has demonstrated that HS results in decreased nucleosome turnover genome-wide within gene bodies; however, the same pattern was observed after drug inhibition of Pol II elongation, arguing that reduced nucleosome turnover may be a consequence rather than the cause of the genome-wide transcriptional repression (Teves and Henikoff 2011). (3) A genome-wide rearrangement of the three-dimensional (3D) chromatin structure could either disrupt long-range interactions that are needed for transcription or allow new long-range interactions that repress transcription initiation, which is supported by a recent study in a different Drosophila cell line that demonstrated that HS induces a genome-wide rearrangement in the 3D nuclear architecture (Li et al. 2015). Any model must accommodate our new observations that (1) recruitment of Pol II is the step in the transcription cycle that is regulated, (2) HSF is not involved in the repression, and (3) the specifically repressed genes identified here and their level of down-regulation must be accommodated by any proposed regulatory factor interactions. Materials and methods GAF-RNAi, HSF-RNAi, M1BP-RNAi, and LacZ-RNAi treatments Drosophila S2 cells were grown in M3 + BPYE medium supplemented with 10% FBS until they reached 3 × 106 to 5 × 106 cells per milliliter. At this point, the cells were split into 1 × 106 cells per milliliter in serum-free M3 + BPYE medium, and the desired volume of cells was mixed with LacZ, GAF, HSF, or M1BP dsRNA to a final concentration of 10 µg/mL. After incubation for 45 min at 25°C, an equal volume of M3 + BPYE medium supplemented with 20% FBS was added to the cells. After 2.5 d, the cells were split 1:2 into two new flasks, and more dsRNA was added to keep the final concentration at 10 µg/mL. After 2.5 d, the cells were HS-treated and harvested for nuclei isolation. The M1BP-RNAi treatment was performed separately with its own LacZ control. The dsRNAs used in these experiments were transcribed from a dsDNA template that had a T7 polymerase promoter at both ends. The DNA templates were generated by PCR using the following primers: LacZ forward (GAATTAATACGACTCACTATAGGGAGAGATATCCTGCTGATGAAGC), LacZ reverse (GAATTAATACGACTCACTATAGGGAGAGCAGGAGCTCGTTATCGC), GAF forward (GAATTAATACGACTCACTATAGGGATGGTTATGTTGGCTGGCGTCAA), GAF reverse (GAATTAATACGACTCACTATAGGGATCTTTACGCGTGGTTTGCGT), HSF forward (GAATTAATACGACTCACTATAGGGAGAGCCTTCCAGGAGAATGCA), HSF reverse (GAATTAATACGACTCACTATAGGGAGAGCTCGTGGATAACCGGTC), M1BP forward (from Li and Gilmour 2013) (GAATTAATACGACTCACTATAGGGAGAGCAGCCAAATTGCTTGTCC), and M1BP reverse (from Li and Gilmour 2013) (GAATTAATACGACTCACTATAGGGAGAAGACGGTGAAGACGCCC). Western blot analysis to assess knockdown levels Western blots were performed using standard conditions, and dilutions of the LacZ-RNAi control samples were used as a quantitative indication of signal linearity. Laboratory stocks of rabbit anti-HSF and anti-GAF antibodies and guinea pig anti-TFIIS antibody were used at dilutions of 1:2000, 1:500, and 1:3000, respectively. The rabbit anti-M1BP antibody was provided by David Gilmour's laboratory and was used at a 1:5000 dilution. We used 1 mg/mL IRDye 800CW donkey anti-rabbit and 1 mg/mL IRDye 680LT donkey anti-guinea pig as secondary antibodies at a 1:15,000 dilution, and the membrane was imaged using the LI-COR Odyssey imaging system. HS treatments For the HS treatments, an equal volume of M3 + BPYE medium (no serum) at 48°C was added to the cells, and the cultures were incubated for the desired time at 37°C. Preparation of PRO-seq libraries Nuclei isolation and PRO-seq library preparation were performed as described previously (Kwak et al. 2013). Preparation of RNA-seq libraries Total RNA from S2 cells was extracted using TRIzol reagent (Thermo Fisher Scientific) and then isolated from the aqueous phase using the EZNA total RNA kit I (Omega Bio-tek). The following steps were performed by the Cornell RNA Sequencing Core (Department of Biomedical Sciences, College of Veterinary Medicine, Cornell University). Poly(A)+ RNA was isolated with the NEBNext Poly(A) mRNA magnetic isolation module (New England Biolabs). TruSeq-barcoded RNA-seq libraries were generated with the NEBNext Ultra Directional RNA library preparation kit (New England Biolabs). Each library was quantified with a Qubit 2.0 (dsDNA HS kit, Thermo Fisher Scientific), and the size distribution was determined with a fragment analyzer (Advanced Analytical) prior to pooling. PRO-seq data acquisition PRO-seq libraries were sequenced in 50-nucleotide (nt) runs on the Illumina HiSeq using standard protocols at the Cornell Biotechnology Resource Center (http://www.BRC.cornell.edu). Raw sequencing reads were processed using the FastX toolkit (http://hannonlab.cshl.edu/fastx_toolkit/index.html). Illumina adapters were removed with the fastx_clipper tool, and reads were trimmed to 26-mers using fastx_trimmer. Sequencing reads <15 nt were discarded. fastx_reverse_complement was then used to generate the reverse complement of the sequencing reads, which correspond to the sense strand of nascent RNA in the nucleus. Reads were aligned uniquely to the D. melanogaster dm3 reference genome using Bowtie (Langmead et al. 2009) with up to two mismatches. Histograms of the 3′ end position of each mapped read in base-pair resolution were generated in bedgraph format and used for all subsequent analyses. Supplemental Table S1 contains a summary of sequencing yields and the number of reads that mapped uniquely to the genome or other annotations. Replicates were highly correlated and pooled for further analyses (Supplemental Fig. S1). Sequencing data sets were deposited at Gene Expression Omnibus (GEO) under accession number GSE77607. PRO-seq normalization method We used a previously published Pol II ChIP-seq data set in Drosophila S2 cells (Teves and Henikoff 2011) to identify genes whose transcription does not change during HS. Unlike ChIP-seq reads, which can originate from both sense and antisense strands, PRO-seq reads are strand-specific. Therefore, in order to increase the likelihood of selecting genes that have the majority of their reads originating from the sense strand, we used our PRO-seq LacZ-RNAi control data sets (NHS and 20 min of HS) to identify and filter out genes that have high levels of transcription in the antisense strand. To identify those genes, we calculated the fraction of PRO-seq reads originated from the antisense strand for each gene and kept only the ones whose fraction was ≤0.2 for both the NHS and 20-min HS conditions. Because of the high background in ChIP-seq data, we then focused on genes with the highest levels of Ser2-P ChIP signal (Z-score > 3) (Core et al. 2012), assuming that these contained the highest densities of transcribing Pol II over background. In order to obtain a final subset of unaffected genes, we filtered out the ones whose gene body fold change between NHS and HS conditions was <0.85 and >1.15, resulting in 335 genes. The mRNA levels of this subset of genes were also unaffected after HS (Supplemental Fig. S2), which is consistent with the transcription levels of these genes not changing after induction. We then used the sum of the total number of gene body reads for all 335 genes to generate normalization factors in our PRO-seq data to normalize the data sets between replicates and different time points. Since the GAF-RNAi, HSF-RNAi, and M1BP-RNAi treatments did not result in genome-wide changes in transcription in both the NHS and 20-min HS time points, we used the same subset of 335 unaffected genes to normalize the data sets between different RNAi treatments (LacZ, GAF, HSF, and M1BP). To assess the efficacy of this normalization method, we examined the correlation between gene body and promoter reads for replicates (Supplemental Fig. S1) and gene body reads across different time points (Supplemental Fig. S3A,B). All replicates showed good correlation and time points that were closer to each other have higher correlation coefficients and better fits to the 1:1 diagonal. Moreover, for the RNAi treatments, we examined the correlation between gene body reads across different conditions (NHS and 20-min HS) and observed that the NHS data sets have higher correlation coefficients and better fits to the 1:1 diagonal when plotted against each other, and the same was observed for the HS treatments (Supplemental Fig. S3C,D). Taken together, these results indicate that the normalization method worked appropriately. Differential expression analysis using DESeq2 We used DESeq2 (Love et al. 2014) to identify genes whose gene body reads significantly change after HS, starting with a list of 9452 nonoverlapping genes described previously (Core et al. 2012). Gene body reads were collected from 200 bp downstream from the TSS, and we used different 3′ limits for each time point, assuming a conservative estimate for Pol II transcription elongation rate of 1 kb/min. We provided our own normalization factors for the DESeq2 calculations, which were obtained as described above. We used an FDR of 0.001 to identify activated and repressed genes. Unchanged genes were defined as having an adjusted P-value of >0.5 and log2 fold change higher than −0.25 and less than 0.25. Upstream transcription filter To minimize the number of false positives caused by changes in run-through transcription originated at the upstream gene, we implemented a filter to exclude from our analyses genes that have high levels of transcription in the region immediately upstream of the TSS. For each gene, we obtained the read counts from a window upstream of the TSS (−500 to −100 bp of the TSS) and a window in the gene body (+300 to +700 bp of the TSS) (Supplemental Fig. S4A). The 3′ limit of the upstream window was defined as −100 bp to the TSS to avoid confounding effects of potentially misannotated TSSs. In the case of the gene body window, the 5′ limit was defined as +300 bp to the TSS to avoid the region immediately downstream from the TSS, which can contain peaks of promoter-proximal paused Pol II. We then took the ratio of the read counts in these two regions for each gene, taking into consideration the number of mappable positions in the two windows (Supplemental Fig. S4A). This ratio was named the “upstream ratio” and was later used to exclude false positive genes from our analyses. In order to verify whether we could distinguish true and false positives based on the upstream ratio and define the appropriate cutoff to filter out false positive genes, we visually inspected 100 randomly selected HS-activated genes and classified each one as true or false positive based on the presence or absence of run-through transcription. The average mRNA levels (Supplemental Fig. S4B) are higher for the genes that were defined as true positives, which provides an independent verification of the criteria that were used to define true and false positives. The distribution of upstream ratios for the LacZ-RNAi HS condition was very distinct for true and false positives, with very little overlap (Supplemental Fig. S4C), indicating that this metric could be used to identify false positives. In order to define the optimal cutoff, we evaluated the performance of all potential cutoffs from 0 to 1 in 0.01 increments and used the accuracy metric (true positives + true negatives)/total to identify the cutoff with the best performance (Supplemental Fig. S4D). The horizontal line in Supplemental Fig. S4C represents the chosen cutoff (0.23). Filtering out genes with upstream ratios >0.23 eliminated all but one false positive, with only a minor loss of true positive genes. We then filtered out genes with upstream ratios >0.23 in the HS-activated and HS-repressed classes to generate the final subsets of genes that were used in all subsequent analyses. In order to use the condition with the highest PRO-seq levels for upstream ratio calculation, we used the NHS upstream ratio to filter the repressed subsets of genes and the HS upstream ratio to filter the activated subsets for every HS and RNAi treatment. GO analysis GO analysis on HS-activated and HS-repressed genes was performed using the Functional Annotation tool from DAVID (Huang et al. 2009), in which “GOTERM_BP_FAT” was selected. Composite profiles The composite profiles in Figures 2, A and C; 3D; 4, A–C; and 6E and Supplemental Figures S7A and S10 represent the median from 1000 subsamplings of 10% of the genes in each class, as described previously (Core et al. 2012; Danko et al. 2013). The shaded areas in Figure 2, A and C, and Supplemental Fig. S7A represent the 75% confidence intervals. Promoter-proximal pausing analysis The “pausing region” was defined as the 50-bp interval with the highest number of reads within −50 to +150 bp of the TSS. This region was defined using the LacZ-RNAi control NHS condition, and the same interval was used for all of the other treatments and conditions. The pausing index was then calculated as the ratio of the read density in the pausing region (reads/mappable bases) and the read density in the gene body (as defined above). Genes were classified as paused as described previously (Core et al. 2008). We used DESeq2 to identify genes whose pausing levels significantly change after HS using an FDR of 0.001. Transcription factor binding analysis We used bedtools closest (Quinlan and Hall 2010) to identify the closest HSF, GAF, or M1BP ChIP-seq peak to the TSS of every transcription unit in our list. HSF-, GAF-, or M1BP-bound genes were defined as having a ChIP-seq peak within ±1000 bp of the TSS. For the modENCODE factors, bound genes were defined as having a “regions_of_sig_enrichment” (from ChIP–chip gff3 file) within ±1000 bp of the TSS. De novo motif search De novo motif analysis of the promoter regions of HS-activated genes (−300 to +50 bp of the TSS) was performed using MEME (Bailey and Elkan 1994). Individual matches to the HSE's position weight matrix were identified by FIMO (Grant et al. 2011). RNA-seq data acquisition and analysis RNA-seq libraries were sequenced in 100-nt runs on the Illumina HiSeq using standard protocols at the Cornell Biotechnology Resource Center (http://www.BRC.cornell.edu). Illumina adapters were removed with the fastx_clipper tool (http://hannonlab.cshl.edu/fastx_toolkit/index.html), and sequencing reads <20 nt were discarded. Reads were aligned to the D. melanogaster dm3 reference genome/transcriptome using TopHat2 (Kim et al. 2013) with the following parameters: “--library-type fr-firststrand --no-novel-juncs.” Supplemental Table S3 contains a summary of sequencing yields and the number of reads that mapped to the genome/transcriptome or other annotations. Sequencing data sets are available under GEO accession number GSE77607. FPKM (fragments per kilobase per million mapped fragments) values for each gene were generated with Cuffnorm (Trapnell et al. 2010) using the BAM files generated by TopHat2 as input. Raw counts for each gene were obtained using HTSeq-count (Anders et al. 2014) and used as input for differential expression analysis using DESeq2 (Love et al. 2014). We used an FDR of 0.001 to identify genes whose mRNA levels significantly increase or decrease upon HS. Supplementary Material Supplemental Material Acknowledgments We thank Jian Li, Doug Baumann, and David Gilmour for providing the M1BP ChIP-seq peak information file and the M1BP antibody. We also thank current and past members of the Lis laboratory for insightful discussions and comments on the manuscript. Funding was provided by the National Institute of General Medical Sciences (R01GM025232). Supplemental material is available for this article. Article published online ahead of print. Article and publication date are online at http://www.genesdev.org/cgi/doi/10.1101/gad.284430.116. Freely available online through the Genes & Development Open Access option. ==== Refs References Anders S, Pyl PT, Huber W. 2014 HTSeq—a Python framework to work with high-throughput sequencing data. Bioinformatics 31 : 166–169.25260700 Ardehali MB, Yao J, Adelman K, Fuda NJ, Petesch SJ, Webb WW, Lis JT. 2009 Spt6 enhances the elongation rate of RNA polymerase II in vivo. EMBO J 28 : 1067–1077.19279664 Bailey TL, Elkan C. 1994 Fitting a mixture model by expectation maximization to discover motifs in biopolymers. Proc Int Conf Intell Syst Mol Biol 2 : 28–36.7584402 Bienz M, Pelham HRB. 1986 Heat shock regulatory elements function as an inducible enhancer in the Xenopus hsp70 gene and when linked to a heterologous promoter. Cell 45 : 753–760.3085957 Buckley MS, Kwak H, Zipfel WR, Lis JT. 2014 Kinetics of promoter Pol II on Hsp70 reveal stable pausing and key insights into its regulation. Genes Dev 28 : 14–19.24395245 Celniker SE, Dillon LAL, Gerstein MB, Gunsalus KC, Henikoff S, Karpen GH, Kellis M, Lai EC, Lieb JD, MacAlpine DM, 2009 Unlocking the secrets of the genome. Nature 459 : 927–930.19536255 Core LJ, Waterfall JJ, Lis JT. 2008 Nascent RNA sequencing reveals widespread pausing and divergent initiation at human promoters. Science 322 : 1845–1848.19056941 Core LJ, Waterfall JJ, Gilchrist DA, Fargo DC, Kwak H, Adelman K, Lis JT. 2012 Defining the status of RNA polymerase at promoters. Cell Rep 2 : 1025–1035.23062713 Danko CG, Hah N, Luo X, Martins AL, Core L, Lis JT, Siepel A, Kraus WL. 2013 Signaling pathways differentially affect RNA polymerase II initiation, pausing, and elongation rate in cells. Mol Cell 50 : 212–222.23523369 Fuda NJ, Guertin MJ, Sharma S, Danko CG, Martins AL, Siepel A, Lis JT. 2015 GAGA factor maintains nucleosome-free regions and has a role in RNA polymerase II recruitment to promoters. PLoS Genet 11 : e1005108.25815464 Gan M, Moebus S, Eggert H, Saumweber H. 2011 The Chriz–Z4 complex recruits JIL-1 to polytene chromosomes,a requirement for interband-specific phosphorylation of H3S10. J Biosci 36 : 425–438.21799255 Glaser RL, Thomas GH, Siegfried E, Elgin SC, Lis JT. 1990 Optimal heat-induced expression of the Drosophila hsp26 gene requires a promoter sequence containing (CT)n.(GA)n repeats. J Mol Biol 211 : 751–761.2313697 Gonsalves SE, Moses AM, Razak Z, Robert F, Westwood JT. 2011 Whole-genome analysis reveals that active heat shock factor binding sites are mostly associated with non-heat shock genes in Drosophila melanogaster. PLoS One 6 : e15934.21264254 Grant CE, Bailey TL, Noble WS. 2011 FIMO: scanning for occurrences of a given motif. Bioinformatics 27 : 1017–1018.21330290 Guertin MJ, Lis JT. 2010 Chromatin landscape dictates HSF binding to target DNA elements. PLoS Genet 6 : e1001114.20844575 Guertin MJ, Petesch SJ, Zobeck KL, Min IM, Lis JT. 2010 Drosophila heat shock system as a general model to investigate transcriptional regulation. Cold Spring Harb Symp Quant Biol 75 : 1–9.21467139 Guhathakurta D, Palomar L, Stormo GD, Tedesco P, Johnson TE, Walker DW, Lithgow G, Kim S, Link CD. 2002 Identification of a novel cis-regulatory element involved in the heat shock response in Caenorhabditis elegans using microarray gene expression and computational methods. Genome Res 712 : 701–712.11997337 Henriques T, Gilchrist DA, Nechaev S, Bern M, Muse GW, Burkholder A, Fargo DC, Adelman K. 2013 Stable pausing by RNA polymerase II provides an opportunity to target and integrate regulatory signals. Mol Cell 52 : 517–528.24184211 Hou C, Li L, Qin ZS, Corces VG. 2012 Gene density, transcription, and insulators contribute to the partition of the Drosophila genome into physical domains. Mol Cell 48 : 471–484.23041285 Huang DW, Sherman BT, Lempicki RA. 2009 Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc 4 : 44–57.19131956 Jamrich M, Greenleaf AL, Bautz EK. 1977 Localization of RNA polymerase in polytene chromosomes of Drosophila melanogaster. Proc Natl Acad Sci 74 : 2079–2083.405671 Jonkers I, Kwak H, Lis JT. 2014 Genome-wide dynamics of Pol II elongation and its interplay with promoter proximal pausing, chromatin, and exons. Elife 3 : e02407.24843027 Kent WJ, Sugnet CW, Furey TS, Roskin KM, Pringle TH, Zahler AM, Haussler D. 2002 The human genome browser at UCSC. Genome Res 12 : 996–1006.12045153 Kharchenko PV, Alekseyenko AA, Schwartz YB, Minoda A, Riddle NC, Ernst J, Sabo PJ, Larschan E, Gorchakov AA, Gu T, 2011 Comprehensive analysis of the chromatin landscape in Drosophila melanogaster. Nature 471 : 480–485.21179089 Kim D, Pertea G, Trapnell C, Pimentel H, Kelley R, Salzberg SL. 2013 TopHat2: accurate alignment of transcriptomes in the presence of insertions, deletions and gene fusions. Genome Biol 14 : R36.23618408 Kwak H, Fuda NJ, Core LJ, Lis JT. 2013 Precise maps of RNA polymerase reveal how promoters direct initiation and pausing. Science 339 : 950–953.23430654 Langmead B, Trapnell C, Pop M, Salzberg SL. 2009 Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol 10 : R25.19261174 Lee H, Kraus KW, Wolfner MF, Lis JT. 1992 DNA sequence requirements for generating paused polymerase at the start of hsp70. Genes Dev 6 : 284–295.1737619 Lee C, Li X, Hechmer A, Eisen M, Biggin MD, Venters BJ, Jiang C, Li J, Pugh BF, Gilmour DS. 2008 NELF and GAGA factor are linked to promoter-proximal pausing at many genes in Drosophila. Mol Cell Biol 28 : 3290–3300.18332113 Leemans R, Egger B, Loop T, Kammermeier L, He H, Hartmann B, Certa U, Hirth F, Reichert H. 2000 Quantitative transcript imaging in normal and heat-shocked Drosophila embryos by using high-density oligonucleotide arrays. Proc Natl Acad Sci 97 : 12138–12143.11035778 Li J, Gilmour DS. 2013 Distinct mechanisms of transcriptional pausing orchestrated by GAGA factor and M1BP, a novel transcription factor. EMBO J 32 : 1829–1841.23708796 Li L, Lyu X, Hou C, Takenaka N, Nguyen HQ, Ong C-T, Cubeñas-Potts C, Hu M, Lei EP, Bosco G, 2015 Widespread rearrangement of 3D chromatin organization underlies polycomb-mediated stress-induced silencing. Mol Cell 58 : 216–231.25818644 Lindquist S, Craig EA. 1988 The heat-shock proteins. Annu Rev Genet 22 : 631–677.2853609 Lindquist S, Petersen R. 1990 Selective translation and degradation of heat-shock messenger RNAs in Drosophila. Enzyme 44 : 147–166.2133647 Lis JT, Neckameyer W, Dubensky R, Costlow N. 1981 Cloning and characterization of nine heat-shock-induced mRNAs of Drosophila melanogaster. Gene 15 : 67–80.6795091 Lis JT, Mason P, Peng J, Price DH, Werner J. 2000 P-TEFb kinase recruitment and function at heat shock loci. Genes Dev 14 : 792–803.10766736 Love MI, Huber W, Anders S. 2014 Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 15 : 550.25516281 Lu Q, Wallrath LL, Granok H, Elgin SC. 1993 (CT)n (GA)n repeats and heat shock elements have distinct roles in chromatin structure and transcriptional activation of the Drosophila hsp26 gene. Mol Cell Biol 13 : 2802–2814.8474442 Mahat DB, Salamanca HH, Duarte FM, Danko CG, Lis JT. 2016 Mammalian heat shock response and mechanisms underlying its genome-wide transcriptional regulation. Mol Cell 62 : 63–78.27052732 Murray JI, Whitfield ML, Trinklein ND, Myers RM, Brown PO, Botstein D. 2004 Diverse and specific gene expression responses to stresses in cultured human cells. Mol Biol Cell 15 : 2361–2374.15004229 O'Brien T, Lis JT. 1993 Rapid changes in Drosophila transcription after an instantaneous heat shock. Mol Cell Biol 13 : 3456–3463.8497261 O'Brien T, Wilkins RC, Giardina C, Lis JT. 1995 Distribution of GAGA protein on Drosophila genes in vivo. Genes Dev 9 : 1098–1110.7744251 Ohtsuki S, Levine M. 1998 GAGA mediates the enhancer blocking activity of the eve promoter in the Drosophila embryo. Genes Dev 12 : 3325–3330.9808619 Omichinski JG, Pedone PV, Felsenfeld G, Gronenborn AM, Clore GM. 1997 The solution structure of a specific GAGA factor–DNA complex reveals a modular binding mode. Nat Struct Biol 4 : 122–132.9033593 Quinlan AR, Hall IM. 2010 BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics 26 : 841–842.20110278 Rahl PB, Lin CY, Seila AC, Flynn RA, McCuine S, Burge CB, Sharp PA, Young RA. 2010 c-Myc regulates transcriptional pause release. Cell 141 : 432–445.20434984 Rasmussen EB, Lis JT. 1993 In vivo transcriptional pausing and cap formation on three Drosophila heat shock genes. Proc Natl Acad Sci 90 : 7923–7927.8367444 Rath U, Ding Y, Deng H, Qi H, Bao X, Zhang W, Girton J, Johansen J, Johansen KM. 2006 The chromodomain protein, Chromator, interacts with JIL-1 kinase and regulates the structure of Drosophila polytene chromosomes. J Cell Sci 119 : 2332–2341.16723739 Rougvie AE, Lis JT. 1988 The RNA polymerase II molecule at the 5′ end of the uninduced hsp70 gene of D. melanogaster is transcriptionally engaged. Cell 54 : 795–804.3136931 Saunders A, Werner J, Andrulis ED, Nakayama T, Hirose S, Reinberg D, Lis JT. 2003 Tracking FACT and the RNA polymerase II elongation complex through chromatin in vivo. Science 301 : 1094–1096.12934007 Schwartz YB, Linder-Basso D, Kharchenko PV, Tolstorukov MY, Kim M, Li H-B, Gorchakov AA, Minoda A, Shanower G, Alekseyenko AA, 2012 Nature and function of insulator protein binding sites in the Drosophila genome. Genome Res 22 : 2188–2198.22767387 Schweinsberg S, Hagstrom K, Gohl D, Schedl P, Kumar RP, Mishra R, Karch F. 2004 The enhancer-blocking activity of the Fab-7 boundary from the Drosophila bithorax complex requires GAGA-factor-binding sites. Genetics 168 : 1371–1384.15579691 Sexton T, Yaffe E, Kenigsberg E, Bantignies F, Leblanc B, Hoichman M, Parrinello H, Tanay A, Cavalli G. 2012 Three-dimensional folding and functional organization principles of the Drosophila genome. Cell 148 : 458–472.22265598 Shalgi R, Hurt JA, Lindquist S, Burge CB. 2014 Widespread inhibition of posttranscriptional splicing shapes the cellular transcriptome following heat shock. Cell Rep 7 : 1362–1370.24857664 Shopland LS, Hirayoshi K, Fernandes M, Lis JT. 1995 HSF access to heat shock elements in vivo depends critically on promoter architecture defined by GAGA factor, TFIID, and RNA polymerase II binding sites. Genes Dev 9 : 2756–2769.7590251 Sørensen JG, Nielsen MM, Kruhøffer M, Justesen J, Loeschcke V. 2005 Full genome gene expression analysis of the heat stress response in Drosophila melanogaster. Cell Stress Chaperones 10 : 312.16333985 Spradling A, Penman S, Pardue ML. 1975 Analysis of Drosophila mRNA by in situ hybridization: sequences transcribed in normal and heat shocked cultured cells. Cell 4 : 395–404.1122559 Teves SS, Henikoff S. 2011 Heat shock reduces stalled RNA polymerase II and nucleosome turnover genome-wide. Genes Dev 25 : 2387–2397.22085965 Trapnell C, Williams BA, Pertea G, Mortazavi A, Kwan G, van Baren MJ, Salzberg SL, Wold BJ, Pachter L. 2010 Transcript assembly and quantification by RNA-Seq reveals unannotated transcripts and isoform switching during cell differentiation. Nat Biotechnol 28 : 511–515.20436464 Trinklein ND, Murray JI, Hartman SJ, Botstein D, Myers RM. 2004 The role of heat shock transcription factor 1 in the genome-wide regulation of the mammalian heat shock response. Mol Biol Cell 15 : 1254–1261.14668476 Tsukiyama T, Becker PB, Wu C. 1994 ATP-dependent nucleosome disruption at a heat-shock promoter mediated by binding of GAGA transcription factor. Nature 367 : 525–532.8107823 Vihervaara A, Sergelius C, Vasara J, Blom MAH, Elsing AN, Roos-Mattjus P, Sistonen L. 2013 Transcriptional response to stress in the dynamic chromatin environment of cycling and mitotic cells. Proc Natl Acad Sci 110 : E3388–E3397.23959860 Wilkins R. 1998 GAGA factor binding to DNA via a single trinucleotide sequence element. Nucleic Acids Res 26 : 2672–2678.9592153 Wu C. 1995 Heat shock transcription factors: structure and regulation. Annu Rev Cell Dev Biol 11 : 441–469.8689565 Xiao H, Lis J. 1988 Germline transformation used to define key features of heat-shock response elements. Science 239 : 1139–1142.3125608 Yost HJ, Lindquist S. 1986 RNA splicing is interrupted by heat shock and is rescued by heat shock protein synthesis. Cell 45 : 185–193.2421918 Zhao K, Hart CM, Laemmli UK. 1995 Visualization of chromosomal domains with boundary element-associated factor BEAF-32. Cell 81 : 879–889.7781065

PMC005xxxxxx/PMC5002982.txt

==== Front eNeuroeNeuroeneuroeneuroeNeuroeNeuro2373-2822Society for Neuroscience 10.1523/ENEURO.0039-16.2016eN-NWR-0039-168New ResearchSensory and Motor SystemsRepresentation of Perceptual Color Space in Macaque Posterior Inferior Temporal Cortex (the V4 Complex) Color Representation in PIT (V4 Complex)Bohon Kaitlin S. 12*Hermann Katherine L. 12*Hansen Thorsten 3Conway Bevil R. 12†1 Program in Neuroscience, Wellesley College, Wellesley, Massachusetts 024812 Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, Massachusetts 021393 Department of Psychology, Justus Liebig University Giessen, 35396 Giessen, GermanyThe authors declare no competing financial interests. Author contributions: B.R.C. designed research; B.R.C. performed research; T.H. contributed unpublished reagents/analytic tools; K.S.B., K.L.H., and B.R.C. analyzed data; B.R.C. wrote the paper. This research was supported by National Institutes of Health Grant EY023322 and National Science Foundation Grant 1353571. * K.S.B. and K.L.H. contributed equally as co-first authors. † Current address: Laboratory of Sensorimotor Research, National Eye Institute, National Institutes of Health, Bethesda MD. Correspondence should be addressed to Bevil R. Conway, Laboratory of Sensorimotor Research, National Eye Institute, National Institutes of Health, Bethesda, Maryland 20892. E-mail: bevil@mit.edu.12 8 2016 29 8 2016 Jul-Aug 2016 3 4 ENEURO.0039-16.201624 2 2016 19 7 2016 4 8 2016 Copyright © 2016 Bohon et al.2016Bohon et al.This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International, which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed.Abstract The lateral geniculate nucleus is thought to represent color using two populations of cone-opponent neurons [L vs M; S vs (L + M)], which establish the cardinal directions in color space (reddish vs cyan; lavender vs lime). How is this representation transformed to bring about color perception? Prior work implicates populations of glob cells in posterior inferior temporal cortex (PIT; the V4 complex), but the correspondence between the neural representation of color in PIT/V4 complex and the organization of perceptual color space is unclear. We compared color-tuning data for populations of glob cells and interglob cells to predictions obtained using models that varied in the color-tuning narrowness of the cells, and the color preference distribution across the populations. Glob cells were best accounted for by simulated neurons that have nonlinear (narrow) tuning and, as a population, represent a color space designed to be perceptually uniform (CIELUV). Multidimensional scaling and representational similarity analyses showed that the color space representations in both glob and interglob populations were correlated with the organization of CIELUV space, but glob cells showed a stronger correlation. Hue could be classified invariant to luminance with high accuracy given glob responses and above-chance accuracy given interglob responses. Luminance could be read out invariant to changes in hue in both populations, but interglob cells tended to prefer stimuli having luminance contrast, regardless of hue, whereas glob cells typically retained hue tuning as luminance contrast was modulated. The combined luminance/hue sensitivity of glob cells is predicted for neurons that can distinguish two colors of the same hue at different luminance levels (orange/brown). colormacaquemonkeyneurophysiologyv4visionNIHEY023322NSF1353571 cover-dateJuly/August 2016 ==== Body Significance Statement This article provides the first quantitative test of the correspondence between the neural representation of color in posterior inferior temporal cortex (PIT; the V4 complex) and the organization of perceptual color space. fMRI-guided micoelectrode recording was used to target two subpopulations of neurons within the PIT/V4 complex, globs and interglobs. The results suggest the following: (1) glob cells have narrow color tuning, and as a population have a uniform representation of color space with a bias for warm colors; and (2) glob cells provide a neural correlate for the psychophysical distinction between two colors that have the same hue but differ in luminance (e.g., orange/brown). The work also underscores the importance of carefully controlled stimuli in neurophysiological studies of color. Introduction Colors can be organized into a uniform color space in which adjacent colors are separated to a similar degree using perceptual thresholds (Munsell, 1907; MacAdam, 1990; Commission Internationale de l′Éclairage, 2004). At the same time, some colors—the unique hues (red, green, blue, yellow, black, white)—are widely considered psychologically more important than other colors (Hurvich and Jameson, 1974). The neural basis for the uniformity of color space, on the one hand, and the specialness of certain colors, on the other hand, is unknown. These features are not a trivial consequence of the spectrum: the spectrum is continuous and linear, whereas color is categorical and color space forms a circle (purple, not in the spectrum, sits where the circle closes). The retina reduces spectral information to three numbers, represented by the activity of the three, broadly tuned, classes of cone photoreceptors (L, M, S). How are cone signals processed by the brain to bring about color perception? We take up this question by asking how color is represented in a mid-tier area halfway along the putative visual-processing hierarchy [posterior inferior temporal cortex (PIT); the V4 complex]. Color-coding cells earlier on in processing, in the retina and lateral geniculate nucleus (LGN), are disproportionately sensitive to some colors (Derrington et al., 1984; Sun et al., 2006): one set of neurons responds best to colors that modulate L/M activity (without altering S activity), appearing reddish (L+, M−) or bluish-green (M+, L−); the second set responds to colors that selectively modulate S activity, appearing lavender (S+) or lime (S−). These color biases define the physiologically important cardinal directions (MacLeod and Boynton, 1979), but their impact on color categorization, if any, is not well understood. They do not correspond to the unique hues (Webster et al., 2000). Is color space represented anywhere in the brain, such that the proportion of cells tuned to each color in the space is equal? Is there a neural correlate for unique hues? Physiology in the retina (and LGN), along with behavioral adaptation experiments (Krauskopf et al., 1982; Eskew, 2009), and some microelectrode recording studies in V1 (Tailby et al., 2008a; Horwitz and Hass, 2012), raise the possibility that color depends on a population code, in which each color is defined by the relative activation of the two sets of cardinal-tuned neurons. Other studies (Webster and Mollon, 1991; Hansen and Gegenfurtner, 2006; Stoughton et al., 2012) suggest that color depends on multiple, independent mechanisms that together comprise a uniform space. Candidate neural substrates include cells in V2 (Moutoussis and Zeki, 2002; Xiao et al., 2003) and V4 (Zeki, 1980; Li et al, 2014), and cells in subregions of inferior temporal cortex (Komatsu et al., 1992; Conway and Tsao, 2006; Conway et al., 2007; Yasuda et al., 2010; Lafer-Sousa and Conway, 2013). One possibility is that color is encoded by a population code early in processing, which is then decoded by subsequent stages (De Valois and De Valois, 1993; Zeki and Marini, 1998; Conway, 2009; Zaidi et al., 2014). We analyzed data from fMRI-guided microelectrode recording of millimeter-sized, color-biased globs and adjacent non-color-biased interglobs in V4/PIT (Conway et al., 2007). The use of fMRI is valuable since it provides an independent means for identifying functional subdomains. Most cells in the globs are not only color tuned but also show tuning that is tolerant to luminance modulation (Conway et al., 2007; Namima et al., 2014). Moreover, glob cells are spatially organized into chromotopic maps (Conway and Tsao, 2009), consistent with a role in representing perceptual color space (Zaidi et al., 2014). A preliminary analysis suggested that the glob cells, as a population, might show a bias for the unique hues (Stoughton and Conway, 2008). Mollon (2009) has challenged this idea, arguing that variation in stimulus saturation caused the apparent biases. Mollon (2009) suggested that the data might be accounted for by a population of linearly tuned neurons biased toward the cardinal directions (such as in the LGN). To address these issues, we compared the color tuning of the population of recorded glob and interglob cells against model predictions that capture a range of theoretical possibilities incorporating the extent to which the neural tuning reflects a linear versus nonlinear combination of cone signals (narrowness of tuning), and the extent to which the color preferences across the population uniformly represent color space. The analyses suggest that the color representation in glob cells is different from the representation in the LGN: glob cells most likely possess nonlinear narrow color tuning that, as a population, represent a perceptually uniform color space with a bias toward “warm” colors (reds/yellows) over “cool” colors (blues/greens). The analyses also underscore the importance of future work to determine neural color tuning using stimuli that fully sample a uniform color space. Materials and Methods Single-unit recording The physiological data were collected as part of a previous report (Conway et al., 2007), and all the details of the recording are described in that report. Tungsten microelectrodes were used to target microelectrode recording to functionally defined domains identified using fMRI in the same animals. Electrodes were inserted into the brain for recording sessions that lasted several hours, and then the electrodes were removed. Anatomical MR images were obtained following many microelectrode recording sessions and confirmed the placement of the electrodes. Single-unit responses were measured in two animals trained to fixate a spot on a computer monitor using standard procedures and apparatus (BAK Electronics). Complete color-tuning responses for 300 glob cells and 181 interglob cells were used in the present analysis. By combining the information from the anatomical scans and the depth information obtained during the recordings, the locations of the recorded cells were correlated with the functional maps and categorized as residing in a glob or an interglob. The anterior boundary of area V4 was initially not obvious in fMRI mapping (Fize et al., 2003), prompting use of the term “the V4 complex” or the anatomical term “posterior inferior temporal,” as used presently; subsequent evidence shows a clear boundary between V4 and inferior temporal cortex (Lafer-Sousa and Conway, 2013). Visual stimuli for single-unit experiments Optimal stimulus dimensions (bar length, width, and position) were used for each cell. The shape and location were fixed for a given cell, and the color of the shape was then varied. A total of 135 colors were used, consisting of three sets of 45 colors; the colors within a set were equiluminant with each other, spanned the full color gamut of the monitor, and were as saturated as the monitor could produce (Fig. 1; Stoughton and Conway, 2008, their Table S1, CIE coordinates). The colors of one set were higher luminance (7.8 cd/m2) than the background; those of another set were photometrically equiluminant with the background (3.05 cd/m2); and those of the third set were of lower luminance than the background (0.62 cd/m2). All colors, including those at the lowest luminance, had discernable color to human observers. The two color sets of equal or high luminance to the adapting background were vividly colored; stimuli of the low-luminance set may be considered mesopic and could have involved rod activation. All stimuli were surrounded by an adapting background of 3.05 cd/m2. Luminance artifacts could cause different amplitude responses to different colors, which could be (inaccurately) interpreted as color tuning. Using stimuli at different luminance levels, and testing for luminance-invariant color tuning, provides one way of controlling for luminance artifacts arising from, for example, chromatic aberration or variability in macular pigmentation. Responses to black (0.02 cd/m2) and white (78.2 cd/m2) were also measured. The different colors were presented in pseudorandom order. Within the time period during which the three sets of colors were presented, white and black versions of the stimulus were each presented three times, so that one complete cycle consisted of 141 stimulus presentations (color set 1, 45 colors; color set 2, 45 colors; color set 3, 45 colors; white, 3; black, 3). The full stimulus set at any given luminance level did not sample the monitor gamut at regular intervals. To do so, in some analyses we subsampled the data, extracting responses to 21 hues (at each luminance level) that were more or less evenly spaced when plotted in CIELUV color space (Commission Internationale De Leclairage, 2004), which is designed to be perceptually uniform (Fig. 1, outlined points connected by spokes to the neutral point). We acknowledge that all color spaces, including CIELUV, are not entirely perceptually uniform, and the deviations from uniformity can be pronounced (Melgosa et al., 1994). To address this issue, we converted our stimuli to CIECAM02 space, a space thought to remedy some of the defects in uniformity found in CIELUV space (these defects are apparent when comparing colors across large distances in the space; Moroney et al., 2002; Luo et al., 2006). We calculated the CIECAM02 Cartesian coordinates and CIECAM02 color angle for each stimulus using the MATLAB (RRID:SCR_001622) function CIECAM02 (Computational Colour Science Toolbox; Ripamonti et al., 2013). Results using CIECAM02 were qualitatively very similar to the results obtained using the CIELUV space and did not change the conclusions (data not shown). We note that the CIECAM02 space is itself not perfect. Indeed, determining a perceptually uniform space remains a persistent challenge, both psychophysically and theoretically. Our long-term goal is to use neurophysiological data from populations of neurons shown to be involved in color to bootstrap a psychophysical definition of perceptual uniformity, and to use this information to determine the theoretical basis for the color space. Figure 1. Chromaticity coordinates of color stimuli. A–C, chromaticity coordinates of the stimuli in CIELUV space for the high-luminance (A), equiluminant (B), and low-luminance (C) stimulus sets. Cell responses to 45 hues at each luminance level were measured. Markers with black outlines denote the 21 subsampled hues. The black lines connect each of the 21 hues with the origin (gray cross). Note that the angles between each of the 21 hues are relatively uniform. Black dots denote 17 hue-matched stimuli used for decoding. Inset in A shows stimuli in MB-DKL color space. The use of monkeys to investigate the neural basis for human color perception is licensed because monkeys have very similar color detection thresholds and psychophysical mechanisms to those found in humans (Stoughton et al., 2012; Gagin et al., 2014). Responses to multiple presentations of the same stimulus were averaged together. Each stimulus was displayed for 200 ms and separated in time from the previous and subsequent stimuli by 200 ms, during which time the animal was rewarded for maintaining constant fixation. Estimates of stimulus saturation The stimuli used in the original study by Conway et al. (2007) were the most saturated that the monitor could produce. The limitation of these stimuli is that there is likely considerable variability in the saturation across stimuli of different hue, confounding saturation, and hue. In an attempt to model the impact of saturation on neural responses, we estimated the saturation for each stimulus. Saturation can be defined in numerous ways, although there is no consensus; moreover, it is unlikely that the neural responses vary linearly with changes in saturation. Nonetheless, we assume linearity because the neural response to saturation has not been empirically determined. We defined saturation for each stimulus in both MB-DKL color space (a physiologically defined cone-opponent space; MacLeod and Boynton, 1979; Derrington et al, 1984) and LUV space (a perceptually defined color space). For MB-DKL saturation, we calculated the distance between the stimulus and the adapting gray point. The MB-DKL location of each stimulus was calculated with a CIE-to-MB-DKL conversion matrix from the spectra of each of the primaries of the monitor at maximum strength (Zaidi and Halevy, 1993; Hansen and Gegenfurtner, 2013). MB-DKL saturation was used to assess the hypothesis that neurophysiological data matches the activity in the LGN. For LUV saturation, we calculated the ratio of the distance between the stimulus and the adapting gray point, over the distance between the gray point and the spectrum locus through the stimulus; this definition was used to test the hypothesis that the neurophysiological data explain psychologically important colors, the unique hues. Preprocessing of cell responses Every visually responsive cell that was tested was included in the analysis if responses to at least two complete stimulus cycles were obtained; in most cases, responses to at least five stimulus cycles were obtained. Most cells responded with higher firing rates compared with baseline values. A small number of cells was suppressed by the majority of stimuli at some or all luminance levels. Five glob cells and one interglob cell were on average suppressed at all luminance levels. Nine glob cells and four interglob cells were on average suppressed at one or two luminance levels. But even among these cells, none were suppressed below baseline activity by all stimuli: there was always at least one hue, at one luminance level, that elicited a response that was higher than the baseline activity. For all cells, we calculated the stimulus responses by summing spikes during a window that was optimized for the response duration of each cell, for each luminance level. The time window began with the visual latency, which was defined as >2.5 SDs above the background firing rate, and ended either when the response rate fell below the background firing rate plus 2.5 SDs, or after a period slightly shorter than the stimulus duration (reduced by one-quarter of the latency time of each cell), whichever was shorter. Capping the integration window avoided integrating over OFF responses. Across all cells, the average latency was 78 ms (SD, 28 ms). The average integration window was 146 ms (SD, 52 ms). Curve fitting In order to estimate the narrowness of the response of each cell, we fit the responses with a curve. Responses to the 21 evenly spaced stimuli at each luminance level were smoothed using a boxcar filter (across one stimulus), and fit with a model tuning curve adapted from Seung and Sompolinsky (1993), according to Equation 1, as follows: rθ=fmin+fmax*cos⁡πθ-α2w2, |θ|<wfmin,  otherwise where fmin is the baseline firing rate, fmax is the maximum firing rate, w is the tuning width (full-width at half-maximum) in radians, and α is the peak tuning angle. As w increases, the cell becomes more broadly tuned; as w decreases, the cell becomes more narrowly tuned. A linear cell, such as those found in the LGN, has a w value of π radians, equivalent to 180°. We chose to curve-fit responses to only the 21 evenly spaced in CIELUV angle stimuli in order to avoid biasing the curves to fit values closer to the monitor primaries, which were oversampled in the 45-hue set. We chose to boxcar smooth responses prior to curve fitting in order to decrease noise and improve the fit. Results obtained using all 45 stimuli and unsmoothed responses yielded similar conclusions. Results obtained using a half wave-rectified cosine exponent curve (De Valois et al., 2000) yielded similar conclusions (data not shown). Is it possible that our sampling of colors space (21 angles ∼17° apart) was too coarse to obtain a good estimate of narrowness? To assess this, we compared the tuning width estimates with those obtained using responses to all 45 hues. If a 21-hue set is insufficient to provide an accurate estimate of tuning width, we would expect the tuning widths to be narrower when using the responses to more dense sampling of color space, especially for cells with tuning peaks located in the part of color space most densely sampled by the 45 colors. The 45-color stimulus set sampled most densely the color space around the monitor primaries. We did not find systematic differences of the tuning widths estimated with either approach (data not shown), suggesting that 21 hues sampled densely enough to accurately reflect the tuning widths of the neurons. Model populations The color space represented by a population of neurons can be defined by varying the following two parameters: the narrowness of the color-tuning function for each cell (w); and the uniformity with which the population of neurons samples color space. Our goal was to determine the combination of these parameters that best describes the color representation in the globs and interglobs, and to compare these parameters to those evident in the LGN. We performed multiple iterations of a model simulation, parametrically varying the linearity and degree of uniformity in a population of model cells; we used 181 model cells on each iteration of the model (corresponding to the number of cells recorded in the interglobs). To compare the model to the glob population, we randomly sampled 181 units from the total 300 glob cells recorded. This subsampling was performed in order to equate the number of cells between the glob and interglob populations to allow a direct comparison of the best-fitting models achieved for the two populations of neurons. To simulate an entirely uniform population of linear glob cells, each model cell was assigned a tuning width of 180, and a random peak tuning angle drawn from a uniform distribution of integers between 1 and 360. A tuning function was generated using Equation 1. The baseline firing rate and maximum firing rate were not varied, as all responses (both model and real cells) were normalized to facilitate comparison between recorded and model cells; for the simulated neurons, the minimum firing rate was set to 0 and the maximum firing rate was set to 1. To account for differences in saturation among the stimuli, the response of the model cell to each stimulus was multiplied prior to normalization by the saturation of that stimulus, as described above. A stimulus with higher saturation would more greatly affect the tuning curve of the model cell than a stimulus with lower saturation. The 181 model cells were then rank-ordered by the angle to which they showed peak response. The neural populations were also rank ordered by the LUV angle at which they maximally fired. R 2 values were then computed between the tuning function of each model cell (defined by both tuning width and peak tuning angle) and its corresponding (rank-ordered) recorded neuron (raw responses to all 45 hues). This procedure was performed 1000 times. The success of the match between the population of recorded cells and the model simulation was defined as the median R 2 value across all 181,000 comparisons. We determined the R 2 values for simulations in which the narrowness varied from 84 to 360 CIELUV degrees. The model responses were compared with unsmoothed responses obtained to all 45 stimuli of a given luminance level. We also determined R 2 values for simulations in which the uniformity of color space varied. We tested the following two hypotheses: first, that the population reflected the distortions of the color space manifest in the retina and LGN; and, second, that the population reflected the distortions of color space predicted by the purported privilege of the unique hues. To simulate a population of LGN cells, each model cell had a randomly assigned α drawn from a distribution of values within 5° of the cardinal angles in LUV space (353°, 100°, 173°, 280°). To simulate a population of unique hue-biased cells, each model cell had a randomly assigned α drawn from a distribution of values within 5° of the unique hue angles (14°, 70°, 139°, and 238°). The degree of nonuniformity within each model simulation was then systematically varied by adjusting the fraction of the model cells that were defined as nonuniform (LGN or unique hue) versus uniform. We summarize the conclusions of the model simulations in heat maps of the median R 2 values across the 181,000 comparisons (1000 iterations × 181 recorded-model cell pairs) at each narrowness–uniformity pairing. The darker the cell of the heat map, the better overall correlation there was between the model population and the real population. Black/white boxes indicate the best-matching simulated population, and numbers report the median R 2 for the best-matching simulated population. Receiver operating characteristic analysis To test whether cells in the glob and interglob populations discriminate stimuli based on luminance, we performed a receiver operating characteristic (ROC) analysis (Britten et al., 1992; Mayo and Sommer, 2013). We z-scored the raw mean firing rates of each cell to the full stimulus set (45 hues at each of three luminances). For each cell, for each pair of luminance categories (equiluminant/low-luminance, high-luminance/equiluminant), we used the perfcurve function in Matlab R2013b to compute an area under the curve (AUC) of the ROC of the cell. An AUC of 0.5 would indicate chance discrimination between the two luminance categories. An AUC < 0.5 would indicate a preference for the first luminance category. An AUC > 0.5 would indicate a preference for the second luminance category. To determine whether the AUC of a cell was significantly different than chance, we performed a permutation test in which, for each of 2000 iterations, we performed the ROC procedure but with randomly shuffled luminance category labels. This yielded a null distribution for which we computed 95% confidence intervals (CIs); if the AUC of a cell fell outside these bounds, it was deemed significant at p < 0.05. Analysis of peak shifting as a function of luminance To determine the effect of stimulus luminance on the glob and interglob hue preferences, we quantified the change in color-tuning preferences across luminance levels. We sought to test whether or not cells maintained the same hue-tuning preference across luminance levels; for example, does a cell that responds best to purple at equiluminance also prefer purple at high-luminance? This test sheds light on the possible involvement of rods. Although the adapting state maintained by the surrounding neutral gray was likely photopic (3.05 cd/m2), the stimuli of the low-luminance set had relatively low luminance, raising the possibility that they activate rods. Moreover, some prior work would also suggest that the stimuli of the other luminance sets might also involve rods. If rods were implicated in driving the neural responses, one might expect systematic shifts in color tuning as the luminance is changed (Shapiro et al, 1996). Our results do not provide conclusive evidence for such shifts (see Fig. 9). To assess the extent to which neurons of each population (glob and interglob) shifted their peak hue tuning, we compared the correlation of the peak tuning determined at different luminance levels. We calculated the Pearson’s correlation coefficient (r) between the peak determined using stimuli at one luminance level and the peak determined using stimuli at a different luminance level following 200 bootstraps of the responses of half of each population. We performed a t test on the glob and interglob distributions of Fisher’s z-transform-corrected r values. In order to calculate 95% CIs on these p values, we performed the 200 bootstraps 1000 times, and calculated the CIs using the percentile method. The reported p value for each comparison is the median of the 1000 p value distribution. In order to test for systematic differences in peak shifting across luminance levels between cells tuned to different hues, we calculated peak shifting within groups of neurons defined by their color preferences assessed using the equiluminant stimulus set. We categorized the cells into eight color categories, each spanning 45° in color space. We used a Mann–Whitney–Wilcoxon U test to determine whether the tuning of the population in each category shifted when tuning was assessed at different luminance levels. The 95% confidence intervals on this p value were obtained by doing 2000 bootstraps, in which the p value was calculated using a random selection of 90% of the cells. The reported p values are the mean p value over these bootstraps. This analysis was performed for both glob cells and interglob cells, comparing tuning at equiluminance to tuning at low-luminance, as well as tuning at high-luminance to tuning at equiluminance. The Seung–Sompolinsky curve fits (Eq. 1) were used to determine the peak tuning angle at each luminance level. Multidimensional scaling To view the population representations of stimuli, we applied multidimensional scaling (MDS). Each stimulus has a high-dimensional neural representation, with each dimension corresponding to the raw mean firing rate of a single neuron in response to that stimulus. MDS attempts to find a k-dimensional embedding of this high-dimensional space that approximately preserves its structure, where k is specified as an input. Given a set of x stimuli S = {s1, … sx} and a function d measuring the pairwise dissimilarities between them, MDS uses an iterative algorithm to find an embedding f: S→Rk for some fixed k such that the distances |fsi-fsj| have approximately the same rank ordering as the dissimilarities d(si, sj). We selected Sammon’s error (Sammon, 1969) as the error function to minimize, and defined the dissimilarity between two stimuli si and sj in terms of their neural representations si^ and sj^: dsi,sj=1-ρ(si^, sj^), where ρ is Pearson’s correlation coefficient. Stimuli are considered dissimilar to the extent that the population responses they evoke are uncorrelated. We performed two separate MDS analyses on the glob and interglob population datasets. The aim of the first analysis was to view how the populations represent the full stimulus set (45 hues at each of three luminances). We z-score normalized the raw response of each neuron to the stimulus set, and applied MDS for a range of k-values. In the second analysis, we examined separately the neural representations of each stimulus luminance class (stimuli at lower luminance than the background, equiluminant with the background, and at higher luminance than the background) in order to view the neural representations of hue within each luminance class. Here, we z-score normalized the responses of each neuron within each luminance set (45 hues each) before performing MDS for k = 2. Representational similarity analysis We used representational similarity analysis (RSA; Kriegeskorte et al., 2008) to compare the neural representations of hue by the two populations to CIELUV color space hue angle. We divided the data into three luminance sets (45 hues each) as above, and z-score normalized the response of each neuron within a set. For each set, we created a neural representational dissimilarity matrix (RDM) in which, as in the MDS analyses, an entry contained the neural correlation dissimilarity between two stimuli. We created a second RDM containing the CIELUV hue angle distance between stimuli. For each luminance class, we then determined the Pearson's r value between the neural and CIELUV RDMs, yielding a measure of similarity between neural and color space representations of hue (Cichy et al., 2014). To test whether these correlations were significant, we performed a Student’s t test, two-tailed on the Fisher’s z-transforms of r. To determine whether there was a significant difference between pairs of correlation coefficients for the glob versus interglob populations, we performed paired t tests, two tailed to z-transforms of the r values. Model LGN populations To test whether or not the MDS and RSA results for the glob and interglob cells were different than those we could expect earlier in the visual system, we performed these analyses on a population of model parvocellular LGN cells. The population of 300 model LGN cells was generated using the narrowness-uniformity model (see Fig. 5). The model LGN cells were linearly tuned (tuning width of 180) and biased for the cardinal axes (28% uniform, 72% cardinal), matching values found in previous studies of the LGN (Derrington et al., 1984; De Valois et al., 2000). The model does not account for luminance, so we analyzed the responses to the luminance levels separately. Limitations of the model also precluded an MDS analysis of the full dataset for comparison with the MDS analysis performed on the glob and interglob cells (see Fig. 10A,B ). Additionally, because the model LGN cells had identical responses to a small number of stimuli, these stimuli were removed from analysis. MDS was run on responses to 41 low-luminance stimuli, 43 equiluminant stimuli, and 45 high-luminance stimuli. Decoding hue and luminance information from the population responses Guided by the results of the MDS analyses and RSAs, which suggested that both populations represent both hue and luminance information, we sought to determine (in the population responses) whether hue information was preserved across changes in luminance, and whether luminance information was preserved across changes in hue. We determined accuracies in classifying (1) hue information invariant to changes in luminance, and (2) luminance information invariant to changes in hue, using a linear support vector machine (SVM; MATLAB_R2013b svmtrain, least-squares method), which attempts to find a hyperplane with maximum margins separating the high-dimensional points (neural responses to stimuli) belonging to two training classes. For these analyses, we used responses to a hue-matched subset of the stimuli (described in the next paragraph), and, prior to applying decoding, z-score normalized the responses of each neuron within the hue-matched stimulus subset. For our decoding analyses, it was important that we used sets of stimuli that had closely corresponding hue angles across the luminance levels. We identified a triplet of hue-matched stimuli as one in which the three hues at the three luminance levels differed no more than 3° in CIELUV angle. Seventeen hue-matched triplets (51 stimuli) were identified (Fig. 1, colors identified by a small dot). In order to test the decoding of hue invariant to changes in luminance, we tested whether a linear SVM could generalize hue information from two luminance classes to a third luminance class. For each possible pair of the 17 hues in the hue-matched stimulus subset, we trained the classifier to distinguish between the pair of hues, h1 and h2, given the population response to these hues at two luminances, and tested whether the classifier assigned the correct labels to h1 and h2, given the population response to the test luminance (e.g., high luminance). We performed decoding for three generalization problems, generalizing to low-luminance, equiluminant, and high-luminance stimuli. We present the mean pairwise decoding accuracies for each of these problems separately. In order to test the decoding of luminance invariant to changes in hue, we tested whether the classifier could generalize luminance information across changes in hue. We performed three decoding problems: we trained and tested a classifier’s ability to distinguish among the (1) low-luminance and equiluminant stimuli, (2) equiluminant and high-luminance stimuli, and (3) low-luminance and high-luminance stimuli. For each classification problem, we trained the classifier on 15 of 17 stimulus hues, and tested on the held-out pair. We performed a decoding run for each possible pair of test hues. We present the mean decoding accuracy across runs for each of these classification problems (see Fig. 12). In both decoding analyses, to account for a mismatch in population size between the globs (N = 300) and interglobs (N = 181), we performed subsampling. For each of 200 subsampling runs, we drew a random subset of 181 glob neurons, and performed the full decoding procedure to obtain decoding accuracies for this population. To test whether decoding accuracies for each classification problem were significantly above chance, we performed a permutation test in which we repeated the full decoding procedure 200 times with randomly shuffled labels, yielding a null distribution of decoding accuracies. We counted as significant decoding accuracies lying above all null points, which enabled us to bound p at 0.005 (1/200). For the glob population, this procedure was repeated for each subsampling run (all subsampling runs achieved accuracy at p < 0.005). To test whether decoding accuracies were significantly different between the glob and interglob populations, for each classification problem, we obtained a p value by taking the average of 200 p values derived by comparing the results for the interglob population and the results for one subsampling run of the glob population using a two-tailed McNemar’s exact test. Comparing narrowness of tuning with estimates obtained in prior work In order to compare the narrowness of the tuning of cells in the glob and interglob populations with values reported by Namima et al. (2014) for cells in V4, PIT, and anterior inferior temporal cortex (AIT) populations, we followed a similar method to compute a narrowness measure. We considered the responses of each cell to the low-luminance and high-luminance stimuli in the hue-matched subset, to match the high- and low-luminance stimulus sets used in the analysis by Namima et al. (2014). For each cell, for each luminance set, we calculated a selectivity index = 1 − (minimum response)/(maximum response), where responses are the raw mean firing rates of the cell in response to the luminance set. If a cell had a selectivity index >0.6 for either luminance set (all cells in the glob and interglob populations met this criterion), we next computed the sparseness index of that cell (Rolls and Tovee, 1995; Vinje and Gallant, 2000) using Equation 2: sparseness index= [1-(∑i=1nrin)2(∑i=1nri2n)]/(1-1n) where ri is the response of the cell to the ith stimulus and n is the number of stimuli in the luminance set. A sparseness index of 1 indicates the cell is sharply selective, whereas a low score indicates the cell responds similarly to all stimuli. Consistent with Namima et al. (2014), we then labeled cells with sparseness indices >0.3 for either set as “sharply selective,” and those with sparseness indices ≤0.3 to both sets as “broadly selective.” We compared the proportions of sharply tuned and broadly tuned cells in the two populations to values reported by Namima et al (2014, their Fig. 5). Proportion of warm- and cool-tuned cells In order to determine whether the glob or interglob populations were biased for warm colors over cool colors, as suggested in the population-tuning distribution for the glob cells (see Fig. 3), we performed a permutation test. We defined warm hues as the CIELUV hue angles lying between Munsell RP and Y (pink, red, orange, and yellow), and cool hues as those between Munsell G and PM (green, cyan, blue, and violet). On each of 2000 permutations, we randomly assigned a population of cells (n = 300 for globs, n = 181 for interglobs) 1 of the 45 hue angles per luminance level. We then calculated the ratio of cells tuned to warm colors to those tuned to cool colors. We then used this distribution to calculate p values. For 2000 permuations each, using a random selection of 90% of the real glob or interglob populations, we calculated the warm-tuned-to-cool-tuned cell ratio. We then determined a p value by counting the number of permutation populations with a higher warm-tuned-to-cool-tuned cell ratio than the bootstrap population. The 95% CIs on the p value were also calculated from the bootstrap distribution using the percentile method. This analysis was performed separately for both populations (globs and interglobs), and each luminance level. Results The neurophysiological data were obtained from the original study by Conway et al. (2007). The stimuli used to characterize the color responses were defined using the CIELUV chromaticity diagram (Komatsu et al., 1992; Conway et al., 2007), which organizes colors in a more or less perceptually uniform manner (Fig. 1). The full stimulus set comprised 45 colors at three luminance levels (all the colors within a set were equiluminant with each other; one set was higher luminance than the adapting background; one set was lower luminance; and one set was equiluminant with the background). Of these 45 colors, 21 hues were selected to be at relatively equal angles in CIELUV space (Fig. 1, points outlined in black); responses to these stimuli were used to quantify the neural color tuning. In other analyses, we analyzed responses to 17 hue-matched (within 3°) stimulus triplets (Fig. 1, dotted colors); responses to these hue-matched stimuli were used in various decoding analyses, which are described below. Figure 2 shows tuning curves obtained for representative cells in the globs (Fig. 2A ) and interglobs (Fig. 2B ). Glob cells typically showed narrower color-tuning curves than interglob cells, and color preferences that were retained across different luminance levels. The median tuning width (see Eq. 1) among glob cells was narrower than that of interglobs for all luminance levels (glob cells: 104°, 84°, and 91°, respectively, for high-luminance, equiluminant, and low-luminance; interglob cells: 120°, 125°, and 121°). For the combined population of glob and interglob cells across all luminance levels, the median narrowness was 100. The median goodness of fit of the tuning-curve fits for glob cells and interglob cells were 0.86 and 0.62. The correlation of peak tuning preferences across luminance levels was higher for globs (0.90 equiluminant/low-luminance; 0.91 high-luminance/equiluminant) than interglobs (0.80 equiluminant/low-luminance; 0.84 high-luminance/equiluminant). See Figures 8 and 9 for more in-depth population analyses of narrowness differences between the two populations and tuning differences across luminance levels. Figure 2. Color tuning of representative glob and interglob cells. A, B, Responses to the 21 subsampled hues, smoothed with a boxcar kernel of 1 hue, at each luminance level from six glob cells (A) and six interglob cells (B). Responses were measured using a bar stimulus optimized for each cell. Points show spike/second (Hz) firing rate response to each stimulus’s LUV hue angle. Lines show the Seung–Sompolinsky curve fit (light gray, high-luminance set; dark gray, equiluminant set; black, low-luminance set). Numbers denote the narrowness (tuning width in CIELUV degrees) for each example cell. Figure 3 shows the number of cells tuned to each of the 45 colors at each luminance level, for globs and interglobs. For each cell, the color tuning was defined as the color corresponding to the stimulus that elicited the peak firing rate. We applied no smoothing of the firing rates across colors, to avoid inflating the extent to which our model predictions, described below (see Fig. 5), correspond to the neural data. (As expected, comparisons of smoothed neural data and the model predictions produced higher R 2 values; data not shown.) The insets in Figure 3 show the distribution of color-tuned cells, binned into categories defined by the 21 equally spaced hues. The plots show lines demarking the cardinal axes (L-M, and S), along with the Munsell principal and intermediate hues (Munsell, 1907). The red, yellow, green, and blue lines are the Munsell coordinates corresponding to the unique hues, although we acknowledge that there is considerable variability within the population with regard to the precise location of the unique hues. Both the glob and interglob populations included neurons tuned to every stimulus we used. Both glob and interglob cells showed a clear over-representation of some colors (red, green, blue, and possibly purple). The pattern of over-representation was evident at all three luminance levels tested, but more clearly consistent across luminance levels for the glob population. Figure 3. Peak color tuning distributions for the glob and interglob populations. The peak color tuning of each cell at each luminance level was defined as the angle of the stimulus to which each cell maximally fired. The number of cells tuned to each hue was counted. The size of the marker at each hue denotes the number of cells tuned to each hue. This analysis was performed separately for the glob (top row) and interglob populations (bottom row), for all 45 stimuli (main plots) and then binned into the 21 evenly spaced hues (insets). In all panels, black axes labeled L-M and S denote the cardinal axes, and colored markers show the nine Munsell primary and intermediate hue coordinates. To test whether the populations are best described by a distribution biased toward the cardinal directions, by the unique hues, or, alternatively, by a uniform distribution, we compared the neurophysiological results to model predictions. The simulated neural populations corresponded to a range of theoretical possibilities by varying the uniformity with which the population represents color space (100% uniform vs a bias toward either the cardinal colors or the unique hues), and the linearity of the color tuning of the cells (100% linear, meaning a tuning curve in the shape of a sine wave, vs highly nonlinear, meaning a tuning curve narrow than a sine wave). Cells with maximally nonlinear tuning would respond to a single color only. Importantly, the model incorporates information about the saturation of the stimuli to account for the differences in saturation among the stimuli used to test the neural responses. Plotted in CIELUV space (Fig. 1) or cone-opponent MB-DKL space (Fig. 1A , inset), the stimuli track a triangle defined by the three monitor primaries. Consider a cell that shows maximal responses to color X when tested with a stimulus set comprising colors of equal saturation (a circular set). “X” would constitute the true color tuning of the neuron. But now consider the response of the same cell when tested with the triangular set: the cell could show peak firing to an adjacent color of higher saturation, especially if the tuning of the cell is relatively broad (Mollon, 2009). We sought to test whether the biases in the population reflect those predicted for a population of linear cells that represent color space uniformly, against the two other predictions: a population of linear cells that over-represents the cardinal directions, and a population of nonlinear cells that over-represents the unique hues. We determined how the different model populations would respond to the 21 evenly spaced colors of the equiluminant triangular stimulus set (Fig. 4). The lower left panel of Figure 4A shows the predicted color-tuning distribution for a population of linear neurons that as a population are biased for the cardinal axes, simulating the known properties of LGN cells. The population shows two dominant peaks, which correspond to the colors of highest cone contrast and highest saturation (within the stimulus set, these colors are closest to the red and blue primaries of the monitor). The lower right panel in Figure 4A shows the distribution for a population of linear neurons that sample color space uniformly; again, the population distribution is weighted toward the color of maximum saturation (blue), although there are subsidiary peaks for intermediate colors. The top panels in Figure 4A show the corresponding distributions for nonlinearly tuned neurons. The predicted population responses are less distorted by differences in saturation: the top left panel in in Figure 4A shows four peaks, corresponding to the poles of the cardinal axes; the top right panel in Figure 4A shows peaks sampling the entire gamut. Figure 4B shows predictions for model populations that are biased toward the unique hues; the model predictions for populations of uniformly tuned neurons shown in Figure 4B are different from the predictions for populations of uniformly tuned neurons shown in Figure 4A: a metric of saturation defined in psychophysical color space (CIELUV) was used in Figure 4B , while a metric of saturation defined in cone-opponent coordinates was used in Figure 4A (see Materials and Methods). Figure 4. Peak color tuning distributions predicted by model populations. Like recorded cells, the peak color tuning of each model cell at each luminance level was defined as the stimulus to which each cell had the largest response. Model cell responses were the product of the Seung–Somplinsky tuning curve and the saturation of each stimulus. A, B, Conventions are as in Figure 3, for model populations bias for the cardinal axes (A) and the unique hues (B). Shown here for the 21 evenly spaced equiluminant hues. Cardinal distribution is 22% uniform and 78% cardinal (our closest approximation of the LGN); unique hue distribution is 100% unique hue biased; uniform distributions are 100% uniform. Comparison of recorded and model populations Figure 5 quantifies the relationship between the measured population responses and those predicted by one set of models. The tuning curves for each cell in the model were drawn either from 10°-wide distributions centered on the cardinal hues or from bins that uniformly sampled color space. To define the population of cells, the proportion of cells with peaks drawn from the cardinal categories was systematically varied, from all cells drawn from drawn from the cardinal categories (a value of 0 on the x-axis) to all cells drawn from the uniform sampling (a value of 1 on the x-axis). All cells in each iteration were assigned the same narrowness, which varied from broader than linear to highly nonlinear (tuning curves were modeled using Eq. 1; we varied the tuning-width parameter to achieve nonlinear tuning). The blackness in the heat map in Figure 5A corresponds to the R 2 value comparing the model prediction to the neural glob data obtained using the equiluminant stimulus set. The model that captures the pattern in the LGN (linear cells tuned to the cardinal axes) yielded a relatively low R 2 value (Fig. 5A , outlined black box in the lower left). We can therefore rule out the first hypothesis that the population is best described by linearly tuned cells that are biased for the cardinal directions. Could the neural data be well described instead by a population of linear neurons regardless of the population distribution? No: the R 2 values obtained for models capturing nonlinear tuning tended to be greater (the rows in the heat map get darker from bottom to top up until the best-fitting width of 132). Across the columns in the heat map, the R 2 values are highest for the models that capture an almost completely uniform representation of color space (far right columns in the heat map). Of these, the optimal model is the one that consists of cells with nonlinear color tuning (R 2 = 0.52; narrowness = 132; Fig. 5B ). That the best model fit consists of nonlinear neurons is reflected in the tuning curves of the recorded data (Fig. 2). Figure 5C shows the tuning curve and raw data for an example glob cell with median narrowness (black line), along with a model tuning curve with a tuning width of 132 (and for comparison, the linear curve, tuning width of 180; see Fig. 8). These results undermine the hypothesis that the population comprises a uniform distribution of linear neurons that manifests as a population with biases for those colors of highest saturation. Figure 5. Model of color-tuning properties. We generated model populations varying in peak tuning distributions and narrowness. Each simulated cell was assigned a peak tuning angle that was drawn either from a distribution centered on the cardinal angles, or a uniform distribution. A, The proportion of simulated peaks drawn from each distribution was systematically varied from 100% cardinal to 100% uniform (bubble plots). A, All cells in each iteration were assigned the same narrowness, which varied from more broad than linear (tuning width of 360) to highly nonlinear (tuning width of 84; tuning curves) and then scaled by the saturation of each stimulus. Each square of the heat map represents one combination of narrowness and proportion uniform peaks. Each model population was compared with the recorded glob or interglob populations. The darker the square of the heat map, the better overall correlation between the model population and the real population. The best-matching simulated population is indicated with a black/white box. The heat map shown here compares the responses of the model population to the responses of the glob population to the equiluminant stimuli. Thick black boxes correspond to the example peak distributions and tuning curves to the left and below the heat map; the black box in the lower left shows the predicted best-matching population for the LGN. B, C, The best-matching model population was similar to the glob cells in both tuning peaks (B; conventions are as in Fig. 3) and tuning width (C; black line average glob tuning curve, black points show representative glob cell, thick gray line shows tuning function of the best model fit, thin gray line shows tuning function of a model cell with linear narrowness.). The best model was similar regardless of the luminance of the stimulus set used to collect the neural data (Fig. 6, left panels): for each stimulus set, the best model was one comprised of nonlinear neurons that represent color space in a uniform fashion. Glob cells were best matched by cells with nonlinear tuning for the high-luminance and equiluminant stimulus sets (tuning width of 120 for the data obtained using the high-luminance sets, 132 for the equiluminant set, and 168 for the low-luminance set). Interglob cells were best fit by a population of broadly tuned neurons with a uniform tuning distribution (Fig. 6, right panels). Figure 6. Comparison with model populations: cardinal. Heat maps in the convention described in Figure 5 show the median R 2 value across the 1000 iterations of the model and 181 recorded-model cell pairs for the glob (left) and interglob (right) populations. Darker shading denotes better fits. White/black outline denotes the parameter combination with the highest median R 2 value for that luminance level. Top row, High-luminance stimulus set; middle row, equiluminant stimulus set; bottom row, low-luminance stimulus set. Black box to lower left denotes the best expected match for LGN cells. The preceding results show that the population of PIT neurons cannot be well described by a model comprising neurons, such as those in the LGN, that are biased for the cardinal directions. But can the population be better explained by a model biased for the unique hues? Figure 7 shows the heat maps comparing the neural data with simulated populations that vary in the extent to which the neurons are biased for the unique hues. The models that account for the most variance among the glob cells are those that comprise nonlinear neurons with a uniform representation of color space. The interglobs are best described by the model populations consisting of broadly tuned neurons. Neurons with broader tuning functions are more sensitive to variation in stimulus saturation, which would lead to peaks in the population distribution for colors of highest saturation (Fig. 4, bottom rows, peaks at the apices of the triangles). The conclusions from Figure 7 are no different if saturation is defined in cone-opponent space (data not shown). Figure 7. Comparison with model populations: unique hues. Heat maps with conventions as in Figure 5 for glob (left) and interglob (right) populations. Peak distributions ranged from 100% unique hues to 100% uniform. Top row, High-luminance stimulus set; middle row, equiluminant stimulus set; bottom row, low-luminance stimulus set. Narrowness in the recorded populations In order to relate the parameters predicted by the model to those found in the recorded cells, we quantified the tuning properties of the glob and interglob cells. First, we fit each cell with the Seung–Somplinsky curve fit (Eq. 1; Figs. 2, 5C, curve fits; we used the same equation, multiplied by the saturation of each stimulus, to simulate cell responses in the model). Glob and interglob cells showed a variety of tuning curve widths, but, in general, glob cells showed narrower tuning compared with interglob cells (Fig. 8). The majority of poor fits for the interglobs were cells with very broad (often almost flat) color-tuning curves (Fig. 2B , bottom left, response to equiluminant stimuli for the example cell). Figure 8. Tuning width distributions. The responses of each cell were fit with a Seung–Sompolinsky curve equation (see Materials and Methods). As the narrowness of the tuning of the cell increases, the tuning width parameter decreases. Glob cells (A-C) had lower tuning width values on average than interglob cells (D-F) for all luminance levels. Top row, Histogram for high-luminance stimuli; middle row, equiluminant stimuli; bottom row, low-luminance stimuli. Dotted line marks the median tuning width. Shading indicates the goodness of fit of the sine exponent curve: white, all cells; light gray, cells with curve fits of R 2 > 0.4; dark gray, R 2 > 0.6; black, R 2 > 0.8. Effect of stimulus luminance on recorded cell responses The tuning curves of example cells (Fig. 2) suggest that the neurons in both the glob and interglob populations carry luminance information: in both sets of cells, the peak response amplitude and hue tuning of the neuron varied somewhat with changes in luminance level. We performed an ideal observer ROC analysis to quantify the luminance sensitivity of glob and interglob cells. Although implementations can differ (see Materials and Methods), the ROC analysis can be described as follows. For each neuron, we computed histograms showing the number of stimuli that elicited a given firing rate: one histogram for responses to the stimuli of the high-luminance set, one for responses to the equiluminant set, and one for responses to the low-luminance set. The ROC analysis compares the extent to which the histograms overlap: the less the overlap, the more likely the cell could distinguish the luminance difference between the stimulus sets. We performed two comparisons: equiluminant versus low-luminance sets; and high-luminance versus equiluminant sets (Fig. 9A ). Given a criterion firing rate (selected from the range of firing rates produced by the neuron), we calculated the proportion of stimuli to which the first histogram exceeded the criterion, and the proportion of stimuli to which the second histogram exceeded the criterion. The calculation was performed for criterion values spanning the response range of the neuron; and we plotted the proportion of stimuli on which the second histogram exceeded the criterion as a function of the proportion of stimuli on which the first histogram exceeded the criterion. From these plots, we computed an AUC. Data points for histograms that perfectly overlap would fall along the diagonal (AUC = 0.5) and indicate that the neuron could not distinguish the luminance of the two sets of stimuli. AUC values <0.5 would indicate that the neuron could distinguish the luminance, and, moreover, that the neuron preferred the luminance associated with the first stimulus set. AUC values >0.5 would indicate that the neuron preferred the luminance associated with the second stimulus set. Figure 9. Effect of stimulus luminance on tuning. A, Luminance sensitivity of cells in the glob (left) and interglob (right) populations assessed by ROC analysis. Interglob cells exhibit a preference for low-luminance or high-luminance stimuli over equiluminant stimuli. The histograms show AUCs for the ROC for each cell, for two luminance discrimination problems. An AUC <0.5 indicates a neuron preferred the first luminance in the pair; an AUC >0.5 indicates a preference for the second. Black bars indicate cells that significantly preferred one of the two luminance categories (p < 0.05, permutation test). The interglob population had significantly more cells whose AUCs were significantly different than chance than did the glob population, for both luminance discriminations (equiluminant/low-luminance: p = 0.00002, Fisher’s exact test, two-tailed; high-luminance/equiluminant: p = 2.6 × 10−13). B, Luminance-invariant color tuning assessment in globs (left) and interglobs (right). For each cell, the peak angle within the low-luminance set is plotted against the peak angle within the equiluminant set (first and third columns), and the peak angle within the equiluminant set is plotted against the peak angle within the high-luminance set (second and fourth columns). The marker color indicates the hue to which the cell maximally responded for the x-axis stimulus set. The marker size increases with the cell's tuning narrowness. Because color space is circular, stimuli are rotated to appear at the y-axis point as close to the x = y line as possible. For example, if a cell had its peak at 10 CIELUV degrees for the equiluminant set and 355° for the low-luminance set, it would be plotted at (10, −5) rather than (10, 355). C, Peak shifting histograms. Each cell was placed into one of eight bins based on the angle to which it maximally responded within a luminance level. Lighter bars show the peak tuning of the cells in each bin at the lighter of the two luminance values shown (equiluminant for the first and third columns, high-luminance for the second and fourth columns). Darker bars show the peak tuning of the same cells, at the darker of the two luminance values shown (low-luminance for the first and third columns, equiluminant for the second and fourth columns). The solid line shows the mean tuning angle for the higher of the two luminance values shown, and the dashed line shows the mean tuning angle of the cells at the darker of the two luminance values shown. Arrowhead designates the peak rod response hue angle. Figure 9A shows AUC measurements for the population of neurons, for the three sets of comparisons. Black bars indicate neurons whose AUCs were significantly different from chance on permutation testa. On average, compared with the interglob population, the glob population contains fewer neurons that were capable of discriminating luminance (there are fewer black bars in the two leftmost plots). The glob and interglob populations contained different proportions of cells that could distinguish luminance, for each discrimination problem (equiluminant vs low-luminance: p = 0.00002b; and high-luminance vs equiluminant: p = 2.6 × 10−13.b The glob cells that could discriminate luminance were equally distributed into ones preferring high-luminance, equiluminant, or low-luminance. By contrast, the majority of interglob cells showed a preference for the high-luminance or the low-luminance stimuli, but rarely the equiluminant stimuli: there is a greater proportion of cells with AUCs >0.5 for the low-luminance versus equiluminant comparison (Fig. 9A , third column), and <0.5 for the equiluminant versus high-luminance comparison (Fig. 9A , rightmost column). These results support the idea that the luminance response among the interglob cells does not contribute to the neural representation of color, whereas the luminance response among the glob cells does contribute to the neural representation of color. We also tested the tolerance to luminance modulation of hue tuning in the globs and interglobs. Figure 9B , leftmost column, shows the peak tuning angle (Eq. 1) of each glob cell for the equiluminant stimulus set plotted against the peak tuning angle of the same cell for the low-luminance stimulus set. Figure 9B shows this analysis for equiluminant versus low-luminance globs (leftmost column), high-luminance versus equiluminant globs (second column), equiluminant versus low-luminance interglobs (third column), and high-luminance versus equiluminant interglobs (rightmost column). The glob cells showed higher peak angle correlation across luminance levels than the interglobs, for both the equiluminant versus low-luminance comparison (glob, r = 0.90; interglob, r = 0.80 equiluminant vs low-luminance, p = 1.27 × 10−155)c and high-luminance versus equiluminant comparison (glob, r = 0.91; interglob, r = 0.84; high-luminance vs equiluminant, p = 4.70 × 10−130)c. To determine whether or not peak shifting varied as a function of hue preference, we divided the glob and interglob cells into categories based on peak tuning preference. We defined eight bins of equal angle sizes (bin edges 0:45:360) and sorted each cell into a category on the basis of its peak color preference using the equiluminant stimulus set (Fig. 9C , left) or the color preference obtained using the high-luminance stimulus set (Fig. 9C , right). We then compared the color preferences obtained using the low-luminance (Fig. 9C , left) and equiluminant (Fig. 9C , right) stimulus sets for each category. This analysis was repeated for both glob and interglob populations, and for the equiluminant versus low-luminance and high-luminance versus equiluminant comparisons. In the high-luminance versus equiluminant comparison, cells were sorted based on their high-luminance peak angle. For the globs, only the bin containing orange-yellow cells (45:90) showed a significant peak shift according to a Mann–Whitney–Wilcoxon U test. The shift for this bin was significant for both the equiluminant versus low-luminance comparison (p = 0.009168, see Table 1 for a complete list of p values for hue categories)d, and high-luminance versus equiluminant comparison (p = 0.02754)d. For the interglobs, both the orange-yellow and yellow-green bins showed significant peak shifting, only for the high-luminance versus equiluminant conditions (orange-yellow, p = 0.009795; yellow-green, p = 0.02961)d. None of the eight hue bins showed significant hue shifting for the interglobs between equiluminant and low-luminance stimuli, despite the large differences in equiluminant and low-luminance mean peak hue angles (Fig. 9C , solid and dashed lines). This effect is likely due to the lower correlation in peak angle across luminance levels (Fig. 9B ). Interglobs show large differences in peak hue tuning between luminance levels, but the shifts are inconsistent across cells: while most glob cells tuned to equiluminant yellow shift tuning such that they are tuned to low-luminance orange (Fig. 9C , leftmost column), some interglob cells tuned to equiluminant purple are tuned to low-luminance red, while others prefer low-luminance green and orange (Fig. 9C , third column). We did not find evidence of rod intrusion in this analysis: there was no systematic shift in peak tuning toward the peak rod sensitivity (Fig. 9C , black arrowhead) at lower luminance levels. Table 1: Statistics Data structure Type of test Power [CIs] aFigure 9A N/APermutation test does not make assumptions about distribution of the data Type of test: permutation test p = 0.030 [0.0269–0.0334] b Figure 9A Description: test of whether there is a significant difference in the proportion of significant AUC cells in the glob vs interglob population (where significant AUC cells are those cells with AUCs significantly different than 0.5, or chance) for a given discrimination problem (e.g., low luminance/equiluminant)Type of test: Fisher’s exact test of proportions, two-tailedMethods used to compute 95% CIs: For each of 200 bootstrap samples of the cells (N = 300 for the globs, N = 181 for the interglobs), we determined the proportions of cells with significant AUCs and computed the test statistic described above. From the distribution of p values, we computed 95% CIs by the percentile method Equiluminant/Low-luminance: p = 0.00002 [95% CIs: 3.95 × 10−10 to 0.006]High-luminance/Equiluminant: p = 2.6 × 10−13 [1.03 × 10−21 to 1.75 × 10−8] c Figure 9B Normally distributed after Fisher’s z transform Description: Test of whether the Pearson’s r values quantifying the correlation between peak tuning angles at two luminance levels is significantly different for the glob and interglob populations. We performed 200 bootstraps to define a distribution of r values for each luminance population combinationType of test: Student’s t test (two-tailed) applied to the z transform of a bootstrapped distribution of r values.Methods used to compute 95% CIs: We performed 200 bootstraps 1000 times in order to get 1000 p values for each comparison. From the distribution of p values, we computed 95% CIs by the percentile method. Significance of glob cell vs interglob cell luminance peak correlationEquilum vs Low Lum: p = 1.27 × 10−155 [6.32 × 10−168 to 1.50 × 10−144]High Lum vs Equilum: p = 4.70 × 10−130 [3.67 × 10−141 to 2.92 × 10−118] d Figure 9C N/AMann–Whitney U rank-sum test does not assume a normal distribution Description: test for peak shifting between luminance levels for eight evenly sized color categoriesType of test: Mann–Whitney U rank-sum test (MATLAB ranksum)Methods used to compute 95% CIs: We performed the rank-sum test on 2000 bootstraps containing 90% of the cells in each population in order to get a distribution of 1000 p values for each comparison. From the distribution of p values, we computed 95% CIs by the percentile method GlobEquilum vs Low Lum (in same order as in panel C, top to bottom): p = 0.1267 [0.1193–0.1342];0.4726 [0.4596–0.4857];0.3052 [0.2923–0.3182];0.2256 [0.2143–0.2369];0.2731 [0.261–0.2853];0.3197 [0.307–0.3324];0.009168 [0.007482–0.01085];0.3883 [0.3748–0.4018];High Lum vs Equilum: p =0.592 [0.5808–0.6031];0.4484 [0.4354–0.4614];0.1211 [0.1125–0.1296];0.4631 [0.4495–0.4766];0.4176 [0.4041–0.4311];0.5271 [0.5149–0.5394];0.02754 [0.02441–0.03067];0.4146 [0.4013–0.428]InterglobsEquilum vs Low Lum (in same order as panel C, bottom to top): p = 0.3541 [0.3409–0.3673];0.3396 [0.3259–0.3533];0.07011 [0.06317–0.07704];0.1924 [0.1805–0.2044];0.4253 [0.4117–0.439];0.312 [0.2988–0.3252];0.4446 [0.4312–0.458];0.152 [0.1417–0.1623];High Lum vs Equilum: p =0.4467 [0.4335–0.4599];0.1771, [0.166–0.1882];0.07582, [0.069–0.08263];0.4654, [0.4517–0.4791];0.1828 [0.172–0.1936];0.009795 [0.007742–0.01185];0.02961, [0.02585–0.03336];0.457 [0.4438–0.4702] e Figure 11A Normally distributed after Fisher’s z transform Description: Test of whether a Pearson’s r quantifying the correlation between a CIELUV RDM and a neural RDM is significantly different from zeroType of test: Student’s t test (two-tailed) applied to the z transform of r Methods used to compute 95% CIs: For each of 200 bootstrap samples of the cells, we created an RDM, computed the correlation between this bootstrap RDM and the CIELUV RDM, and computed the test statistic described above. From the distribution of p values, we computed 95% CIs Significance of Glob and CIELUV RDM Correlation High Lum set: p = 4.31 × 10−296 [95% CIs: 2.11 × 10−316, 6.63 × 10−259]Equilum set: p = 1.62 × 10−271 [2.82 × 10−293 to 3.76 × 10−236]Low Lum set: p = 3.71 × 10−321 [0 to 2.44 × 10−276]Significance of Interglob and CIELUV RDM CorrelationHigh Lum set: p = 3.33 × 10−157 [1.08 × 10−171 to 1.40 × 10−114]Equilum set: p = 5.16 × 10−166 [1.70 × 10−174 to 6.68 × 10−115]Low Lum set: p = 4.80 × 10−93 [2.03 × 10−100 to 3.31 × 10−65]Significance of Model LGN and CIELUV RDM CorrelationAll values are highly significant, with p < 2.23 × 10−308 (upper bound on 95% CI 0, equivalent to 2.23 × 10−308 in MATLAB 2015b) f Figure 11A Normally distributed after Fisher’s z-transform Description: Test of significance of difference between glob and interglob Pearson’s r correlation coefficientsType of test: Paired t test (two-tailed) applied to z-transforms of r’sMethods used to compute 95% CIs: As described in d above, we computed correlations between bootstrap RDMs and the CIELUV RDM, but here, subsequently performed the test statistic described above. High Lum set: p = 2.05 × 10−8 [95% CIs, 6.15 × 10−5 to 4.00 × 10−15]Equilum set: p = 1.70 × 10−5 [2.20 × 10−3 to 6.17 × 10−11]Low Lum set: p < 2.0 × 10−16 [2.22 × 10−16, < 2.0 × 10−16] g Figure 11B Classification accuracies; permutation test does not make assumptions about distribution of the data Description: Test of whether classification accuracy is significantly above chance (50%)Type of test: Permutation testMethods used to compute 95% CIs: For each of 200 bootstrap samples, we determined a classification accuracy given the bootstrap sample, and performed a permutation test to obtain a p value. Since our null distribution contained 200 points, p was bound at 0.005, permitting calculation of only an upper 95% confidence bound.Because the glob and interglob populations were of different sizes, we subsampled the glob population to N = 181 for each of 200 subsample runs, and consider the ps for all subsamples Hue decodingGlobsGeneralizing to High Lum: p < 0.005. [No null points lay above the observed decoding accuracy for any subsample of any bootstrap sample.]Generalizing to Equilum: p < 0.005. [No null points lay above the observed decoding accuracy for any subsample of any bootstrap sample.]Generalizing to Low Lum: p < 0.005. [No null points lay above the observed decoding accuracy for any subsample of any bootstrap sample.]InterglobsGeneralizing to High Lum: p < 0.005. [No null points lay above the observed decoding accuracy for any bootstrap sample.]Generalizing to equiluminant: p < 0.005 [upper 95% confidence bound = 0.08]Generalizing to Low Lum: p < 0.005 [upper 95% confidence bound = 0.315]Luminance decodingGlobsLow Lum/High Lum: p < 0.005. [No null points lay above the observed decoding accuracy for any bootstrap sample.]Low Lum/Equilum: p < 0.005 [No null points lay above the observed decoding accuracy for any bootstrap sample.]Equilum/High Lum: p < 0.005. [No null points lay above the observed decoding accuracy for any bootstrap sample.]InterglobsLow Lum/High Lum: p < 0.005. [No null points lay above the observed decoding accuracy for any bootstrap sample.]Low Lum/Equilum: p < 0.005. [No null points lay above the observed decoding accuracy for any bootstrap sample.]Equilum/High Lum: p < 0.005. [No null points lay above the observed decoding accuracy for any bootstrap sample.] h Figure 11B Binomial distribution Description: Significance of difference between classification accuracies for the glob versus interglob populationsType of test: McNemar’s exact test, two-tailed on paired binomial data, with α = 0.05Methods used to compute 95% CIs: For each of 200 bootstrap samples, we obtained a p value by taking the average of 200 p values derived by comparing the results for the interglob population and the results for one subsampling run of the glob population using McNemar’s extact test, two-tailed (for paired binomial data). Hue decodingGeneralizing to High Lum: p = 5.29 × 10−7 [95% CIs: 1.05 × 10−10 to 0.003]Generalizing to Equilum: p = 6.27 × 10−12 [1.88 × 10−15 to 1.12 × 10−6]Generalizing to Low Lum: p = 1.69 × 10−5 [1.88 × 10−15 to 0.006]Luminance decodingLow Lum/High Lum: p = 0.213 [95% CIs: 0.005–0.939]Low Lum/Equilum: p = 0.463 [0.003–0.570]Equiluminant/High-luminance: p = 0.074 [2.6 × 10−6, 0.375] iDiscussion N/A, permutation test does not assume a normal distribution Description: significance of difference between proportion of warm tuned and cool tuned cellsType of test: permutation test. For 2000 permutations, each cell was randomly assigned one of the 45 stimulus angles to be tuned to. A null distribution of warm-tuned-to-cool-tuned cell ratios was calculated from this permutation.Methods used to compute 95% CIs: We ran 2000 bootstraps using 90% of each population. For each bootstrap, we calculated a p value as the proportion of permuted populations with a higher warm-tuned-to-cool-tuned cell ratio than the warm-tuned-to-cool-tuned cell ratio of the bootstrap population. The 95% CIs are calculated from the bootstrap populations. Because we had 2000 bootstrap permutations, the lowest bound of the p value possible is p < 5 × 10−4 GlobsHigh Lum: p = 0.001 [5 × 10−4 to 0.14]Equi Lum: 5 × 10−4 [5 × 10−4 to 0.001]Low Lum: p < 5 × 10−4 5 × 10−4 to 5 × 10−4] (globs had higher warm-tuned-to-cool-tuned cell ratio than all permutations on all bootstraps)InterglobsHigh Lum: p = 0.0035 [5 × 10−4 to 0.15]Equi Lum: p = .51 [0.06–0.95]Low Lum: p = p < 5 × 10−4 5 × 10−4 to 5 × 10−4] Equilum, Equiluminant; High Lum, high-luminance; Low Lum, low-luminance; N/A, not applicable. These results are consistent with the idea that the interglob population is sensitive to luminance contrast independent of the hue of the stimulus, whereas the glob population is sensitive to a combination of luminance contrast and hue. Such a combined sensitivity is predicted for neurons that represent a psychophysical color space in which the same hue at different luminance levels can be distinguished as having a different color. For example, orange and brown have the same hue but differ in luminance contrast. Certain glob cells would be capable of signaling both the hue and the luminance that distinguish orange from all other hues, and orange from brown. Quantitative comparison of population coding to color space The tuning curves of example cells (Fig. 2) also suggest that both glob and interglob cells carry information about hue. In our next set of analyses, we considered how the globs and interglobs represent color information at the population level. We used MDS to first obtain a picture of neural color space, allowing our results to be compared with those of Namima et al. (2014). We then deployed RSA to quantitatively compare the neural representations of colors to the spatial relationships between those colors in CIELUV color space. To our knowledge, our analysis is the first in the single-unit literature that quantitatively tests the fit between neural representations and color space. Each stimulus has a high-dimensional neural representation in which each dimension corresponds to the mean firing of one cell in the population of recorded neurons. For the glob population (N = 300), the representation of a given stimulus is 300-dimensional; for the interglob population (N = 181), it is 181-dimensional. From the high-dimensional neural representation, we can read out information about how similar the two stimuli are: we deem stimuli to be similar to the extent that their neural response vectors are correlated. MDS allows us to reduce (or “embed”) the high-dimensional neural representation to a lower-dimensional representation that we can plot, and that still captures (most of) the similarity structure of the original representation. We can compute an error measure (we used Sammon’s stress) to determine how much the similarity relationships between pairs of points in the high-dimensional representation are uncaptured by a best-fitting lower-dimensional representation (Fig. 10, insets, top panels). While higher numbers of dimensions always produce better dissimilarity approximations, this comes at the cost of increased model complexity; for our embeddings, stress drops for two dimensions, and then decreases gradually thereafter. Figure 10 shows plots of the two-dimensional MDS embeddings for the two populations of neurons in V4/PIT, as well as a population of model LGN cells. The color of each data point corresponds to the color of the stimulus, and the spatial relationship of the points reflects the extent to which the neural responses to the stimuli were related: data points are plotted closest together when the neural responses to the two stimuli were most similar (see Fig. 12 for the three-dimensional MDS embeddings). Figure 10. Glob, interglob, and LGN neural color spaces as calculated by MDS. Both the glob and interglob populations represent hue information, and these representations resemble the structure of perceptual CIELUV color space. A–K, MDS applied to the glob (A, B, C, D), interglob (E, F, G, H), and model LGN (I, J, K) responses to stimuli yields a picture of neural color space. Each stimulus has a high-dimensional neural representation, where each dimension corresponds to the mean firing rate of a single cell. MDS produces a low-dimensional embedding that seeks to preserve the distances between stimuli in the original, high-dimensional space. We use 1 − ρ (Pearson’s correlation coefficient) as our distance metric: stimuli are distant to the extent that the patterns of activity they evoke are uncorrelated. Stimuli are plotted by their color in coordinates determined by the new, two-dimensional embedding. A, B, Glob and interglob MDS embeddings of the full stimulus set (45 hues at three luminance each). Insets show Sammon’s stress as a function of embedding dimensions used. Glob (B–D), interglob (F–H), and model LGN (I–K) MDS embeddings of stimulus hues for each luminance level (Sammon’s stress for globs: low-luminance = 0.05; equiluminant = 0.06; high-luminance = 0.05; for interglobs: low-luminance = 0.12; equiluminant = 0.10; high-luminance = 0.10; for model LGN cells: low-luminance = 0.10; equiluminant = 0.10; high-luminance = 0.10). For the glob population, the arrangement of the stimuli clearly reflects CIELUV color space: points of the same hue irrespective of luminance level are plotted next to each other, and the progression of the points forms a circle that proceeds according the color wheel: following clockwise, blue is next to cyan, which is next to green, followed by yellow (brown), orange, red, and, closing the circle, purple. The pattern obtained for the interglob cells also bears some similarity to perceptual color space, but is not as clearly organized (e.g., cyan dots are intermingled with blue and purple dots). The MDS analyses in the top panels of Figure 10 were performed using the neural responses to all the stimuli at all luminance levels. How do the populations represent hue information when luminance is removed as a variable? In other words, to what extent is the neural representation of color preserved across luminance levels? To answer this question, we performed MDS separately on responses to stimuli in each luminance set. For the globs, the representation of the stimuli largely reflects perceptual color space for all luminance levels tested (Fig. 10, B-D). For the interglobs, the representation of the stimuli corresponds less clearly to perceptual color space, especially when assessed using the equiluminant and low-luminance sets (Fig. 10, F-H). The MDS analysis suggests that behavioral judgments of the similarity between colors closely match the similarities between the neural responses to these colors by the glob population, and, to a lesser extent, by the interglob population. For comparison, we performed the same MDS analysis on a simulated population of parvocellular LGN neurons (see Materials and Methods). The MDS representation recovers the sequence of colors found in perceptual color space (Fig. 10, right column) but is distorted toward the stimuli that had the highest saturation, as expected from a population of neurons with linear tuning (Fig. 4). If the sensitivity to color is reflective of a causal role in color perception, we hypothesized that the relative response of a population to pairs of stimuli should reflect the similarity of the colors of the stimuli defined by a uniform perceptual color space. To test this idea quantitatively, we performed RSA, in which we looked at the extent to which the representational dissimilarities between stimuli according to neural response are correlated with the dissimilarities according to perceptual CIELUV color space hue angle. Given any two stimuli, we can measure their dissimilarity both by the angular distance between their hues in CIELUV space, and by the correlation distance between their neural response vectors. Performing these two computations for each pair of stimuli, we obtain two 45 × 45 dissimilarity matrices, where 45 is the number of hues. To assess whether the two measures induce similar representations, we calculate the Pearson’s correlation coefficient between the two dissimilarity matrices. A high correlation indicates that the neural responses agree well with the color space. To focus on hue, we split the stimuli into the three luminance sets (low-luminance, equiluminant, and high-luminance), and for each luminance set, we performed RSA on each of the glob and interglob population responses. Given our MDS results (Fig. 10), we predicted that the glob representations of hue at each luminance would strongly correlate with CIELUV hue space for all luminance sets, but that for the interglobs this correlation would be high only for the high-luminance set. Further, we expected there to be a significant difference in the correlation values for the two populations. As shown in Figure 11A , we found that the representations by the glob population of the low-luminance, equiluminant, and high-luminance sets were indeed significantly correlated with the angular distances between stimuli (high-luminance set: Pearson’s r = 0.70, p = 4.31 × 10−296, two-tailed t teste; equiluminant set: r = 0.68, p = 1.62 × 10−271; low-luminance set: r = 0.72, p = 3.71 × 10−321). The representations of the interglob population were also significantly correlated for each set (high-luminance set: r = 0.55, p = 3.33 × 10−157; equiluminant set: r = 0.56, p = 5.16 × 10−166; low-luminance set: r = 0.43, p = 4.80 × 10−93). But the glob representations were more similar to CIELUV than were the interglob representations (high-luminance set: p = 2.05 × 10−8, paired two-tailed t testf; equiluminant set: p = 1.70 × 10−5; low-luminance set: p < 2.0 × 10−16), showing that hue information in the glob population more closely resembles perceptual color space. We performed the same analysis on a population of model LGN cells and, similarly, found high correlations between their representation of the stimulus sets and the angular distances between stimuli: (high-luminance set: r = 0.83; equiluminant set: r = 0.96; low-luminance set: r = 0.89; all correlations were significant, p < 2.23 × 10−308e, two-tailed t test). Figure 11. Neural color space and CIELUV hue RSA (A) and hue decoding (B). A, Neural distances between stimuli are significantly correlated with the hue angle distances between these stimuli in perceptual CIELUV color space. Both glob and interglob population representations are significantly correlated with CIELUV for the high-luminance, equiluminant, and low-luminance stimulus sets (p < 0.0001, t test, two-tailed). For all sets, the glob representations are significantly more correlated with CIELUV than are the interglob representations (high-luminance set: p = 2.05 × 10−8, t test two-tailed, Fisher r-to-z; equiluminant set: p = 1.70 × 10−5; low-luminance set: p < 2.0 × 10−16). B, Representations of the same hue at different luminance levels are sufficiently similar that hue information can be read out by a linear SVM invariant to changes in luminance. Mean pairwise hue classification accuracies for three generalization problems. A classifier was trained to distinguish between hues given two stimulus sets (e.g., low-luminance and equiluminant), and tested on the held-out stimulus set (e.g., high-luminance). The test luminance appears below each set of bars. Classification accuracy was significantly above chance for all generalization problems in both populations (p < 0.005, permutation test), though significantly higher for the glob than interglob population (all generalization problems: p < 0.0001, McNemar’s exact test, two-tailed). Error bars indicate the SD of the mean across 200 subsampling runs. Decoding of hue information across changes in luminance Contemporary color-ordering systems treat hue and luminance as separable parameters. To examine whether the extraction of hue and luminance from colored stimuli could be supported by the neural data, we used invariant decoding: we used a pattern classifier to predict the hue of a stimulus from its neural representation. During training, we presented the classifier with labeled examples of the neural response vectors for two hues at two luminances each (e.g., low-luminance green, equiluminant green, low-luminance blue, equiluminant blue). We then tested the classifier by asking it to predict the hues of two new neural response vectors, where our test cases were the same two hues seen during training, but at a new luminance (e.g., high-luminance green and high-luminance blue). To do well on this task, then, our classifier must generalize the hue information it learned during training to correctly classify these hues at a luminance it has never seen before. We obtain a comprehensive measure of classification accuracy by averaging the results for each pair of hues. High classification accuracy indicates that the population represents hue in such a way that it is sufficiently similar as to be recognizable across changes in luminance. When a classifier successfully “reads out” an experimental variable given a pattern of activity, this suggests that the information may be accessible to upstream neurons. Similar analyses have been used to look at how neural populations represent object identity invariant to changes in position and scale (Hung et al., 2005; Rust and diCarlo, 2010) and to recover color information from fMRI data (Brouwer and Heeger, 2009). Given that our RSA and MDS results showed that the glob population captures hue information to a greater degree than does the interglob population, we predicted that classification accuracy would be higher for the glob cells. Figure 11B shows results for both populations for three classification problems, namely, problems in which we (1) train on low-luminance and equiluminant sets, and test on high-luminance; (2) train on low-luminance and high-luminance, and test on equiluminant; and (3) train on equiluminant and high-luminance, and test on low-luminance. Classification accuracy was significantly above chance for all three problems for both the glob and interglob populations (all p < 0.005g, permutation test). But classification accuracies were significantly higher for the glob than the interglob population (generalizing to high-luminance: p = 5.29 × 10−7, two-tailed McNemar’s exact testh; generalizing to equiluminant: p = 6.27 × 10−12; generalizing to low-luminance: p = 1.69 × 10−5). Analyses of luminance information present in the population responses In Figure 9, we showed that many of our individual cells can discriminate stimulus luminance. How does this ability manifest at the population level? As we did with hue, we can use MDS embeddings to visualize stimuli by luminance category. Figure 12 shows the three-dimensional MDS embeddings for the glob (Fig. 12A ; Sammon’s stress = 0.03) and interglob populations (Fig. 12B ; stress = 0.06). In both populations, the stimuli are well segregated by luminance. Striking is the fact that, in the globs, each luminance set forms a discrete ring, and each of these rings contains a full hue map, similar to three-dimensional renderings of CIELUV color space. Movies 1 and 2 show the rotations of the three-dimensional embedding for the glob population colored by stimulus color (1) and luminance category (2). Movies 3 and 4 show the rotations of the interglob embedding. Figure 12. Luminance clustering within neural color space (A, B) and luminance decoding (C). Both the glob and interglob populations segregate stimuli by luminance. A, B, Three-dimensional MDS embedding for the glob (A) and interglob (B) populations. Squares indicate low-luminance stimuli, circles indicate equiluminant stimuli, and triangles indicate high-luminance stimuli. C, Luminance information is accessible to readout by a linear classifier invariant to changes in hue for both populations, showing that these populations carry information about stimulus luminance as well as hue. Mean classification accuracies across decoding runs were high (>89%) for distinguishing between all luminance class pairs, for both populations (all p < 0.005, permutation test). Error bars indicate the SD of the mean across 200 subsampling runs. Movie 1. Three-dimensional MDS stimulus embedding for the glob population, colored by stimulus color. Rotations reveal hue-organized rings. 10.1523/ENEURO.0039-16.2016.video.1 Movie 2. Three-dimensional MDS stimulus embedding for the glob population, colored by luminance category. Squares indicate low-luminance stimuli, circles indicate equiluminant stimuli, and triangles indicate high-luminance stimuli. Rotations reveal three luminance rings. 10.1523/ENEURO.0039-16.2016.video.2 Movie 3. Three-dimensional MDS stimulus embedding for the interglob population, colored by stimulus color. 10.1523/ENEURO.0039-16.2016.video.3 Movie 4. Three-dimensional MDS stimulus embedding for the interglob population, colored by luminance category. Squares indicate low-luminance stimuli, circles indicate equiluminant stimuli, and triangles indicate high-luminance stimuli. Rotations reveal that the population segregates stimuli by luminance. 10.1523/ENEURO.0039-16.2016.video.4 Given the luminance category segregation apparent in the MDS plots, we expected that hue-invariant luminance decoding should be possible for both populations. Proceeding analogously to our hue-decoding analysis, we trained a classifier to distinguish between each pair of stimuli of constant hue and different luminance. For each pair, we trained the classifier by showing it luminance-labeled examples of 15 of 17 hues. Next, we tested its ability to predict the luminance of the two hues that had been left out. We found that the classifier was able to read out luminance information with high accuracy invariant to changes in hue for both neural populations (>89%, p < 0.005, permutation testg; Fig. 12C ). Classification accuracies were not significantly different between the glob and interglob populations. Discussion The organization of colors can safely count as one of the longest-lived problems in science (Kuehni and Schwartz, 2008). Newton was perhaps the first to systematically arrange colors in a circle, making meaningful use of geometry; his insight suggested that color organization was determined not solely by physical parameters but also by the way signals are processed in the brain. Contemporary color-ordering schemes adopt three dimensions: hue (e.g., “red,” “purple,” “green”), saturation (“red” vs “pink”), and brightness (or value). Unknown are the neural rules that determine the geometric relationships within and between these dimensions. One tradition proposes that color is organized around six unique hues (Hering, 1905; Hurvich, 1981), which were initially thought to reflect color tuning in the LGN (De Valois et al, 1966). The psychological importance of these colors is not unquestioned (Saunders and van Brakel, 1997; Witzel et al., 2015; Wool et al., 2015). And careful analysis shows that LGN color tuning does not underlie the unique hues (Webster et al., 2000); the neural basis for the unique hues remains unknown. Where Newton’s color-ordering system was launched by physics, and Hering’s color-opponent scheme began with psychology, our approach starts with the structural organization of perception and the representation of color in the brain. Our goal is to determine what brain areas (and cells) represent color, and to interrogate the neural responses to reverse engineer the rules that govern color space geometry. We have not yet reached this goal, but here we present evidence showing that a population of neurons (glob cells) in PIT/V4 not only contains a representation of color space that bears remarkable similarity to uniform perceptual color space, but also possesses nonlinear (narrow) color tuning, two features that suggest an important role in color perception. Color-tuned neurons have been found earlier than V4/PIT in the visual-processing hierarchy. V1 has received considerable attention (Gegenfurtner and Kiper, 2003; Solomon and Lennie, 2007; Conway, 2009; Conway et al., 2010; Shapley and Hawken, 2011), although linear systems analysis has not uncovered a meaningful correspondence between V1 color-tuning properties and color space (Lennie et al., 1990). An analysis of just the population of cone-opponent V1 cells—cells most likely involved in color processing—uncovered an over-representation of the colors associated with daylight (Conway, 2001; Lafer-Sousa et al., 2012), possibly providing a Bayesian prior used to resolve stimulus ambiguity (Lafer-Sousa et al., 2015). Other work, applying a nonlinear analysis, suggests that V1 cells inherit the chromatic-tuning biases of the LGN (Horwitz and Hass, 2012).Together, this research suggests that V1 constitutes an intermediate step in the computation of color. Consistent with this hypothesis, imaging in humans shows that the color space derived from the covariation across V1 voxels in response to different colors does not correspond to perceptual color space; instead, higher-order areas (hV4, V-O1) possess a representation that more closely corresponds to perception (Brouwer and Heeger, 2009). Cells carrying color information are found in V2 (Burkhalter and Van Essen, 1986; Hubel and Livingstone, 1987; Moutoussis and Zeki, 2002); as a population, V2 cells show a bias for colors of daylight (Kiper et al., 1997), similar to V1. One study suggests that V2 contains maps that match color space (Xiao et al., 2003; Lim et al., 2009), but quantitative tests of this correspondence have not been performed. Neurons in V3 either appear indistinguishable in their color-tuning properties from V2 neurons (Gegenfurtner et al., 1997) or they carry considerably less color information (Baizer, 1982). A substantial transformation of color signals takes place in or before area V4/PIT (Zeki, 1980; Desimone et al., 1985). The importance of V4 in color processing was initially challenged (Schein et al., 1982), but the controversy was resolved with experiments combining fMRI and fMRI-guided microelectrode recording, which showed that V4 contains color-biased subdomains, “globs,” that are separated by interglob regions showing lower color bias (Conway and Tsao, 2006; Conway et al., 2007). These findings have been confirmed using optical imaging (Tanigawa et al., 2010; Li et al, 2014). Prior work suggests that glob cells have narrow color tuning (Conway et al., 2007), but the narrowness has not been quantified until now. Here we show that glob cells are probably much narrower in their tuning than LGN cells (and most V1 cells). The stimuli used presently, which are typical of many neurophysiological studies of color, preclude a definitive conclusion because they confound saturation and hue. Nonetheless, the likely narrow tuning, coupled with the representation of color space encompassed by the population of glob cells, suggests a neural basis for higher-order psychophysical chromatic mechanisms (Webster and Mollon, 1991; Hansen and Gegenfurtner, 2007; Stoughton et al., 2012). A preliminary analysis suggested that the glob cell population is biased toward the unique hues (Stoughton and Conway, 2008). The stimuli used to obtain these data were similar to stimuli used in other studies (Komatsu et al., 1992): they consist of the most vivid colors permitted by the monitor. These colors lie on a triangle in CIE space. Cone contrast varies among stimuli within this set, and is highest for the red and blue apices. It has been suggested that the color-tuning biases observed in the population of glob cells would be observed in the LGN, if LGN responses were assessed with the triangular stimulus set (Mollon, 2009). Alternatively the color-tuning distribution may reflect the responses of a population that has color-tuned neurons uniformly representing color space, in which each neuron is sensitive to saturation (Conway and Stoughton, 2009). We tested these alternative predictions by comparing model simulations to recorded data. Simulations that assume linearly tuned neurons did not match the glob cell data: the simulations only yielded peaks to colors located at the apices of the triangle, unlike the measured neural population that also had peaks at intermediate colors. These findings rule out the first possibility: glob cells do not appear to have color tuning as found in the LGN. Among nonlinear models, does the best-fitting model show a bias toward the cardinal directions? It seems not. The nonlinear models biased to the cardinal directions showed too few responses to green compared with the glob cell data. Instead, the optimal model was a uniform representation. Among glob cells, those tuned to purple showed the narrowest color tuning, a feature that may reflect the peculiar properties of S cone-opponent neurons in V1 (Cottaris and De Valois, 1998; Conway and Livingstone, 2006; Horwitz and Hass, 2012), LGN (Tailby et al., 2008b), and retina (Dacey and Lee, 1994). These cells probably account for the band corresponding to models with a tuning width of 96 (Fig. 6). The nonlinear models biased for the unique hues did not fare any better at each level of nonlinearity than the models of populations biased to the cardinal colors: the best model was still close to a uniform distribution. These results suggest that the variation in saturation among the stimuli is the cause of the bias toward red, green, and blue that was reported previously. The best-matching models in all the comparisons have R 2 values hovering around 0.5, leaving plenty of variation unaccounted for. Some of this variability may be attributed to uneven sampling of neurons. It is also possible that the color space sampling, every ∼17°, was too sparse; we suspect this is not the case because we did not find systematic changes of narrowness as a function of color tuning when using stimuli that more finely sample color space. The stimuli were all relatively low luminance, raising the possibility of rod intrusion (Stiles and Burch, 1959). The analysis of color responses at different luminance levels does not indicate a bias predicted by the rod peak, although the impact of rod intrusion could be complex (Shapiro et al., 1996; Buck, 1997). The complications of interpreting the neurophysiological data described here underscore the critical importance of the careful choice of stimuli, and draw attention to the overlooked gaps in our understanding of what constitutes seemingly elementary properties of color (hue, saturation, and brightness) and their interaction. In particular, we note that there remains no consensus on the spatial organization of color space: CIELUV, the perceptual space used here, is a standard, but it has defects (Melgosa et al., 1994) Prior work has shown that glob cells are spatially clustered by color preference (Conway and Tsao, 2009). Conway and Tsao (2009) present the first microelectrode evidence for chromotopic maps anywhere in the visual system, but they perform no quantitative tests of the correspondence between the neural representation of color and the organization of perceptual color space. The data presented here fill this gap, and strongly suggest that the population of glob cells contains narrowly tuned neurons representing most directions in color space, not just the cardinal directions favored by the LGN. Moreover, using representational similarity analysis, we found that the representation of color space encoded by the glob, as compared with the interglob, population, shows a better correspondence to CIELUV color space; the glob cell population also shows a bias toward warm colors (reds, yellows). RSA of a model LGN population also showed similarity to CIELUV color space, which demonstrates that a correspondence with color space can occur in the absence of single cells narrowly tuned to each color in space. The RSAs on the two cortical populations suggest that between glob and interglob populations, it is the glob cells that perform the readout of the color space representation found in the LGN. This conclusion is supported by the narrow tuning found among glob cells and the decoding algorithms showing that stimulus color can be predicted from the color tuning of the most active neurons in the glob cell population (Zaidi et al., 2014). The responses to the luminance of the different sets of stimuli, by both glob and interglob populations, could be read out by a linear classifier. But a linear classifier decoded the hue invariant to luminance better for the glob than for the interglob population. Luminance classification accuracy (luminance invariant to hue) was high for both populations. One might have thought that neurons implicated in color perception (glob cells) would show color tuning that was entirely invariant to changes in luminance. But luminance is an important dimension of color, and can change a hue from one color into another: an increase in luminance contrast (brightness) converts brown into orange. The sensitivity to luminance contrast of the glob cells is consistent with prior observations (Conway et al., 2007; Namima et al., 2014), and may underlie the Bezold–Brucke hue shift (Stoughton and Conway, 2008). The interglob cells, meanwhile, typically did not retain their hue tuning across luminance levels. We quantified this result, and confirm that interglob cells tended to show clear preferences for stimuli that were either darker or brighter than the background, regardless of hue. Together, the results on the interglob cells suggest that this population is using chromatic information in the service of something besides color perception, such as the detection of color boundaries for object recognition. Such a computation would be useful in object segmentation and in defeating camouflage, where sensitivity, but not selectivity, for color is crucial; this property of interglob cells suggests that they are part of the network that includes the “complex equiluminance” cells found in V1 (Conway and Livingstone, 2006; Conway, 2014). Imaging experiments show that considerable cortical territory within and anterior to V4 is implicated in color (Tootell et al., 2004; Conway et al., 2007; Harada et al., 2009; Lafer-Sousa and Conway, 2013); there is a high degree of homology in the functional organization of the ventral visual pathway in humans and IT cortex in monkeys (Wade et al., 2008; Lafer-Sousa et al., 2016). Microelectrode recording and anatomical track-tracing experiments in monkeys confirm that extrastriate regions identified as color biased using fMRI typically are enriched for color-tuned neurons (Conway et al., 2007) and probably connected to each other (Banno et al., 2011). Namima et al. (2014) offer the first systematic comparison of the narrowness of color tuning at multiple stages spanning the temporal lobe, from V4 to AIT. They found that neurons in AIT showed more nonlinear tuning than neurons in V4/PIT, although they did not target recordings to color-biased domains identified independently with imaging. We wondered whether the relatively narrower tuning of neurons in AIT might be evident in V4/PIT, if one restricted the analysis to just globs. If so, then the high nonlinearity in AIT might not be generated cumulatively through serial stages along IT, as suggested by Namima et al. (2014), but rather created early in processing and inherited by AIT. Consistent with our hypothesis, glob cells had considerably narrower color tuning than interglob cells. Indeed, the narrowness of the glob cells was higher than the estimates for neurons in AIT (93% sharply selective glob cells, compared with ∼75% in AIT, as estimated by Namima et al., 2014). The results presented here suggest that the neural correlate of perceived color is computed in V4/PIT. What then is the functional role of the vast amount of color-biased cortical tissue within the rest of IT? At this point, we can only speculate. Chromatic information informs many behaviors ranging from attentional recruitment, to memory, to social cognition (Conway, 2016).These behaviors underscore an important advantage of color perception: the system is trainable and can forge associations not only between colors and objects, but also between colors and abstract concepts such as emotional states (red/anger) and words. IT is implicated in many aspects of high-level object vision, including recognition, categorization, memory, and attention. Koida and Komatsu (2007) have shown that the firing of color-tuned neurons within AIT is influenced by task demands, lending support to the idea that the job of more anterior regions of IT is not to compute color perception (a job taken care of by PIT), but rather to use this information to direct behavior. It seems likely, then, that the color-biased regions within AIT are involved in high-level behaviors that depend on or involve color. Hering (1964) made a powerful case that opponency is the basis for color vision. He took this observation further to argue that color is constructed by three sets of exclusive color pairs: red/green; blue/yellow; and black/white. The results from the study by Hering (1964) have been interpreted as supporting the theory that basic color categories are universal and derive from the hardwiring of color tuning in the nervous system (Berlin and Kay, 1969; Lindsey and Brown, 2009). But the universalist theory has received some hard blows challenging the methodology (Saunders and van Brakel, 1997). Psychophysical work has also challenged the theory: the unique hues are not more salient than other colors, as they should be if they are privileged (Wool et al., 2015). Moreover, there is considerable variability among people (Webster et al., 2002), and across language groups (Lindsey et al., 2015), in the location and boundaries of the unique hues within color space; and the specific colors that are assigned special status vary across cultures (Davidoff et al., 1999; Roberson et al., 2000), which suggests that the unique hues are not as special as widely assumed. Neurophysiological work strongly suggests that the building blocks of color processing in the retina and LGN depend on cone opponency (Conway, 2009); but this opponency does not account for the unique hues or basic color categories (Webster et al., 2000). Is it possible that the importance of the unique hues is learned, reflecting some special behavioral importance, and as such depends on processing in higher-order areas, rather than being innate and accounted for by activity early in the visual-processing hierarchy? The ultimate goal of the visual system is to transform an inherently ambiguous retinal stimulus into an unequivocal signal that can guide action. The retinal stimulus for color is typically ambiguous, which seems to be at odds with our experience of color: most observers are under a powerful illusion that color (including the importance of the unique hues) is tied in some direct way to the physical world (Conway, 2009, 2016)—the apparent unequivocal nature of our experience of color is one of the triumphs of the brain. Each of us has great conviction about the accuracy and validity of our color experience, even when there is no clear consensus among people considering the same stimulus (Lafer-Sousa et al., 2015). Our convictions could corrupt experimental design: in retrospect, it seems likely that the methods of the important World Color Survey (http://www1.icsi.berkeley.edu/wcs/) begged the existence of basic color terms (Saunders and van Brakel, 1997). Rather than reflecting a fundamental input to the color vision machinery, the unique hues more likely represent the computational product of this machinery, reflecting the integration of task demands, behaviorally relevant statistics, and language, computations that we hypothesize are implemented in IT and the frontal cortex (Lafer-Sousa and Conway, 2013; Romero et al., 2014). These arguments reformulate the question regarding the unique hues: do they have some special behavioral relevance that underlies their privileged status? Throughout this article, we have referred to “color-biased regions” rather than “color areas.” We draw this distinction because we are not yet in a position to conclude what computations these regions are performing. The term “color area” implies a specific and exclusive role in computing color. Given the extensive amount of cortical real estate in the temporal lobe that responds to chromatic stimulation, the wide range of operations that could benefit from sensitivity to color, and the likely sensitivity of color-biased regions to other stimulus attributes such as texture and material properties (Goda et al., 2014), it would seem premature (and wrong) to conclude that color-biased regions, including the globs of V4/PIT, are functionally discrete areas dedicated to processing only color. The analysis presented here provides a method for quantitatively testing the relationship between color behavior and neural activity that we hope to exploit in tracing the transformation of color signals through these regions to figure out what the regions are doing and how color tuning emerges. The close relationship between perceptual color space and the neural representation of color found among glob cells of PIT/V4 shows what information is available to subsequent stages in the putative visual-processing hierarchy, and will guide future work measuring neural activity with stimuli that densely sample uniform color space in an attempt to find a solution to the geometry of color space. Acknowledgments: We thank Doris Tsao, who was instrumental in collecting the neurophysiological data (see Conway et al., 2007); Michael Wiest for valuable discussions and comments on the manuscript; and Cleo Stoughton and Galina Gagin, who helped in the initial modeling experiments. Synthesis The decision was a result of the Reviewing Editor Tim Bussey and the peer reviewers coming together and discussing their recommendations until a consensus was reached. A fact-based synthesis statement explaining their decision and outlining what is needed to prepare a revision is listed below. The following reviewer(s) agreed to reveal their identity: Qasim Zaidi The reviewers and I felt the MS was very well written, with an impressive amount of analysis of colour responses of PIT cells, much of it systematically done. The paper carefully considers some important questions in color theory, and the results are likely to be of significant interest to the field. As pointed out by one reviewer in particular, all of the analyses rely on the measurements being tuning curves, and it was felt that this needed more justification. As this reviewer put it: "The original measurements were made to 45 colors along the triangle formed by the three phosphors of a color monitor. The authors try to approximate tuning curves by sampling test colors to be roughly equally spaced in the CIELUV diagram. There is not sufficient justification of this assumption in the paper. As far as I know, even industry people only consider CIELUV to be a locally uniform space, and even that is not based on any data more recent than MacAdam's color matches. It would be useful to see references to empirical evidence that CIELUV is a uniform space for color differences as big as the ones used in this study. Even with this assumption, the sampling is 22.5 deg around the color circle. That would be fine for cosine-tuned cells, but is unlikely to provide good estimates for narrowly tuned cells." At the same time this same reviewer felt that an argument could be made that the Conway data is the most complete existing characterization of the color responses of PIT cells, and without a huge amount of data collection, it is not possible to improve on CIELUV as a uniform space. The reviewers ask you to consider a number of additional points in your revision: 1. "The colors of one set were higher luminance (7.8 cd/m2) than the background; those of another set were photometrically equiluminant with the background (3.05 cd/m2); and those of the third set were lower luminance than the background (0.62 cd/m2)." A number of standard sources do claim 3cd/m2 as the lower limit of the photopic regime, but it has been repeatedly shown that even at low photopic levels, for fields covering more than the central fovea, there is rod intrusion in color matching, e.g. the failure of scalar invariance in the Stiles 10 degree CMFs. Hence for each of these luminance levels, the authors need to show that rod intrusion does not play a part in distorting responses across the color space. A simple way to test this is to assume ceteris paribus that there should be more rod intrusion at the lower levels. That would predict a systematic shift in the tuning curves from high to low levels. The nature of the shift could be derived from published calculations based on CRT phosphor spectra (e.g. Shapiro AG, Pokorny J, Smith VC. Cone-rod receptor spaces with illustrations that use CRT phosphor and light-emitting-diode spectra. JOSA A. 1996 Dec 1;13(12):2319-28.). If that shift does not occur, then there is less reason to be worried about rod intrusion. 2. Please explain how: "Using stimuli at different luminance levels provides one way of controlling for luminance artifacts arising from, for example, chromatic aberration or variability in macular pigmentation." 3. "To do so, we subsampled the data, extracting responses to 16 hues (at each luminance level) that were relatively evenly spaced in perceptually uniform CIELUV (C.I.E., 2004) angle". What is the evidence that CIELUV is a uniform color space? My understanding is that it is just a linear correction to CIELAB, which was based on MacAdams color-matching errors, not thresholds, and certainly did not address uniformity of even moderate size color differences. I am raising this point because it is very important in evaluating the results of this paper, since the interpretation of the recordings as tuning curves depends critically on whether the colors are equally spaced on the important metric. Also were a total of 16 hues used for all the cells? Isn't this too sparse a set for narrowly tuned neurons, essentially 22.5 deg apart in angle. 4. Even if the hues were equally spaced, their saturations need to be equated. "We defined saturation in DKL space". This would rely critically on the relative scaling of the axes. How was this done? Were any measurements made to show equal saturations for equal excursions from gray in different directions? Why use DKL for saturation and not CIELUV? 5. "The time window began with the visual latency, which was defined as >2.5 standard deviations above the background firing rate, and ended when the response rate fell below the background firing rate + 2.5 standard deviations." Are all responses increases in firing rate, and no responses decreases? 6. For fitting narrow tuning curves, there are much better alternatives than the DeValois equation, in which the exponent controls the slope near threshold as in this equation, but a divisor inside the sine function controls the width. 7. "We chose to curve fit responses to only the 16 evenly-spaced stimuli" Evenly spaced in what metric? 8. "To account for differences in saturation among the stimuli, the model cell's response to each stimulus was multiplied by that stimulus's saturation." To justify this adjustment, the authors would have to show that the contrast response curve is linear, and that is highly unlikely. 9. "To test whether the populations are best described by a distribution biased towards the cardinal directions, by the unique hues, or alternatively by a uniform distribution, we compared the neurophysiological results to model predictions." Doesn't this comparison require random sampling? How close to that assumption was the actual data collection? 10. "Figure 2 shows tuning curves obtained for representative cells, in the globs (panel A) and interglobs (panel B). Glob cells typically showed narrower color-tuning curves than interglob cells, and color preferences that were retained across different luminance levels." This results is both interesting and incomplete. Does it mean that both glob and inter-glob cells are color selective? It would be useful to have summary statistics about narrowness of tuning, and shifts of tuning, that quantify the assertions in this statement here rather than later in the paper. 11. "The best model was similar regardless of the luminance of the stimulus set used to collect the neural data (Figure 6, left panels): for each stimulus set, the best model was one comprised of non-linear neurons that represent color space in a uniform fashion. Interglob cells were similarly best fit by a model comprising a uniform representation of color space (Figure 6, right panels)." Could this be an artifact of the discrete sampling of colors used for making the tuning curves? 12. "The interglobs are best described by the model consisting of broadly tuned neurons that are biased for the unique hues." This needs some explanation. What are the implications of this result? 13. The analysis in Fig 9 is weak. The sampling is not uniform across the circle, and the inference is particularly susceptible to non-uniformity in the metric used to sample colors. Unless there is some significant implication of this result, it can be left out. 14. The analysis in Fig 10 would probably be better captioned as Luminance Contrast Sensitivity. The issue of sensitivity to dark colors, such as brown, navy and maroon is too interesting to be given such short-shrift. The histogram comparison is fine because it is invariant to changes in tuning curves with luminance contrast, but it does not quantify how color sensitivity changes with luminance contrast. Also, in the absence of a proper analysis, the comment that "These results are consistent with the idea that the interglob population is sensitive to luminance contrast independent of the hue of the stimulus" is difficult to reconcile with the tuning curves of inter-glob cells in Fig 2. Clearly a closer analysis is required of how color tuning changes or not with luminance contrast. 15. There are two problems with the similarity analysis. First, I would need to see some justification that extended distances in CIELUV represent similarities between widely separated colors for humans or macaques. Second, I am not sure how to interpret the stress criteria. Ramsay, Takane, Young and others have developed maximum likelihood procedures for MDS that give probability estimates of fits etc, and it would be preferable to use these methods. 16. Decoding section. Given that the authors have just shown that glob cells can be selective for darker or lighter colors, and use a histogram analysis that enables them to ignore changes in color tuning with luminance contrast, I am not sure why different cells would not do decoding at different contrasts, so why look for decoding invariance across luminance contrast. 17. Figures 11-13. Given that the color signal from photoreceptors is 3-D, the authors need to identify under what scenario one would expect anything different from the results in Figure 11-13? Perhaps the authors should repeat the MDS etc exercise on the responses of the three (or four) classes of LGN cells, and see if there is any difference. More minor comments: The table detailing statistical comparisons is very helpful. Note however that the last two rows in this table state that p-value are Pending due to computation time. Perhaps this is a typo, because the tet states that p values are &amp;lg;0.0001 while referring to these two rows. Further, when reference to this table cannot be made as simple superscripts (a,b,c,d,e,f,g,h) -- it is not clear where to look for the corresponding footnotes. You should add Table 1, row a, etc. The last analysis does raise some questions. For example, the difference in analyses underlying Figure 6 and Figure 7 was not clear. In the case of interglob regions, Figure 6 shows that more uniform pooling than Figure 7. Figure 6 does not show qualitative differences between pooling for globs and interglobs, but Figure 7 does. Mutimedia: We found the movies useful in seeing the 3D structure of the synthesized color space. However, would it perhaps be better, or possible, to provide a matlab/mathametica/python script that the readers could run to produce this figure. In this way, one can control the viewing angle and speed. ==== Refs References Baizer JS (1982 ) Receptive field properties of V3 neurons in monkey. Invest Ophthalmol Vis Sci 23 :87 –95 . 7085225 Banno T , Ichinohe N , Rockland KS , Komatsu H (2011 ) Reciprocal connectivity of identified color-processing modules in the monkey inferior temporal cortex. Cereb Cortex 21 :1295 –1310 . 10.1093/cercor/bhq211 21060111 Berlin B , Kay P (1969 ) Basic color terms: their universality and evolution . Berkeley, CA : University of California . Britten KH , Shadlen MN , Newsome WT , Moversushon JA (1992 ) The analysis of visual motion: a comparison of neuronal and psychophysical performance. J Neurosci 12 :4745 –4765 . 1464765 Brouwer GJ , Heeger DJ (2009 ) Decoding and reconstructing color from responses in human visual cortex. J Neurosci 29 :13992 –14003 . 10.1523/JNEUROSCI.3577-09.2009 19890009 Buck SL (1997 ) Influence of rod signals on hue perception: evidence from successive scotopic contrast. Vision Res 37 :1295 –1301 . 9205721 Burkhalter A , Van Essen DC (1986 ) Processing of color, form and disparity information in visual areas VP and V2 of ventral extrastriate cortex in the macaque monkey. J Neurosci 6 :2327 –2351 . 3746412 Cichy RM , Pantazis D , Oliva A (2014 ) Resolving human object recognition in space and time. Nat Neurosci 17 :455 –462 . 10.1038/nn.3635 24464044 Commission Internationale De Leclairage (2004 ) CIE 15: technical report: colorimetry , Ed 3. CFR 430 Subpart B, App. R, 4.1.1. Vienna, Austria: Commission Internationale De Leclairage. Conway BR (2001 ) Spatial structure of cone inputs to color cells in alert macaque primary visual cortex (V-1). J Neurosci 21 :2768 –2783 . 11306629 Conway BR (2009 ) Color vision, cones, and color-coding in the cortex. Neuroscientist 15 :274 –290 . 10.1177/1073858408331369 19436076 Conway BR (2014 ) Color signals through dorsal and ventral visual pathways. Vis Neurosci 31 :197 –209 . 10.1017/S0952523813000382 24103417 Conway BR (2016 ) Processing In: Experience: culture, cognition, and the common sense (Jones CA , Mather D , Uchill R , eds), pp 86 –109 . Cambridge MA : MIT . Conway BR , Livingstone MS (2006 ) Spatial and temporal properties of cone signals in alert macaque primary visual cortex. J Neurosci 26 :10826 –10846 . 10.1523/JNEUROSCI.2091-06.2006 17050721 Conway BR , Stoughton CM (2009 ) Towards a neural representation for unique hues. Curr Biol 19 :R442 –R443 . 10.1016/j.cub.2009.04.056 Conway BR , Tsao DY (2006 ) Color architecture in alert macaque cortex revealed by fMRI. Cereb Cortex 16 :1604 –1613 . 10.1093/cercor/bhj099 16400160 Conway BR , Tsao DY (2009 ) Color-tuned neurons are spatially clustered according to color preference within alert macaque posterior inferior temporal cortex. Proc Natl Acad Sci U S A 106 :18034 –18039 . 10.1073/pnas.0810943106 19805195 Conway BR , Moeller S , Tsao DY (2007 ) Specialized color modules in macaque extrastriate cortex. Neuron 56 :560 –573 . 10.1016/j.neuron.2007.10.008 17988638 Conway BR , Chatterjee S , Field GD , Horwitz GD , Johnson EN , Koida K , Mancuso K (2010 ) Advances in color science: from retina to behavior. J Neurosci 30 :14955 –14963 . 10.1523/JNEUROSCI.4348-10.2010 21068298 Cottaris NP , De Valois RL (1998 ) Temporal dynamics of chromatic tuning in macaque primary visual cortex. Nature 395 :896 –900 . 10.1038/27666 9804422 Dacey DM , Lee BB (1994 ) The “blue-on” opponent pathway in primate retina originates from a distinct bistratified ganglion cell type. Nature 367 :731 –735 . 10.1038/367731a0 8107868 Davidoff J , Davies I , Roberson D (1999 ) Colour categories in a stone-age tribe. Nature 398 :203 –204 . 10.1038/18335 10094043 De Valois RL , De Valois KK (1993 ) A multi-stage color model. Vision Res 33 :1053 –1065 . 10.1016/0042-6989(93)90240-W 8506645 De Valois RL , Abramov I , Jacobs GH (1966 ) Analysis of response patterns of LGN cells. J Opt Soc Am 56 :966 –977 . 4959282 De Valois RL , Cottaris NP , Elfar SD , Mahon LE , Wilson JA (2000 ) Some transformations of color information from lateral geniculate nucleus to striate cortex. Proc Natl Acad Sci U S A 97 :4997 –5002 . 10781111 Derrington AM , Krauskopf J , Lennie P (1984 ) Chromatic mechanisms in lateral geniculate nucleus of macaque. J Physiol 357 :241 –265 . 6512691 Desimone R , Schein SJ , Moran J , Ungerleider LG (1985 ) Contour, color and shape analysis beyond the striate cortex. Vision Res 25 :441 –452 . 4024463 Eskew RT Jr (2009 ) Higher order color mechanisms: a critical review. Vision Res 49 :2686 –2704 . 10.1016/j.visres.2009.07.005 [Mismatch]19616020 Fize D , Vanduffel W , Nelissen K , Denys K , Chef d'Hotel C , Faugeras O , Orban GA (2003 ) The retinotopic organization of primate dorsal V4 and surrounding areas: a functional magnetic resonance imaging study in awake monkeys. J Neurosci 23 :7395 –7406 . 12917375 Gagin G , Bohon KS , Butensky A , Gates MA , Hu JY , Lafer-Sousa R , Pulumo RL , Qu J , Stoughton CM , Swanbeck SN , Conway BR (2014 ) Color-detection thresholds in rhesus macaque monkeys and humans. J Vis 14 (8 ):12 1 –15 . 10.1167/14.8.12 25027164 Gegenfurtner KR , Kiper DC (2003 ) Color vision. Annu Rev Neurosci 26 :181 –206 . 10.1146/annurev.neuro.26.041002.131116 12574494 Gegenfurtner KR , Kiper DC , Levitt JB (1997 ) Functional properties of neurons in macaque area V3. J Neurophysiol 77 :1906 –1923 . 9114244 Goda N , Tachibana A , Okazawa G , Komatsu H (2014 ) Representation of the material properties of objects in the visual cortex of nonhuman primates. J Neurosci 34 :2660 –2673 . 10.1523/JNEUROSCI.2593-13.2014 24523555 Hansen T , Gegenfurtner KR (2006 ) Higher level chromatic mechanisms for image segmentation. J Vis 6 (3 ):239 –259 . 10.1167/6.3.5 16643093 Hansen T , Gegenfurtner KR (2007 ) Higher order color mechanisms for image segmentation In: Brain, vision and artificial intelligence 2007 LNCS/LNAI (Mele F , Ramella G , Santillo S , Ventriglia F , eds). Heidelberg, Germany : Springer-Verlag , 72 –83 . 10.1007/978-3-540-75555-5_8 Hansen T , Gegenfurtner KR (2013 ) Higher order color mechanisms: evidence from noise-masking experiments in cone contrast space. J Vis 13 (1 ):26 1 –21 . Harada T , Goda N , Ogawa T , Ito M , Toyoda H , Sadato N , Komatsu H (2009 ) Distribution of colour-selective activity in the monkey inferior temporal cortex revealed by functional magnetic resonance imaging. Eur J Neurosci 30 :1960 –1970 . 10.1111/j.1460-9568.2009.06995.x 19912328 Hering E (1905 ) Outline of a theory of the light sense (Hurvich L , Jameson D , translators). Cambridge, MA : Harvard UP . Hering E (1964 ) Outlines of a theory of the light sense (Hurvich , LM , Jameson D , translators). Cambridge, MA : Harvard UP . Horwitz GD , Hass CA (2012 ) Nonlinear analysis of macaque V1 color tuning reveals cardinal directions for cortical color processing. Nat Neurosci 15 :913 –919 . 10.1038/nn.3105 22581184 Hubel DH , Livingstone MS (1987 ) Segregation of form, color, and stereopsis in primate area 18. J Neurosci 7 :3378 –3415 . 2824714 Hung CP , Kreiman G , Poggio T , DiCarlo JJ (2005 ) Fast readout of object identity from macaque inferior temporal cortex. Science 310 :863 –866 . 10.1126/science.1117593 16272124 Hurvich LM (1981 ) Color vision . Sunderland, MA : Sinauer Associates . Hurvich LM , Jameson D (1974 ) Opponent processes as a model of neural organization. Am Psychol 29 :88 –102 . 4204851 Kiper DC , Fenstemaker SB , Gegenfurtner KR (1997 ) Chromatic properties of neurons in macaque area V2. Vis Neurosci 14 :1061 –1072 . 9447688 Koida K , Komatsu H (2007 ) Effects of task demands on the responses of color-selective neurons in the inferior temporal cortex. Nat Neurosci 10 :108 –116 . 10.1038/nn1823 17173044 Komatsu H , Ideura Y , Kaji S , Yamane S (1992 ) Color selectivity of neurons in the inferior temporal cortex of the awake macaque monkey. J Neurosci 12 :408 –424 . 1740688 Krauskopf J , Williams DR , Heeley DW (1982 ) Cardinal directions of color space. Vision Res 22 :1123 –1131 . 7147723 Kriegeskorte N , Mur M , Bandettini P (2008 ) Representational similarity analysis—connecting the branches of systems neuroscience. Front Syst Neurosci 2 :4 . 10.3389/neuro.06.004.2008 19104670 Kuehni R , Schwartz A (2008 ) Color ordered: a survey of color systems from antiquity to the present . New York : Oxford UP . Lafer-Sousa R , Conway BR (2013 ) Parallel, multi-stage processing of colors, faces and shapes in macaque inferior temporal cortex. Nat Neurosci 16 :1870 –1878 . 10.1038/nn.3555 24141314 Lafer-Sousa R , Liu YO , Lafer-Sousa L , Wiest MC , Conway BR (2012 ) Color tuning in alert macaque V1 assessed with fMRI and single-unit recording shows a bias toward daylight colors. J Opt Soc Am A Opt Image Sci Vis 29 :657 –670 . 10.1364/JOSAA.29.000657 22561924 Lafer-Sousa R , Hermann KL , Conway BR (2015 ) Striking individual differences in color perception uncovered by “the dress” photograph. Curr Biol 25 :R545 –R546 . 10.1016/j.cub.2015.04.053 25981795 Lafer-Sousa R , Conway BR , Kanwisher NG (2016 ) Color-biased regions of the ventral visual pathway lie between face-and place-selective regions in humans, as in macaques. J Neurosci 36 :1682 –1697 . 10.1523/JNEUROSCI.3164-15.2016 26843649 Lennie P , Krauskopf J , Sclar G (1990 ) Chromatic mechanisms in striate cortex of macaque. J Neurosci 10 :649 –669 . 2303866 Li M , Liu F , Juusola M , Tang S (2014 ) Perceptual color map in macaque visual area v4. J Neurosci 34 :202 –217 . 10.1523/JNEUROSCI.4549-12.2014 24381282 Lim H , Wang Y , Xiao Y , Hu M , Felleman DJ (2009 ) Organization of hue selectivity in macaque V2 thin stripes. J Neurophysiol 102 :2603 –2615 . 10.1152/jn.91255.2008 19571184 Lindsey DT , Brown AM (2009 ) World Color Survey color naming reveals universal motifs and their within-language diversity. Proc Natl Acad Sci U S A 106 :19785 –19790 . 10.1073/pnas.0910981106 19901327 Lindsey DT , Brown AM , Brainard DH , Apicella CL (2015 ) Hunter-gatherer color naming provides new insight into the evolution of color terms. Curr Biol 25 :2441 –2446 . 10.1016/j.cub.2015.08.006 26365254 Luo MR , Cui G , Li C (2006 ) Uniform colour spaces based on CIECAM02 colour appearance model. Color Res Appl 31 :320 –330 . 10.1002/col.20227 MacAdam DL (1990 ) Redetermination of colors for uniform scales. J Opt Soc Am A 7 :113 –115 . 2299442 MacLeod DI , Boynton RM (1979 ) Chromaticity diagram showing cone excitation by stimuli of equal luminance. J Opt Soc Am 69 :1183 –1186 . 490231 Mayo JP , Sommer MA (2013 ) Neuronal correlates of visual time perception at brief timescales. Proc Natl Acad Sci U S A 110 :1506 –1511 . 10.1073/pnas.1217177110 23297217 Melgosa M , Quesada JJ , Hita E (1994 ) Uniformity of some recent color metrics tested with an accurate color-difference tolerance dataset. Appl Opt 33 :8069 –8077 . 20963027 Mollon JD (2009 ) A neural basis for unique hues? Curr Biol 19 :R441 –R442 . 10.1016/j.cub.2009.05.008 19515347 Moroney N , Fairchild MD , Hunt RW , Li C , Luo MR , Newman T (2002 ) The CIECAM02 color appearance model . In: Color and Imaging Conference , pp 23 –27 . Springfield, VA : Society for Imaging Science and Technology . Moutoussis K , Zeki S (2002 ) Responses of spectrally selective cells in macaque area V2 to wavelengths and colors. J Neurophysiol 87 :2104 –2112 . 10.1152/jn.00248.2001 11929928 Munsell AH (1907 ) A Color Notation, a measured color system, based on the three qualities hue, value, and chroma , Ed 2. Boston : Geo H Ellis Co . Namima T , Yasuda M , Banno T , Okazawa G , Komatsu H (2014 ) Effects of luminance contrast on the color selectivity of neurons in the macaque area v4 and inferior temporal cortex. J Neurosci 34 :14934 –14947 . 10.1523/JNEUROSCI.2289-14.2014 25378160 Ripamonti C , Cheung V , Westland S (2013 ) Computational colour science using Matlab . New York : Wiley . Roberson D , Davies I , Davidoff J (2000 ) Color categories are not universal: replications and new evidence from a stone-age culture. J Exp Psychol Gen 129 :369 –398 . 10.1037/0096-3445.129.3.369 11006906 Rolls ET , Tovee MJ (1995 ) Sparseness of the neuronal representation of stimuli in the primate temporal visual cortex. J Neurophysiol 73 :713 –726 . 7760130 Romero MC , Bohon KS , Lafer-Sousa L , Conway BR (2014 ) Functional organization of colors, places and faces in alert macaque frontal cortex . Washington, DC : Society for Neuroscience . Rust NC , diCarlo JJ (2010 ) Selectivity and tolerance (“invariance”) both increase as visual information propagates from cortical area V4 to IT. J Neurosci 30 :12978 –12995 . 10.1523/JNEUROSCI.0179-10.2010 20881116 Sammon JW (1969 ) A nonlinear mapping for data structure analysis. IEEE Trans Comput C-18 :401 –409 . 10.1109/T-C.1969.222678 Saunders BA , van Brakel J (1997 ) Are there nontrivial constraints on colour categorization? Behav Brain Sci 20 :167 –179 . 10096997 Schein SJ , Marrocco RT , de Monasterio FM (1982 ) Is there a high concentration of color-selective cells in area V4 of monkey visual cortex? J Neurophysiol 47 :193 –213 . 7062096 Seung HS , Sompolinsky S (1993 ). Simple models for reading neuronal population codes. Proc. Natl. Acad. Sci. 90 :10749 –10753 . 10.1073/pnas.90.22.10749 8248166 Shapiro AG , Pokorny J , Smith VC (1996 ) Cone-rod receptor spaces with illustrations that use CRT phosphor and light-emitting-diode spectra. J Opt Soc Am A Opt Image Sci Vis 13 :2319 –2328 . 8972587 Shapley R , Hawken MJ (2011 ) Color in the cortex: single- and double-opponent cells. Vision Res 51 :701 –717 . 10.1016/j.visres.2011.02.012 21333672 Solomon SG , Lennie P (2007 ) The machinery of colour vision. Nat Rev Neurosci 8 :276 –286 . 10.1038/nrn2094 17375040 Stiles WS , Burch JM (1959 ) N.P.L. colour-matching investigation: final report (1958). Opt Acta (Lond) 6 :1 –26 . DOI: 10.1080/713826267 10.1080/713826267 Stoughton CM , Conway BR (2008 ) Neural basis for unique hues. Curr Biol 18 :R698 –R699 . 10.1016/j.cub.2008.06.018 18727902 Stoughton CM , Lafer-Sousa R , Gagin G , Conway BR (2012 ) Psychophysical chromatic mechanisms in macaque monkey. J Neurosci 32 :15216 –15226 . 10.1523/JNEUROSCI.2048-12.2012 23100442 Sun H , Smithson HE , Zaidi Q , Lee BB (2006 ) Specificity of cone inputs to macaque retinal ganglion cells. J Neurophysiol 95 :837 –849 . 10.1152/jn.00714.2005 16424455 Tailby C , Solomon SG , Dhruv NT , Lennie P (2008a ) Habituation reveals fundamental chromatic mechanisms in striate cortex of macaque. J Neurosci 28 :1131 –1139 . 10.1523/JNEUROSCI.4682-07.2008 18234891 Tailby C , Solomon SG , Lennie P (2008b ) Functional asymmetries in visual pathways carrying S-cone signals in macaque. J Neurosci 28 :4078 –4087 . 10.1523/JNEUROSCI.5338-07.2008 18400907 Tanigawa H , Lu HD , Roe AW (2010 ) Functional organization for color and orientation in macaque V4. Nat Neurosci 13 :1542 –1548 . 10.1038/nn.2676 21076422 Tootell RB , Nelissen K , Vanduffel W , Orban GA (2004 ) Search for color 'center(s)' in macaque visual cortex. Cereb Cortex 14 :353 –363 . 15028640 Vinje WE , Gallant JL (2000 ) Sparse coding and decorrelation in primary visual cortex during natural vision. Science 287 :1273 –1276 . 10678835 Wade A , Augath M , Logothetis N , Wandell B (2008 ) fMRI measurements of color in macaque and human. J Vis 8 (10 ):6 1 –19 . 10.1167/8.10.6 Webster MA , Mollon JD (1991 ) Changes in colour appearance following post-receptoral adaptation. Nature 349 :235 –238 . 10.1038/349235a0 1987475 Webster MA , Miyahara E , Malkoc G , Raker VE (2000 ) Variations in normal color vision. II. Unique hues. J Opt Soc Am a Opt Image Sci Vis 17 :1545 –1555 . 10975364 Webster MA , Webster SM , Bharadwaj S , Verma R , Jaikumar J , Madan G , Vaithilingham E (2002 ) Variations in normal color vision. III. Unique hues in Indian and United States observers. J Opt Soc Am a Opt Image Sci Vis 19 :1951 –1962 . 12365615 Witzel C , Cinotti F , O'Regan JK (2015 ) What determines the relationship between color naming, unique hues, and sensory singularities: illuminations, surfaces, or photoreceptors? J Vis 15 (8 ):19 1 –32 . 10.1167/15.8.19 26114682 Wool LE , Komban SJ , Kremkow J , Jansen M , Li X , Alonso JM , Zaidi Q (2015 ) Salience of unique hues and implications for color theory. J Vis 15 (2 ):10 1 –11 . 10.1167/15.2.10 25761328 Xiao Y , Wang Y , Felleman DJ (2003 ) A spatially organized representation of colour in macaque cortical area V2. Nature 421 :535 –539 . 10.1038/nature01372 12556893 Yasuda M , Banno T , Komatsu H (2010 ) Color selectivity of neurons in the posterior inferior temporal cortex of the macaque monkey. Cereb Cortex 20 :1630 –1646 . 10.1093/cercor/bhp227 19880593 Zaidi Q , Halevy D (1993 ) Visual mechanisms that signal the direction of color changes. Vision Res 33 :8 . Zaidi Q , Marshall J , Thoen H , Conway BR (2014 ) Evolution of neural computations: Mantis shrimp and human color decoding. Iperception 5 :492 –496 . 10.1068/i0662sas 26034560 Zeki S (1980 ) The representation of colours in the cerebral cortex. Nature 284 :412 –418 . 6767195 Zeki S , Marini L (1998 ) Three cortical stages of colour processing in the human brain. Brain 121 :1669 –1685 . 10.1093/brain/121.9.1669 9762956

PMC005xxxxxx/PMC5002983.txt

==== Front eNeuroeNeuroeneuroeneuroeNeuroeNeuro2373-2822Society for Neuroscience 10.1523/ENEURO.0074-16.2016eN-NWR-0074-162New ResearchDevelopmentHepatocyte Growth Factor Modulates MET Receptor Tyrosine Kinase and β-Catenin Functional Interactions to Enhance Synapse Formation MET Signaling in Synapse FormationXie Zhihui 1Eagleson Kathie L. 1http://orcid.org/0000-0002-9538-9017Wu Hsiao-Huei 1http://orcid.org/0000-0002-9717-1695Levitt Pat 121 Department of Pediatrics, The Saban Research Institute, Children’s Hospital Los Angeles, Keck School of Medicine, University of Southern California, Los Angeles, California 900272 Institute for the Developing Mind, The Saban Research Institute, Children’s Hospital Los Angeles, Keck School of Medicine, University of Southern California, Los Angeles, California 90027The authors declare no competing financial interests. Author contributions: Z.X., K.L.E., and P.L. designed research; Z.X. and H.-H.W. performed research; Z.X., K.L.E., and P.L. analyzed data; Z.X., K.L.E., H.-H.W., and P.L. wrote the paper. This research was supported by National Institute of Mental Health Grant MH-067842, and The Simms/Mann Chair in Developmental Neurogenetics to P.L. Correspondence should be addressed to Dr. Pat Levitt, The Saban Research Institute, Children’s Hospital Los Angeles, 4650 Sunset Boulevard, MS #135, Los Angeles, CA 90027. E-mail: plevitt@med.usc.edu.01 8 2016 29 8 2016 Jul-Aug 2016 3 4 ENEURO.0074-16.20165 4 2016 19 7 2016 25 7 2016 Copyright © 2016 Xie et al.2016Xie et al.This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International, which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed.Visual Abstract MET, a pleiotropic receptor tyrosine kinase implicated in autism risk, influences multiple neurodevelopmental processes. There is a knowledge gap, however, in the molecular mechanism through which MET mediates developmental events related to disorder risk. In the neocortex, MET is expressed transiently during periods of peak dendritic outgrowth and synaptogenesis, with expression enriched at developing synapses, consistent with demonstrated roles in dendritic morphogenesis, modulation of spine volume, and excitatory synapse development. In a recent coimmunoprecipitation/mass spectrometry screen, β-catenin was identified as part of the MET interactome in developing neocortical synaptosomes. Here, we investigated the influence of the MET/β-catenin complex in mouse neocortical synaptogenesis. Western blot analysis confirms that MET and β-catenin coimmunoprecipitate, but N-cadherin is not associated with the MET complex. Following stimulation with hepatocyte growth factor (HGF), β-catenin is phosphorylated at tyrosine142 (Y142) and dissociates from MET, accompanied by an increase in β-catenin/N-cadherin and MET/synapsin 1 protein complexes. In neocortical neurons in vitro, proximity ligation assays confirmed the close proximity of these proteins. Moreover, in neurons transfected with synaptophysin-GFP, HGF stimulation increases the density of synaptophysin/bassoon (a presynaptic marker) and synaptophysin/PSD-95 (a postsynaptic marker) clusters. Mutation of β-catenin at Y142 disrupts the dissociation of the MET/β-catenin complex and prevents the increase in clusters in response to HGF. The data demonstrate a new mechanism for the modulation of synapse formation, whereby MET activation induces an alignment of presynaptic and postsynaptic elements that are necessary for assembly and formation of functional synapses by subsets of neocortical neurons that express MET/β-catenin complex. β-cateninautismMET receptor tyrosine kinasesynapse developmentNational Institute of Mental HealthMH067842The Simms/Mann Chair in Developmental Neurogenetics cover-dateJuly/August 2016 ==== Body Significance Statement The gene encoding the MET receptor tyrosine kinase is associated with autism spectrum disorder, and influences typical and atypical synapse development and cortical circuit function. The present studies focus on determining potential molecular mechanisms through which the receptor functions in neocortical neurons during synaptogenesis. The findings show that the MET receptor interacts functionally with other proteins also implicated in promoting new synapse assembly, which is reduced upon disruption of the interactions. Thus, in some instances of autism spectrum disorder, disturbances of these molecular interactions may relate to the pathophysiology of cortical circuit development. Introduction The MET receptor tyrosine kinase has been implicated in multiple neurodevelopmental processes (Peng et al., 2013), and thus outcomes from disruptions in MET function vary according to cell context. For example, in the forebrain, a risk allele for autism spectrum disorder (ASD) in the MET promoter, which reduces MET transcript and protein levels (Campbell et al., 2006, 2007), is correlated with altered circuit function in typical and ASD human populations (Rudie et al., 2012) and in gray matter growth (Hedrick et al., 2012). Further, following conditional deletion of Met in mice, there is an increase in local interlaminar drive onto layer V neurons in the neocortex and premature maturation of excitatory synapse function in the hippocampus (Qiu et al., 2011, 2014). Analyses in vivo and in vitro demonstrate that MET signaling modulates dendritic morphogenesis; spine volume; the clustering of postsynaptic proteins; excitatory synapse formation; and maturation in the neocortex, striatum, and hippocampus (Gutierrez et al., 2004; Tyndall and Walikonis, 2006; Nakano et al., 2007; Lim and Walikonis, 2008; Judson et al., 2010; Finsterwald and Martin, 2011; Qiu et al., 2011, 2014; Kawas et al., 2013; Eagleson et al., 2016; Peng et al., 2016). These developmental influences likely underlie the mature forebrain circuit phenotypes observed in the context of altered MET signaling. How MET receptor activation mediates these discrete cellular outcomes is only beginning to be addressed, with most focus on the diversity of downstream signaling pathways initiated following the activation of MET (Finsterwald and Martin, 2011; Eagleson et al., 2016). Evidence from cell lines, however, indicates that the repertoire of MET protein-interacting partners expressed by a cell also can modulate MET signaling to influence biological outcomes (Smyth and Brady, 2005; Wang et al., 2005; Zeng et al., 2006; Reshetnikova et al., 2007; DeAngelis et al., 2010; Bozkaya et al., 2012; Burghy et al., 2012; Lu et al., 2012; Niland et al., 2013). A recent coimmunoprecipitation (Co-IP)/mass spectrometry (MS) study identified the MET interactome with 72 proteins, including β-catenin, in isolated murine neocortical synaptosomes during the peak of synaptogenesis (Xie et al., 2016). In the current study, we focused on the role of the MET/β-catenin protein complex in hepatocyte growth factor (HGF)-mediated neocortical synapse formation. Previous studies have shown the following: (1) that MET and β-catenin are expressed at the developing neocortical synapse (Phillips et al., 2001; Murase et al., 2002; Eagleson et al., 2013); (2) that MET activation increases synapse density on neocortical neurons in vitro (Eagleson et al., 2016); (3) that β-catenin regulates synaptic vesicle localization during presynaptic development in the hippocampus (Bamji et al., 2003; Yu and Malenka, 2003); and (4) that functional interactions between MET and β-catenin can be observed in hippocampal neurons, as well as cancer cell lines (Monga et al., 2002; Herynk et al., 2003; David et al., 2008), with the stability of the complex dependent upon the presence of HGF. MET and β-catenin physically interact with each other in vitro, and the activated MET receptor directly phosphorylates β-catenin at tyrosine142 (Y142; David et al., 2008). Consistently, following the addition of HGF in hippocampal neurons, β-catenin is phosphorylated at Y142 and dissociates from MET (Herynk et al., 2003; Rasola et al., 2007; David et al., 2008; Bhardwaj et al., 2013). Here, we used Co-IP/Western blot, proximity ligation assays (PLAs), and immunocytochemical analyses to determine how the MET/β-catenin complex might modulate neocortical synapse development in response to HGF. We report, following stimulation with HGF, a dynamic regulation of MET/β-catenin- and MET/synapsin 1-containing complexes in synaptosomes within minutes, a rapid increase in synapses in primary cultures of neocortical neurons. Both outcomes are dependent upon phosphorylation of β-catenin at Y142. We propose a model in which an axis of HGF/MET/β-catenin signaling modulates neocortical synapse development. Disruption of this signaling complex may contribute to ASD etiology. Materials and Methods Mice Timed-pregnant C57BL/6J mice were purchased from The Jackson Laboratory and the day of birth was considered postnatal day 0 (P0). Animals had free access to food and water and were housed in a 13/11 h light/dark cycle. All research procedures using mice were approved by the Institutional Animal Care and Use Committee at Children’s Hospital Los Angeles. All efforts were made to minimize animal suffering and to reduce the number of animals used. Plasmid construction Mouse β-catenin full-length cDNA was cloned by PCR from an adult mouse brain cDNA library using high proof PfuUltra II Fusion HS DNA Polymerase (Agilent) according to the manufacturer protocol, using the following primer pair: 5' CTAGCTAGCTAGATGGATACGTATCGCTACATAATGGCTACTCAAGC 3' and 5' TGCTCTAGAGCATTACAGGTCAGTATCAAACCAGGCCAGCTGATT 3'. Purified β-catenin cDNA fragments were subcloned into a PCI expression vector (Promega) and transformed into DH5α competent cells (Invitrogen). PCI-β-catenin plasmids were purified using the Zyppy Plasmid Maxiprep Kit (Zymo Research) and PCR site-directed mutagenesis of β-catenin (β-catenin Y142F) performed according to a published strategy (Zheng et al., 2004) using the following primer pair: 5' GTTGTCAATTTGATTAACTTCCAGGATGACGCGGAACTTG 3' and 5' CAAGTTCCGCGTCATCCTGGAAGTTAATCAAATTGACAAC 3'. The β-catenin and β-catenin Y142F fragments were PCR amplified using the following primer pair: 5' CGGGATCCATGGATACGTATCGCTACATAATGGCTACTCAAGC 3' and 5' CGGGATCCTTACAGGTCAGTATCAAACCAGGCCAGCTGATT 3'. The purified fragments were subcloned into a p3XFLAG-CMV-10 vector (Sigma-Aldrich) to generate p3XFLAG-CMV-10-β-catenin and p3XFLAG-CMV-10-β-cateninY142F plasmids. The fidelity of the entire coding sequences of all plasmids was confirmed by DNA sequencing (Genewiz). RNAscope P14 mouse brains were fresh frozen in ice-cold isopentane and sectioned in the coronal plane at 25 μm. Sections were subjected to dual fluorescent in situ hybridization using the RNAscope Multiplex Fluorescent Reagent kit (Advanced Cell Diagnostics) according to manufacturer instructions. RNAscope probes and the regions used to generate the probes were as follows: Met (catalog #405301-C2, Advanced Cell Diagnostics; accession #NM_008591.2 region 3370-4286) and β-catenin (catalog #311741, Advanced Cell Diagnostics; accession #NM_007614.3 region 342-2511). Alexa Fluor 488 and Atto550 detection reagents were used to visualize Met and β-catenin, respectively. Images were acquired using a Zeiss LSM 710 confocal microscope with a 20× objective. The imaging parameters and z-axis were adjusted to bring the sample into focus. The parameters were maintained to capture focused optical images in each wavelength. Co-IP and Western blot analysis All reagents for Co-IP and Western blots were from Sigma-Aldrich, unless otherwise noted. Crude synaptosomes were isolated from the neocortex of male and female P14 mice (Judson et al., 2010) were and resuspended in sodium bicarbonate-buffered oxygenated artificial CSF containing the following (in mM): NaCl 12.4, KCl 0.4, KH2PO4 0.1 (Baker), CaCl2 0.25 (Baker), MgCl2 0.1, dextrose 1 (VWR). Either 25 ng/ml HGF (R&D Systems) or the same volume of vehicle (PBS) was added for 5 min (Co-IP experiments) or for 5, 10, and 20 min (β-catenin phosphorylation experiment) at 37°C to the synaptosomes. For Co-IP experiments, the synaptosomes were centrifuged at 16,000 × g for 15 min, and the pellets were lysed in Co-IP lysis buffer containing the following (in mM): HEPES 50, pH 7.4, EGTA 2, EDTA 2, NaF 30, sodium orthovanadate 10, β-glycerol phosphate 40, 1% Triton X-100, and a protease inhibitor cocktail. The lysate was centrifuged at 16,000 × g for 30 min, and the resulting supernatant used for Co-IP, with a goat anti-MET antibody (R&D Systems), a mouse anti-β-catenin antibody (BD Biosciences), or a rabbit anti-N-cadherin antibody (Santa Cruz Biotechnology). An equivalent amount of goat, mouse or rabbit IgG antibody (Jackson ImmunoResearch) was used in parallel lysates as a negative control. The Co-IP complexes were bound to protein G plus agarose beads (Pierce), after which the beads were washed in Co-IP lysis buffer plus 150 mM NaCl. The complexes were eluted from the beads by boiling in the final sample buffer (12.5 mM Tris-HCl, pH 6.8, 5% Glycerol, 0.4% SDS, 1% 2-mercaptoethanol, 0.02% bromophenol blue) and analyzed by Western blot. For the β-catenin phosphorylation experiment, the synaptosomes were centrifuged at 16,000 × g for 15 min followed by lysis in final sample buffer. Blots were probed with antibodies directed against β-catenin (1:2000; BD Biosciences), N-cadherin (1:500; Santa Cruz Biotechnology), synapsin1 (1:4000; EMD Millipore), synaptophysin1 (1:2000; EMD Millipore), phospho-MET (1:500; Cell Signaling Technology), and MET (1:500; Santa Cruz Biotechnology). Digital images of the Western blots were acquired using a CCD camera coupled to a UVP BioImaging System using VisionWorksLS Image Acquisition software (version 8.0, UVP). Semiquantification of the Co-IPs Western blot analyses of Co-IPs in neocortical synaptosomes in the presence or absence of HGF were performed in three independent experiments. For each blot, representing an independent Co-IP experiment, the density of each immunoreactive band was measured in ImageJ (version 1.46r), and a background subtraction was applied. First, to account for different efficiencies of each pull down in each experiment, a ratio of coimmunoprecipitated protein (for example, Fig. 1A , β-catenin) to immunoprecipitated protein (for example, Fig. 1A , MET) was generated. Then, for each candidate, the data are expressed as the fold-change levels in the HGF-treated group compared with the PBS-treated group. The data are presented as box plots using GraphPad Prism 6. Figure 1. MET/β-catenin complexes during synapse development. A–C, Representative Western blots of complexes immunoprecipitated from P14 cortical crude synaptosomes using anti-MET (A), anti-β-catenin (B), anti-N-cadherin (C), or control IgG antibody. This experiment was repeated three times using independent synaptosomal preparations. In PBS-treated synaptosomes, β-catenin (α-β-cat) and MET (α-MET) are detected in MET and β-catenin, but not in control IgG, pulldowns. Similarly, β-catenin and N-cadherin (α-N-cad) are evident in N-cadherin and β-catenin, but not in control IgG, pulldowns. MET is not detected in the N-cadherin pulldown. Stimulation of synaptosomes for 5 min with HGF results in reduced MET and β-catenin complexes, with a concomitant increase in the β-catenin and N-cadherin complexes (HGF lane vs PBS lane). An anti-phospo-MET antibody (α-pMET) was used to confirm HGF-induced activation of the MET receptor. D, The fold change in the HGF-stimulated group compared with the PBS-stimulated group for each IP is presented as a box plot. The line bisecting the box represents the median. The horizontal red dash line indicates unchanged level (1.0) for comparison between HGF and PBS. N = 3 independent Co-IP experiments for each interaction. E, Representative confocal microscopy images of PLA staining in primary cultures of neocortical neurons at 14 DIV following treatment with PBS or HGF for 5, 10, and 30 min. Red fluorescent profiles represent regions of PLA signal amplification denoting MET and β-catenin colocalization. For comparison, the total MET immunoreactivity (green fluorescence) in the same field is illustrated. Scale bar, 5 μm. F, Quantitative analysis of the MET/β-catenin PLA signals. Error bars represent the SEM. N = 30 neurons from four independent culturing sessions in each group. **p < 0.01 (HGF vs PBS). G–G″, Dual RNAscope in situ hybridization for Met (red) and β-catenin (green) in the P14 mouse cortex. Nuclei were labeled with DAPI (blue) to distinguish the cortical layers. Representative confocal microscopy images show Met expression in superficial and deep layers, with β-catenin expression across all layers, with more intense labeling in layers V and VI (G). Higher-magnification images from superficial (G') and deep (G″) cortical layers. Dotted circles in G' and G″ indicate RNAscope-labeled single cells. Arrows in G' and G″ indicate Met and β-catenin colabeled cells. Arrowheads indicate Met (G') or β-catenin (G″) single-labeled neurons. Scale bars: G, 200 μm; G″ (for G' and G″), 25 μm. Primary neocortical neuron cultures Primary cultures of neocortical neurons were prepared from P1 mice (Beaudoin et al., 2012) with the following minor modifications. In each culturing session, tissue from two male and female pups was pooled, and ∼50,000 cells/cm2 were seeded onto 12 mm coverslips (Carolina Biological Supply Company) in 24-well plates, which were precoated with poly-d-lysine (Sigma-Aldrich). Cells were initially plated in DMEM (Invitrogen) supplemented with 10% fetal bovine serum (FBS). After 4 h, the medium was replaced with neurobasal medium (Invitrogen) supplemented with B27 (Invitrogen) and l-glutamine (Invitrogen), and one-half of the volume of medium was replaced every 3 d. This condition results in slower growth than when using glial conditioned medium or a glial feeder layer. To achieve the sparse labeling of neurons, at 5 d in vitro (DIV), cultures were transfected with a p3XFLAG-CMV10-β-catenin or p3XFLAG-CMV10-β-catenin Y142F plasmid using a calcium phosphate transfection kit according to manufacturer instructions (Clontech). In some experiments, a synaptophysin-GFP (Syn-GFP) plasmid (obtained from L. Reichardt, University of California, San Francisco, San Francisco, CA) was cotransfected to label synaptic vesicles. At 14 DIV, 25 ng/ml HGF or the same volume of PBS was added to the medium for 5, 10, or 30 min (PLA assays) or for 10 min (Syn-GFP cluster assays). The experiments were repeated in at least three independent culturing sessions. At the end of the assay period, coverslips were fixed with 4% paraformaldehyde for 15 min at room temperature and processed for immunocytochemistry. The PLA was used to determine spatial proximity between proteins that are immunolabeled with the Duolink in situ PLA kit (Sigma-Aldrich), as described previously (Eagleson et al., 2013). Immunofluorescent signals using dual imaging channels represent proteins that are within 40 nm or less of each other. The following antibody combinations with α-MET (1:50; R&D Systems) were used: (1) α-β-catenin (1:50; BD Biosciences); (2) α-synapsin1 (1:100; EMD Millipore); or (3) α-flag (1:200; Sigma-Aldrich). For the Syn-GFP cluster assays, coverslips were incubated with prechilled 100% methanol for 10 min at 4°C, then permeabilized with 0.1% Triton X-100 in PBS (PBST) for 20 min and blocked in blocking buffer (5% FBS in PBST) for 1 h at room temperature. Coverslips were incubated overnight at 4°C in the following primary antibody cocktails diluted in blocking buffer: (1) rabbit anti-bassoon (1:500; Cell Signaling Technology) and mouse anti-flag (1:1000; Sigma-Aldrich); or (2) mouse anti-PSD-95 (1:1000; Thermo Scientific) and rabbit anti-flag (1:1000). Following three washes with PBS, coverslips were incubated for 1 h at room temperature in the following cocktails of Alexa Fluor-labeled secondary antibodies (all at 1:1000; Life Technologies): (1) 546 goat-anti-mouse and 633 goat-anti-rabbit for bassoon/flag staining; or (2) 546 goat anti-rabbit and 633 goat-anti-mouse for PSD-95/flag staining. After three washes in PBS, coverslips were mounted onto glass slides with Prolong Mounting Medium (Life Technologies). Image analysis Images were captured using an automated laser-scanning confocal microscope (LSM 710, Zeiss) with a 60× oil-immersion objective. The focal point of the beam and the z-axis were adjusted until an appropriate focus was reached. All images in a given culturing session were captured and analyzed with the same exposure time and settings. Note that the visualization of spines requires a longer exposure than that needed to visualize the labeling of synaptic proteins in the linear range. Because the settings were optimized for the analysis of synapses, there is an apparent absence of dendritic protrusions in the captured images. For PLA analyses, in each culturing session, six to eight fields were randomly imaged for each treatment group. Quantitative measures of MET association with select protein partners were obtained by counts of PLA clusters/area of dendrite using published methods (Eagleson et al., 2013). For the Syn-GFP cluster assays, axons from eight transfected neurons, based on Syn-GFP labeling, were imaged for each treatment group in each culturing session. Images were imported into ImageJ (NIH) for analysis. Syn-GFP puncta colabeled with bassoon or PSD-95, and single-labeled Syn-GFP puncta, were identified using an automated plugin to ImageJ (https://github.com/Pat-Levitt/SynapseCounter; Wang et al., 2015), with thresholds set in each channel independently. Once the threshold was set for a given culturing session, the same threshold was used throughout the analyses. The following parameters were then measured for each colabeled Syn-GFP+ puncta using the “analyze particles” tool, as follows: (1) density (average number of puncta per 100 μm of axon length); (2) integrated density (the product of the puncta area and the average gray value within that area); (3) major length (the length of the major axis of Syn-GFP fluorescence signal expressed as the average Feret’s diameter); and (4) the size of the puncta. Statistics Data were expressed as the mean ± SEM. For each experimental manipulation, data were collected from at least three independent culturing sessions. Individual neurons were considered as samples (Sun and Bamji, 2011; Crowell et al., 2015), and sample size varied between studies and is indicated in the figure legends. The normality of the data was tested using the D’Agostino and Pearson omnibus normality test (D'Agostino and Pearson, 1986). If the data were not normally distributed, either the data were transformed to meet the assumption of normal distribution or a nonparametric test was applied. Specifically, the data in Figures 1F, 4D, and 6C were transformed with a square root, and the data in Figure 6D were transformed with a natural logarithm to meet the assumption of normality for two-way ANOVA analyses. For two-way ANOVAs, means were compared to determine the effects of treatment (PBS and HGF) × treatment time, or treatment × transfected plasmid (wild-type β-catenin and β-catenin-Y142F), and the interaction between those factors. If a significant effect was detected, a Bonferroni’s multiple-comparisons test was performed to determine the possible source of interactions. When the data were not normally distributed, the Mann–Whitney U test was used to compare differences between PBS- and HGF-treated groups. The unpaired t test with Welch’s correction was applied to compare differences between PBS- and HGF-treated groups if the data were normally distributed. For all tests, p values are reported to the fourth decimal place, and values <0.05 are considered to be significant. Statistical analyses and the preparation of graphs were performed using GraphPad Prism 6.0. Results HGF downregulates the MET and β-catenin complex during synapse development Previously, a discovery-based Co-IP/MS method was used to detect the MET interactome in neocortical synaptosomes at the peak of synaptogenesis (P14) and identified β-catenin as a member of the MET interactome (Xie et al., 2016). In this study, we took several approaches to validate and characterize the functional relevance of the MET/β-catenin complex. First, we confirmed the presence of a MET/β-catenin complex in neocortical crude synaptosomes using Western blot analysis of Co-IPs. Specifically, β-catenin was detected in MET-immunoprecipitated complexes (Fig. 1A ) and, conversely, MET was present in β-catenin-immunoprecipitated complexes (Fig. 1B ). Neither protein was detected in complexes that had been immunoprecipitated with control IgG. Activation of the MET receptor following treatment of the synaptosomes with HGF for 5 min was accompanied by a significant decrease in MET/β-catenin complex [mean fold change (HGF/PBS), 0.3987 (95% CI, 0.0773–0.7201) for β-catenin in MET IP; mean fold change (HGF/PBS), 0.3760 (95% CI 0.0606–0.6915) for MET in β-catenin IP; Fig. 1D ]. The β-catenin and N-cadherin complex also occurs during synapse formation (Uchida et al., 1996). Interestingly, while we confirmed this complex following immunoprecipitation with β-catenin and N-cadherin antibodies (Fig. 1B,C ), N-cadherin could not be detected in MET pulldowns (Fig. 1A ). These data show that β-catenin forms separate complexes with MET and with N-cadherin. Moreover, following HGF stimulation, there is an increase in N-cadherin/β-catenin complex [mean fold change (HGF/PBS), 1.309 (95% CI 1.018–1.600) for N-cadherin in β-catenin IP; mean fold change (HGF/PBS), 1.388 (95% CI 0.6342–2.141) for β-catenin in N-cadherin IP; Fig. 1D ] that complement the reduced MET/β-catenin complex (Fig. 1A,B,D ). We next used the PLA to examine the proximity of MET/β-catenin in primary cultures of neocortical neurons. At 14 DIV, PLA signal was detected in the presence of MET and β-catenin antibodies (Fig. 1E ), indicating a close proximity between these two proteins. There was a significant effect of treatment (PBS vs HGF) on the magnitude of the PLA signal (F(1,174) = 9.401, p = 0.0025, Fig. 1F ). Pairwise analysis revealed that there was no effect for HGF stimulation for 5 min (PBS vs HGF: p a = 0.8383, df = 174; Table 1), but there was a significant reduction in PLA signal after HGF stimulation for 10 min (p b = 0.0002, df = 174). Specifically, after stimulation for 10 min, there was an ∼50% decrease in the density of PLA clusters in HGF-treated compared with PBS-treated cultures (Fig. 1F ). The PLA signal returned to prestimulation levels 30 min following HGF treatment (PBS vs HGF: p c = 0.0723, df = 174; Fig. 1F ). Together, the Co-IP/Western and PLA analyses demonstrate that MET/β-catenin proximity is regulated in an HGF-dependent manner. Table 1: Statistical table Figures Comparisons Label Data structure Type of test Power 1F PBS vs HGF (5 min) a Normal distribution after transformation Two-way ANOVA with Bonferroni's multiple comparisons test 0.26 PBS vs HGF (10 min) b 1.00 PBS vs HGF (30 min) c 0.93 2C PBS vs HGF d Non-normal distribution Mann–Whitney test 0.99 3C PBS vs HGF e Normal distribution Unpaired t test with Welch's correction 1.00 4D β-catenin vs β-cateninY142F PBS treatment) f Normal distribution after transformation Two-way ANOVA Bonferroni’s multiple comparisons test 0.17 β-catenin vs β-cateninY142F HGF treatment) g 1.00 PBS vs HGF (β-catenin transfection) h 1.00 PBS vs HGF (β-cateninY142F transfection) i 0.01 5E PBS vs HGF (Syn-GFP/bassoon) j Non-normal distribution Mann–Whitney test 0.99 PBS vs HGF (Syn-GFP/PSD-95) k Non-normal distribution Mann–Whitney test 0.99 PBS vs HGF (Syn-GFP) l Non-normal distribution Mann–Whitney test 0.94 5F PBS vs HGF (Syn-GFP/bassoon) m Non-normal distribution Mann–Whitney test 0.94 PBS vs HGF (Syn-GFP/PSD-95) n Non-normal distribution Mann–Whitney test 0.78 PBS vs HGF (Syn-GFP) o Non-normal distribution Mann–Whitney test 0.82 5G PBS vs HGF (Syn-GFP/bassoon) p Non-normal distribution Mann–Whitney test 0.90 PBS vs HGF (Syn-GFP/PSD-95) q Non-normal distribution Mann–Whitney test 0.75 PBS vs HGF (Syn-GFP) r Normal distribution Unpaired t test with Welch's correction 0.92 5H PBS vs HGF (Syn-GFP/bassoon) s Non-normal distribution Mann–Whitney test 0.91 PBS vs HGF (Syn-GFP/PSD-95) t Non-normal distribution Mann–Whitney test 0.90 PBS vs HGF (Syn-GFP) u Non-normal distribution Mann–Whitney test 0.88 6C β-catenin vs β-cateninY142F PBS treatment, Syn-GFP/bassoon) v Normal distribution after transformation Two-way ANOVA Bonferroni's multiple comparisons test 0.00 PBS vs HGF (β-catenin transfection, Syn-GFP/bassoon) w 0.98 PBS vs HGF (β-cateninY142F transfection, Syn-GFP/bassoon) x 0.69 6D β-catenin vs β-catenin Y142F PBS treatment, Syn-GFP/PSD-95) y Normal distribution after transformation Two-way ANOVA Bonferroni's multiple comparisons test 0.00 PBS vs HGF (β-catenin transfection, Syn-GFP/PSD-95) z 1.00 PBS vs HGF (β-cateninY142F transfection, Syn-GFP/PSD-95) aa 0.00 To meet the assumption of normality, the data from Figures 1F, 4D, and 6C were transformed with the square root, and the data from Figure 6D were transformed with the natural logarithm. The normality of the data was tested using the D’Agostino and Pearson omnibus normality test. In the third set of experiments, we used RNAscope to examine the coexpression of Met and β-catenin in the mouse neocortex at P14, providing an anatomical context for MET and β-catenin complex in vivo. There was a gradient of Met and β-catenin coexpression across neocortical layers (Fig. 1G ). Specifically, in layers II/III, there were many neurons that coexpressed the Met and β-catenin transcripts (Fig. 1G' ). In deeper layers, the signal intensity of puncta was greater in colabeled neurons (Fig. 1G″ ). In both superficial and deep layers, there also were single-labeled Met or β-catenin pyramidal neurons (Fig. 1G',G″ ). These data indicate that the MET/β-catenin complex resides in subsets of neocortical neurons during the peak period of synapse formation. This heterogeneity in neuronal coexpression may account in part for the disruption of physiological functions in a subset of neocortical synapses after Met deletion (Qiu et al., 2011). HGF increases the MET/synapsin 1 complex during synapse development Following HGF stimulation, β-catenin dissociates from the MET complex. We hypothesized that the activation of MET could result in the recruitment of other protein partners into the receptor complex. MET is localized in presynaptic and postsynaptic compartments, but with predominant enrichment in presynaptic compartments in the developing neocortex (Eagleson et al., 2013) and cultured neocortical neurons. β-Catenin also is densely colocalized with presynaptic markers synapsin 1 and neurexin 1 compared with postsynaptic marker PSD-95 in cultured neocortical neurons (Fig. 2A ). Thus, the experiments here focused on the impact of MET receptor activation on presynaptic proteins. In P14 neocortical synaptosomes, synapsin 1 is coimmunoprecipitated with MET (Fig. 2B ). Further, in contrast to β-catenin, following the activation of MET by HGF stimulation for 5 min, additional synapsin 1 was recruited to MET complexes [mean fold change (HGF/PBS), 1.804; 95% CI 1.114–2.494; Fig. 2B ]. Consistent with this, the density of the PLA signal generated by MET and synapsin 1 antibody labeling in neocortical neurons in vitro was significantly increased (∼1.5 fold, p d = 0.0109, df = 58) following HGF addition, compared with PBS treatment (Fig. 2C,D ). In contrast with the MET/synapsin 1 complex, MET and synaptophysin 1 do not coimmunoprecipitate under conditions with or without HGF stimulation (Fig. 2E ). These data suggest that there is an increase in a functional MET/synapsin 1 complex following HGF stimulation. Figure 2. HGF recruits additional synapsin 1 to MET receptor complex. A, Quantitative analysis of PLA signals generated with β-catenin alone (β-cat), β-catenin with MET (+ MET), β-catenin with synapsin 1 (+ Syn 1), β-catenin with Neurexin 1 (+ Nxn 1), and β-catenin with PSD-95 (+ PSD-95). Error bars represent the SEM; N = 6–8 cells from independent cultures for each group. B, Representative Western blots of complexes immunoprecipitated from P14 cortical crude synaptosomes using an anti-MET or control IgG antibody. This experiment was repeated three times using independent synaptosomal preparations. In PBS-treated synaptomsomes, synapsin 1 and MET are detected in the MET, but not IgG, pulldowns. Stimulation of the synaptosomes for 5 min with HGF increased the MET/synapsin 1 complex (HGF vs PBS lane). An anti-phospo-MET antibody (α-pMET) was used to confirm HGF-induced activation of the MET receptor. C, Representative confocal microscopy images of PLA staining in primary cultures of neocortical neurons at 14 DIV following treatment with PBS or HGF for 10 min. Red fluorescent profiles represent regions of PLA signal amplification denoting MET and synapsin 1 colocalization. For comparison, the total MET immunoreactivity (green fluorescence) in the same field is illustrated. Scale bar, C (for B, C), 5 μm. D, Quantitative analysis of the MET/synapsin 1 PLA signals. Error bars represent the SEM; N = 30 cells from five independent cultures in each group. *p < 0.05 (HGF vs PBS). E, Representative Western blots of complexes that were immunoprecipitated from P14 cortical crude synaptosomes using an anti-MET or control IgG antibody. MET is detected in the MET, but not IgG, pulldowns. Stimulation of the synaptosomes for 5 min with HGF results in phospho-MET detection in MET pulldowns (HGF vs PBS lane). A single synaptophysin 1 band is readily detected in the input sample prior to IP. In contrast, the postimmunoprecipitation sample has only nonspecific bands in all Co-IP groups, indicating that synaptophysin 1 does not coimmunoprecipitate with MET under these conditions. HGF regulates MET and β-catenin complex through phosphorylation of β-catenin at Y142 Previous reports demonstrated phosphorylation of β-catenin at Y142 in response to HGF in cancer cells and cultured hippocampal neurons (Herynk et al., 2003; Rasola et al., 2007; David et al., 2008; Bhardwaj et al., 2013). Similarly, we found that, following the addition of HGF to P14 neocortical synaptosomes, the level of phosphorylated of β-catenin (Y142) was increased, reaching a peak after 5 min, and declining toward prestimulation levels by 20 min (Fig. 3A ). Moreover, in cultured neocortical neurons at 14 DIV, the number of immunoreactive puncta, labeled with an antibody that recognizes Y142-phosphorylated β-catenin (Strom et al., 2007; David et al., 2008), significantly increased after HGF treatment for 10 min (approximately twofold; p e = 0.003, t = 3.882, df = 51; Fig. 3B,C ). Together, these results indicate that HGF can regulate β-catenin phosphorylation at Y142 in the neocortex during the period of synapse formation. Figure 3. HGF promotes phosphorylation of β-catenin at Y142. A, Representative Western blots of crude neocortical synaptosomes following stimulation with HGF stimulation for 0, 5, 10, and 20 min. This experiment was repeated three times using independent synaptosomal preparations. The level of β-catenin phosphorylated at Y142 (p142-β-cat) increased in the presence of HGF, peaking at 5 min. Note the expected increase, followed by a time-dependent decrease, in phospho-MET (pMET) levels in response to HGF. Total levels of β-catenin (β-cat) and MET are unchanged. B, Representative confocal microscopy images of primary neocortical neurons at 14 DIV following stimulation for 5 min with PBS or HGF (25 ng/ml). This experiment was repeated in two independent culturing sessions. Note the increase in immunostaining of p142- β-catenin in the presence of HGF. Scale bar, 20 μm. C, Quantitative analysis of the p142- β-catenin clusters. Error bars represent the SEM; N = 26 cells from two independent cultures in each group. *p < 0.05 (HGF vs PBS). To address the possibility that phosphorylation of β-catenin at Y142 modulates the extent to which MET and β-catenin functionally interact, neocortical neurons were transfected with wild-type β-catenin that can be phosphorylated at Y142 with HGF treatment (Fig. 4A,B ) or a mutant form of β-catenin that cannot be phosphorylated at Y142 (β-cateninY142F; David et al., 2008). PLA analyses at 14 DIV demonstrated that, consistent with our previous data with endogenous β-catenin and MET associated, either directly or indirectly, with transfected β-catenin and MET, this association is downregulated by HGF (Fig. 4C ). Quantitative analysis revealed a significant functional association between the construct transfected (wild-type β-catenin vs β-cateninY142F) and treatment (PBS vs HGF) on the density of PLA clusters (F(1,84) = 9.570, p = 0.0027; Fig. 4C,D ). Pairwise analysis revealed that this effect was due to differences in the PLA signal following HGF stimulation. Specifically, there was no significant difference in the density of PLA clusters between wild-type β-catenin and β-cateninY142F in the absence of HGF (p f = 0.8269, df = 84; Fig. 4C,D ), demonstrating that β-cateninY142F still associated in a complex with MET. Following stimulation with HGF, however, there was a significant difference between wild-type β-catenin and β-cateninY142F (p g = 0.0013, df = 84; Fig. 4C,D ). Specifically, there was an ∼70% decrease in the density of PLA clusters in neurons transfected with wild-type β-catenin (p h = 0.0002, df = 84), but no significant difference in PLA signal in flag-β-cateninY142F-transfected neurons (p i > 0.9999, df = 84; Fig. 4C,D ), compared with PBS. These data demonstrate that HGF-stimulated phosphorylation of β-catenin at Y142 is required for the subsequent dissociation of the MET/β-catenin complex. We noted that the transfected β-catenin or β-cateninY142F was distributed in the entire neuron (Fig. 4A ), but the PLA signals are rarely present in dendrites. The labeling pattern may suggest that HGF stimulation impacts the functional association of MET with transfected β-catenin at presynaptic sites, but we cannot exclude postsynaptic complex interactions. Figure 4. HGF modulates MET/β-catenin complex via phosphorylation of β-catenin at Y142. A, Neurons were transfected with flag-tagged wild-type β-catenin (Flag-β-cat). Representative confocal microscopy image of transfected neuron with total flag immunoreactivity (white) was shown. Note that transfected β-catenin was distributed along the entire neuron and processes. Scale bar, 50 μm. B, Representative confocal microscopy images of Flag-β-cat-transfected neurons with total flag (green) and p142- β-catenin immunoreactivity (white). Note the positive immunostaining of p142- β-catenin in the Flag-β-cat-transfected neuron with stimulation of HGF. Scale bar, 25 μm. C, Representative confocal microscopy images of PLA staining of MET and flag in primary cultures of neocortical neurons at 14 DIV following treatment with PBS or HGF for 10 min. Neurons were transfected with Flag-β-cat or β-cateninY142F (Flag-β-catY142F) at 5 DIV. Red fluorescent profiles represent regions of PLA signal amplification denoting MET and flag colocalization. For comparison, the total flag immunoreactivity (green fluorescence) in the same field is illustrated. Scale bar, 5 μm. D, Quantitative analysis of the MET/flag PLA signals. Error bars represent the SEM; N = 22 cells from three independent cultures in each group. Note that there is a decrease in PLA signal with HGF in the wild-type Flag-β-catY, but that there is no change with HGF in the Flag-β-catY142F condition. **p < 0.01 (HGF vs PBS in Flag-β-cat-transfected group); p = n.s. (Flag-β-cat vs Flag-β-catY142F with PBS stimulation, HGF vs PBS in Flag-β-catY142F-transfected group). HGF increases the density of Syn-GFP/bassoon and Syn-GFP/PSD-95 clusters in neocortical neurons Given the increased MET/synapsin 1 complex following activation of the receptor, we reasoned that HGF/MET signaling might contribute to presynaptic development through an alteration in protein complexes. To address this, we transfected neurons with Syn-GFP, which is a marker of synaptic vesicles, localized at presynaptic sites and axon but not at dendrites (Fig. 5B; Bamji et al., 2003). MET/synaptophysin are not found in the same protein complex. Following transfection, neurons were treated with HGF for 10 min at 14 DIV. At the end of the treatment period, we categorized Syn-GFP puncta (Fig. 5A ) according to whether they were (1) colabeled with bassoon (Fig. 5C ), an active zone marker; (2) colabeled with PSD-95 (Fig. 5D ), a postsynaptic marker; or (3) total labeled Syn-GFP puncta (Syn-GFP), including puncta colabeled with bassoon or PSD-95 and puncta labeled with Syn-GFP alone. Mann–Whitney statistical analyses revealed a significant increase in the density of Syn-GFP puncta colabeled with bassoon (∼2.1-fold, U(34) = 80, p j = 0.0087; Fig. 5C,E ) or with PSD-95 (∼1.8-fold, U(34) = 84, p k = 0.0129, Fig. 5D,E ) in HGF-treated compared with PBS-treated cultures. In contrast, there was no significant difference in the density of clusters labeled with Syn-GFP (U(34) = 103, p l = 0.0633). We further characterized Syn-GFP puncta colabeled with bassoon or PSD-95, measuring parameters previously shown to be influenced by β-catenin (Sun et al., 2009). There was no significant effect of HGF treatment on the integrated density (Fig. 5F ), major length (Fig. 5G ), or size (Fig. 5H ) of Syn-GFP puncta, which approximate the size of the synaptic vesicle pool (Sun et al., 2009). Together, these data suggest that HGF promotes the rapid assembly of synaptic vesicles at active zones to increase the formation of nascent synapses, but does not further cause the accumulation of synaptic vesicles at existing synaptic sites. Figure 5. MET activation increases the density of Syn-GFP/bassoon and Syn-GFP/PSD-95 clusters in neocortical neurons. A, Diagram of Syn-GFP cluster assay. Neocortical neurons were transfected with Syn-GFP at 5 DIV and treated with PBS or HGF (25 ng/ml) for 10 min at 14 DIV. Measurements were made of (1) Syn-GFP/bassoon (a marker of the active zone) colabeled clusters, (2) Syn-GFP/PSD-95 (a marker of the postsynaptic density) colabeled clusters, and (3) clusters labeled with Syn-GFP along the axon (with active zone and synapse included). B, Representative confocal microscopy image of Syn-GFP-transfected (green) neurons. Note that Syn-GFP clusters are present along axons but not dendrites. Scale bar, 50 μm. C, D, Representative confocal microscopy images of Syn-GFP (green), bassoon (red), and PSD-95 (magenta) immunoreactivity in neocortical neurons. White arrows indicate clusters colabeled with Syn-GFP/bassoon (C) or Syn-GFP/PSD-95 (D). Scale bars: C, D, 5 μm. E–H, Quantitative analysis of the density (E), integrated density (F), major length (G), and size (H) of Syn-GFP clusters. Each parameter was normalized in each culturing session to the mean value of Syn-GFP clusters in the PBS-treated group. Error bars represent the SEM; N = 18 cells from three independent cultures in each group. Increases in colabeling of presynaptic/postsynaptic markers are evident following HGF stimulation. *p < 0.05, **p < 0.01 (HGF vs PBS). Phosphorylation of β-catenin at Y142 is required for the HGF-induced increase in the density of Syn-GFP/bassoon and Syn-GFP/PSD-95 clusters Previous studies have demonstrated a role for an N-cadherin/β-catenin complex in synaptic vesicle localization at the synapse (Brigidi and Bamji, 2011). Given our data showing that, following HGF stimulation, there is a reduction of the MET/β-catenin complex that is accompanied by an increase in the β-catenin/N-cadherin complex, we wondered whether the HGF upregulation of Syn-GFP/bassoon and Syn-GFP/PSD-95 clusters requires dissociation of the MET/β-catenin complex. To address this, we transfected primary cultures of neocortical neurons with wild-type β-catenin or β-cateninY142F, which can form a complex with MET, but does not dissociate from the complex after HGF stimulation. Consistent with our data in untransfected cultures, HGF increased the density of Syn-GFP/bassoon and Syn-GFP/PSD-95 clusters in neurons transfected with wild-type β-catenin (Fig. 6A ). In contrast, there was no increase in the density of these clusters in β-cateninY142F-transfected neurons (Fig. 6B ). Quantitative analysis demonstrated a significant interaction between the construct transfected (wild-type β-catenin or β-cateninY142F) and treatment (PBS or HGF) on the density of Syn-GFP/bassoon (F(1,92) = 8.056, p = 0.0056) and Syn-GFP/PSD-95 (F(1,116) = 4.141, p = 0.0441) clusters. Pairwise analysis revealed that, in the absence of HGF, there was no significant difference in the density of Syn-GFP/bassoon (p v > 0.9999, df = 92; Fig. 6C ) or Syn-GFP/PSD-95 (p y > 0.9999, df = 116; Fig. 6D ) clusters in neurons transfected with wild-type β-catenin or β-cateninY142F. This result suggests that β-cateninY142F does not disrupt Syn-GFP/bassoon or Syn-GFP/PSD-95 cluster density under conditions in which MET is not stimulated. Following the addition of HGF, however, there was a significant increase in the density of Syn-GFP/bassoon (p w = 0.0232, df = 92) and Syn-GFP/PSD-95 (p z = 0.0018, df = 116) clusters in neurons transfected with wild-type β-catenin, but not in flag-β-cateninY142F-transfected neurons (Syn-GFP/bassoon: p x = 0.3078, df = 92; Syn-GFP/PSD-95: p aa > 0.9999, df = 116), compared with PBS. These results demonstrate that HGF activation of MET promotes an increased density of Syn-GFP/bassoon and Syn-GFP/PSD-95 clusters by regulating the MET/β-catenin complex through the phosphorylation of β-catenin at Y142. Figure 6. Phosphorylation of β-catenin at Y142 is required for the HGF-induced increase in Syn-GFP/bassoon and Syn-GFP/PSD-95 clusters. A, B, Representative confocal microscopy images of Syn-GFP (green), bassoon (red), PSD-95 (magenta), and flag (blue) immunoreactivity in neocortical neurons at 14 DIV after treatment with PBS or HGF (25 ng/ml) for 10 min. The neurons were cotransfected with Syn-GFP and either flag-β-catenin (A) or flag-β-cateninY142F (B) plasmids at 5 DIV. White arrows indicate clusters colabeled with Syn-GFP/bassoon (top panels) or Syn-GFP/PSD-95 (bottom panels). Scale bars: A, B, 5 μm. C, D, Quantitative analysis of the density of Syn-GFP/bassoon (C) and Syn-GFP/PSD-95-colabeled clusters in β-catenin (β-cat)- or β-cateninY142F (β-catY142F)-transfected neurons. Each parameter was normalized in each culturing session to the mean value of Syn-GFP clusters in the PBS-treated group. Error bars represent the SEM; N = 24 cells from three independent culturing sessions for bassoon colabeling assay; N =30 cells from three independent culturing sessions for PSD-95 colabeling assay. Note that the inability to phosphorylate Y142 residue following HGF treatment results in no change in the colabeling of presynaptic/postsynaptic markers. *p < 0.05, **p < 0.01 (HGF vs PBS). Discussion Many genes have been identified as components of molecular networks involved in ASD risk (Bourgeron, 2009; Pinto et al., 2014). These putative relations are placed in a functional context by examining how interactions at the protein level impact typical and atypical neurodevelopment (Xie et al., 2016). In this context, we demonstrate here a dynamic, ligand-dependent functional interaction, either directly or indirectly, between two proteins encoded by ASD risk genes, MET (Campbell et al., 2006, 2008; Jackson et al., 2009; Thanseem et al., 2010; Zhou et al., 2011; Rudie et al., 2012; Abrahams et al., 2013; Lambert et al., 2014) and β-catenin (O’Roak et al., 2012a,b). A possible association between MET and WNT/β-catenin was recently put forth as a contributing mechanism through which neurodevelopmental events impacted in ASD are coordinated (Mullins et al., 2016). Previous reports of functional interactions between MET and β-catenin have focused on transcriptional regulation, in which HGF promotes the phosphorylation of β-catenin at Y142 directly via activating the MET receptor (David et al., 2008), followed by the dissociation of β-catenin from a MET complex and then translocated to the nucleus (Herynk et al., 2003; Rasola et al., 2007; Bhardwaj et al., 2013). We demonstrate a similar modulation of the MET/β-catenin functional interaction by HGF at the neocortical synapse. In contrast with nuclear translocation, however, the released β-catenin is, at least in part, sequestered into N-cadherin complexes, with activated MET receptor recruiting additional synapsin 1 to form functional complexes that may be mediated through other proteins in a complex. In addition to MET, phosphorylation of β-catenin at Y142 could also be induced by other non-receptor tyrosine kinases such as Fer or Fyn tyrosine kinase. This latter phosphorylation downregulates the interactions of β-catenin and α-catenin, but does not affect the β-catenin and cadherin adhesive complex, which is controlled by the phosphorylation of β-catenin at Y654 (Roura et al., 1999; Piedra et al., 2003; Tai et al., 2007). Independent of HGF/MET signaling, the regulation of cadherin/β-catenin/α-catenin complex by Fer and Fyn tyrosine kinases also contributes to synapse development (Bamji et al., 2006; Arikkath and Reichardt, 2008; Lee et al., 2008). The dissociation of the MET/β-catenin complex is required for the HGF-induced increase in the density of Syn-GFP/bassoon and Syn-GFP/PSD-95 clusters in neocortical neurons. This increase, which is observed over a short assay period of 10 min, is suggestive of an increased number of nascent synapses in the presence of HGF and is reminiscent of increased synapse density, defined by synapsin 1 and PSD-95 colocalization, in the same culture paradigm after stimulation with HGF for 24 h (Eagleson et al., 2016). It should be noted that MET signaling also appears to modulate excitatory synapse maturation in the hippocampus and neocortex (Qiu et al., 2014; Peng et al., 2016), such that genetic deletion of Met results in premature maturation, assessed morphologically, electrophysiologically, and by the increased membrane insertion of AMPA receptor subunits (Qiu et al., 2014). Overexpression of MET in hippocampal neurons or slices in vitro results in both immature spine growth and electrophysiological properties. The data together are consistent with a unique dual role for MET during synapse development: initiating synapse formation at early stages and maintaining an immature functional state until MET signaling is eliminated. In vivo, MET receptor activation is robust during the period of rapid synapse formation between P7 and P14, but then rapidly falls to near-negligible levels by P16 (Eagleson et al., 2016), a time when neocortical synapses are undergoing maturation. While speculative, MET contribution to new synapse formation, followed by a decline to permit maturation, may contribute to the generation of an appropriate number or function of mature excitatory synapses in the developing neocortex. The disruption of MET signaling does increase excitatory drive and synapse maturation, and thus may alter excitation/inhibition balance, a key element contributing to neurodevelopmental disorders, including ASD (Rubenstein and Merzenich, 2003; Levitt et al., 2004; Dani et al., 2005; Gogolla et al., 2009; Gatto and Broadie, 2010; Bateup et al., 2013). In the current study, we focused on the role of MET/β-catenin complex in modulating presynaptic development. Specifically, HGF increases synaptic vesicles clustering at the active zone and at the synapse through regulation of MET/β-catenin complex by the phosphorylation of β-catenin at Y142. We also showed that, under basal culture condition, β-catenin as well as β-catenin Y142F itself could promote synapse formation. This may reflect the role of β-catenin independent of HGF/MET signaling and phosphorylation of β-catenin at Y142 as discussed previously (Arikkath and Reichardt, 2008; Sun et al., 2009; Brigidi and Bamji, 2011), as well as the low concentration of HGF in our basal culture condition. The low concentration of HGF may not generate sufficient phosphorylation of β-catenin at Y142 to induce difference between β-catenin and β-cateninY142F-transfected neurons. It should be noted that the regulation of MET/β-catenin complex by HGF may occur at both presynaptic and postsynaptic sites, because in vivo, both β-catenin and MET are also localized at postsynaptic sites in the neocortex and hippocampus (Phillips et al., 2001; Murase et al., 2002; Eagleson et al., 2013). Thus, the increased alignment of synaptic vesicles at active zone and at synapse induced by HGF could be triggered through regulation of both presynaptic and postsynaptic MET/β-catenin complexes in the cultured neurons. It also should be noted that MET is localized predominantly in presynaptic compartments, and there is no individual synapse with MET distributed at both presynaptic and postsynaptic sites in the developing neocortex in vivo (Eagleson et al., 2013). While it awaits formal testing, we favor the hypothesis that the major regulation of MET/β-catenin complexes would occur at the presynaptic site in the developing neocortex in vivo. However, we note that β-catenin and MET have been demonstrated independently to modulate several features of postsynaptic development. For example, β-catenin regulates dendritic morphogenesis, dendritic spine density, and postsynaptic structure and function (Murase et al., 2002; Yu and Malenka, 2003, 2004; Abe et al., 2004; Gao et al., 2007; Okuda et al., 2007). Similarly, there is increasing evidence that MET signaling modulates dendritic and spine morphogenesis, as well as the clustering of postsynaptic proteins, including PSD-95 (Gutierrez et al., 2004; Lim and Walikonis, 2008; Judson et al., 2010, 2011; Finsterwald and Martin, 2011; Qiu et al., 2014; Peng et al., 2016). Thus, it is plausible that the MET/β-catenin complex, and the regulation of this complex by HGF, may influence different aspects of presynaptic and postsynaptic development. The analyses of HGF-induced regulation of MET/β-catenin functional interactions in neocortical neurons in vitro raise the issue of defining the cellular and circuit context in which this signaling system may operate in vivo. At P14, we found that there are both colabeled and single-labeled neurons expressing Met and β-catenin transcripts in layers II-III and V-VI. The double-labeled neurons located superficially are almost entirely intrinsic intratelencephalic cortico-cortical or callosal neurons, whereas deep layer MET+ neurons could be both callosal and cortico-fugal in nature. During the active period of synaptogenesis in the mouse neocortex, Hgf mRNA is evident mostly in deep layers of the neocortex at P14 (Eagleson et al., 2016). Thus, both ligand and receptor are positioned to modulate MET/β-catenin complexes in subsets of neurons in vivo. Determining the specific subpopulation identities is currently under investigation. Our current findings are consistent with converging evidence that the developing neocortical synapse is disrupted in ASD, with many ASD risk genes having implicated or demonstrated roles in synapse development and plasticity (Zoghbi, 2003; Garber, 2007; Südhof, 2008; Zoghbi and Bear, 2012; De Rubeis et al., 2014; Xie et al., 2016). Less progress has been made in understanding the heterogeneity in clinical presentation, which likely reflects the polygenic nature of the disorder. Our data suggest that an understanding of ASD risk at the level of protein functional interactions, including the identification of the specific subpopulations of neurons and circuits in which these interactions occur, will provide insight into how such heterogeneity arises. For example, MET expression in the primate brain is enriched in temporal, posterior parietal, and occipital regions, with very limited expression in a few frontal lobe areas. Neuroimaging studies confirm that the MET promoter risk variant impacts the structure and function of circuits in which it is enriched (Rudie et al., 2012). At the cellular level, the RNAscope analyses reveal that a subset of neocortical neurons coexpress MET and β-catenin during the peak period of synaptogenesis. This suggests that the biological impact of reducing MET expression, which occurs in ASD and Rett syndrome (Campbell et al., 2007; Voineagu et al., 2011; Plummer et al., 2013), may differentially disrupt the development of subpopulations of neurons, with specific changes being dependent on the specific repertoire of MET-interacting proteins expressed by different neurons and circuits. Advances in multiplex in situ techniques will provide opportunities to more carefully characterize the coexpression of multiple members of the MET interactome, 11% of which have been associated with neurodevelopmental disorders (Xie et al., 2016), in discrete neocortical neuron subpopulations. Synthesis The decision was a result of the Reviewing Editor Robert Kalb and the peer reviewers coming together and discussing their recommendations until a consensus was reached. A fact-based synthesis statement explaining their decision and outlining what is needed to prepare a revision is listed below. The following reviewers agreed to reveal their identity: Randall Walikonis, Charles Finsterwald Both reviewers find the topic interesting and novel. There is enthusiasm for the report. The two reviewers differ in their opinions about whether the work is appropriate controlled and cautiously interpreted. One of the key issues is localization and direct v. indirect protein-protein interactions. I think the authors can address this with new experiments and so I favor returning the ms. to them for revision and re-review Dear Authors, I have redacted portions of both reviewers comments and my synthesis is highlighted by ***. I think you will find my comments helpful as enthusiasm of the external reviewers for the revision will be largely determined by how well the highlighted issues are addressed. 1. The conclusions of this paper are heavily dependent upon the assumption that the proximity ligation assay shows direct interaction between proteins. However, signals from PLA simply mean that the target proteins are within about 40 nm of each other, and proteins in proximity to each other within a cellular subdomain or protein complex can give a PLA signal even if they are not directly interacting. The conclusions about the data that is shown must be tempered with this understanding. ***The authors could be more careful about their conclusion as for the direct interaction between Met and b-catenin or synapsin. Bringing new methods for addressing this critical issue would strongly help the cause. A yeast 2 hybrid system approach or GST-fusion protein approach are potential paths forward 2. MET and beta-catenin are probably also present at the postsynaptic compartment. Thus interpreting the data as if it was only in the presynaptic compartment is unfounded. ***I agree, the authors should be more careful about that interpretation. It is very important to be circumspect with interpretation. Figure 1. Panels A-C. This is the best data showing a possible interaction between MET and beta-catenin, although it is possible that there are other proteins that are interposed between MET and beta-catenin. The blot against N-cadherin is an important piece of data, but additional control blots against other proteins could help eliminate the possibility of other intervening proteins. Postsynaptic proteins should also be included - even though the paper is written with a focus on presynaptic events, synaptosomes include postsynaptic complexes that can contain both MET and beta-catenin. ***Yes, additional pre and post synaptic proteins could be included as controls. Panels B and C. The bands from multiple replicates should be quantified and graphed, particularly the N-cadherin (B) and beta-catenin (C) panels. ***I clearly agree with that comment. Panel D. The staining for MET shows a very broad distribution, apparently along dendrites in these images. Is beta-catenin only found in the areas where the PLA gave a reaction? Are these synaptic sites where MET is also found? The PLA data could be showing postsynaptic proximity between MET and beta-catenin, so even though the manuscript is written from a presynaptic perspective the data doesn't necessarily result from presynaptic MET only. Parallel data showing double staining against beta-catenin with the same PLA conditions would be helpful. In the methods, the authors state that spines couldn't be seen based upon the imaging conditions. However, staining against a synaptic marker such as PSD-95 in a third channel to show the relative location of the PLA signal would be very informative. ***The authors must clarify MET and beta-catenin subcellular distribution. An in vivo characterization would be useful. Figure 2. Panel A. This is the sole evidence presented that MET interacts with synapsin, although it doesn't eliminate the possibility that there are other intervening proteins linking MET with synapsin. Other synaptic vesicular proteins should also be stained against, and multiple replicates quantified. Other interaction assays between synapsin and MET should also be included, as this putative interaction forms a significant part of their model. ***Data should be quantified. Panel B. The proximity ligation assay between MET and synapsin should be interpreted much more narrowly. It shows a close proximity of these two proteins, but not necessarily a direct interaction. In addition, if HGF induces synaptic activity (as has been shown previously), synapsin can translocate with vesicles to the active zone near where MET may be located, thus allowing for greater PLA signal that is independent of a direct interaction. ***Again the direct v. indirect issue is raised and needs better experiments to resolve the issue. Figure 3. The western blot against pY142-beta-catenin and the number of clusters in neurons should be quantified. ***I agree. 2) the FLAG-beta-catenin signal appears to be dendritic (an image of a transfected whole neuron or a costain against an axonal or dendritic marker would be revealing), and its presynaptic function is thus unclear; ***Yes, a better discrimination of the pre vs. postsyanptic localization of b-catenin is essential 3) the apparent FLAG beta-catenin stain in dendrites is similar to the MET stain in dendrites (Figure 1), so why is there no PLA signal throughout the dendrite? Are they only colocalized at synaptic sites? ***This is a very good point raised. In general, I would also suggest to present all PLA data with the 2 channels used. 4) This Y142 phosphorylation site regulates the switch for beta-catenin between its adhesive role and its transcriptional role. Wouldn't the Y142F mutant thus cause beta-catenin to remain in an adhesive state with alpha-catenin and thus remain at synapses, independent of a putative interaction with MET? Similarly, phosphorylation of WT beta-catenin would cause it to be released from its interaction with alpha-catenin and thus translocate away from sites near MET. ***Separation between phospho-b-catenin and MET could indeed be mediated by its interaction with alpha-catenin. This however does not change the fact that HGF treatment leads to b-catenin phosphorylation and separation from MET. The authors could discuss this interesting point. 5) Given the limitation of the PLA in showing direct interactions, it would be more convincing if a Co-IP was conducted similar to figure 1 against endogenous proteins showing a loss of interaction upon phosphorylation of beta-catenin. At minimum, the FLAG-beta-catenin should be shown to become phosphorylated upon HGF treatment. ***A control showing that exogenous flag-b-catenin becomes phosphorylated upon HGF stimulation should be added. Figure 5. Why is GFP-synapsin used instead of simply staining against endogenous synapsin? This would eliminate artifacts due to long transfection times and overexpression artifacts. The data on the size of the Syn-GFP clusters suggests that the clusters get larger with treatment with HGF. Larger clusters, simply due to their increased size, will overlap to a greater extent with the abundant puncta of other synaptic markers. ***Endogenous staining of synapsin would be more physiological than Synapsin-GFP expression. Figure 6. This data would be more convincing if done with antibodies against endogenous synapsin. The Syn-GFP forms very large clusters that may or may not mimic endogenous events. It is not clear from the images whether the Syn-GFP and FLAG-beta-catenin are in dendrites or axons. An image of an entire transfected neuron or showing a costain against a dendritic or axonal marker would clarify. The caveat about overexpressing these proteins for an extended amount of time is also applicable to this experiment. ***Endogenous staining of synapsin would be more physiological. Sincerely, Robert Kalb, MD ==== Refs References Abe K , Chisaka O , Van Roy F , Takeichi M (2004 ) Stability of dendritic spines and synaptic contacts is controlled by alpha N-catenin. Nat Neurosci 7 :357 -363 . 10.1038/nn1212 15034585 Abrahams BS , Arking DE , Campbell DB , Mefford HC , Morrow EM , Weiss LA , Menashe I , Wadkins T , Banerjee-Basu S , Packer A (2013 ) SFARI Gene 2.0: a community-driven knowledgebase for the autism spectrum disorders (ASDs). Mol Autism 4 :36 10.1186/2040-2392-4-36 24090431 Arikkath J , Reichardt LF (2008 ) Cadherins and catenins at synapses: roles in synaptogenesis and synaptic plasticity. Trends Neurosci 31 :487 -494 . 10.1016/j.tins.2008.07.001 18684518 Bamji SX , Shimazu K , Kimes N , Huelsken J , Birchmeier W , Lu B , Reichardt LF (2003 ) Role of beta-catenin in synaptic vesicle localization and presynaptic assembly. Neuron 40 :719 -731 . 14622577 Bamji SX , Rico B , Kimes N , Reichardt LF (2006 ) BDNF mobilizes synaptic vesicles and enhances synapse formation by disrupting cadherin-beta-catenin interactions. J Cell Biol 174 :289 -299 . 10.1083/jcb.200601087 16831887 Bateup HS , Johnson CA , Denefrio CL , Saulnier JL , Kornacker K , Sabatini BL (2013 ) Excitatory/inhibitory synaptic imbalance leads to hippocampal hyperexcitability in mouse models of tuberous sclerosis. Neuron 78 :510 -522 . 10.1016/j.neuron.2013.03.017 23664616 Beaudoin GM 3rd, Lee SH , Singh D , Yuan Y , Ng YG , Reichardt LF , Arikkath J (2012 ) Culturing pyramidal neurons from the early postnatal mouse hippocampus and cortex. Nat Protoc 7 :1741 -1754 . 10.1038/nprot.2012.099 22936216 Bhardwaj D , Náger M , Camats J , David M , Benguria A , Dopazo A , Cantí C , Herreros J (2013 ) Chemokines induce axon outgrowth downstream of hepatocyte growth factor and TCF/β-catenin signaling. Front Cell Neurosci 7 :52 . 10.3389/fncel.2013.00052 23641195 Bourgeron T (2009 ) A synaptic trek to autism. Curr Opin Neurobiol 19 :231 -234 . 10.1016/j.conb.2009.06.003 19545994 Bozkaya G , Korhan P , Cokaklı M , Erdal E , Sağol O , Karademir S , Korch C , Atabey N (2012 ) Cooperative interaction of MUC1 with the HGF/c-Met pathway during hepatocarcinogenesis. Mol Cancer 11 :64 . 10.1186/1476-4598-11-64 22962849 Brigidi GS , Bamji SX (2011 ) Cadherin-catenin adhesion complexes at the synapse. Curr Opin Neurobiol 21 :208 -214 . 10.1016/j.conb.2010.12.004 21255999 Burghy CA , Stodola DE , Ruttle PL , Molloy EK , Armstrong JM , Oler JA , Fox ME , Hayes AS , Kalin NH , Essex MJ , Davidson RJ , Birn RM (2012 ) Developmental pathways to amygdala-prefrontal function and internalizing symptoms in adolescence. Nat Neurosci 15 :1736 -1741 . 10.1038/nn.3257 23143517 Campbell DB , Sutcliffe JS , Ebert PJ , Militerni R , Bravaccio C , Trillo S , Elia M , Schneider C , Melmed R , Sacco R , Persico AM , Levitt P (2006 ) A genetic variant that disrupts MET transcription is associated with autism. Proc Natl Acad Sci U S A 103 :16834 -16839 . 10.1073/pnas.0605296103 17053076 Campbell DB , D'Oronzio R , Garbett K , Ebert PJ , Mirnics K , Levitt P , Persico AM (2007 ) Disruption of cerebral cortex MET signaling in autism spectrum disorder. Ann Neurol 62 :243 -250 . 10.1002/ana.21180 17696172 Campbell DB , Li C , Sutcliffe JS , Persico AM , Levitt P (2008 ) Genetic evidence implicating multiple genes in the MET receptor tyrosine kinase pathway in autism spectrum disorder. Autism Res 1 :159 -168 . 10.1002/aur.27 19360663 Crowell B , Lee GH , Nikolaeva I , Dal Pozzo V , D'Arcangelo G (2015 ) Complex neurological phenotype in mutant mice lacking Tsc2 in excitatory neurons of the developing forebrain. Eneuro 2 (6 ). 10.1523/ENEURO.0046-15.2015 26693177 D'Agostino RB (1986 ) Tests for normal distribution. Goodness-of-fit techniques . New York, NY: Macel Decker. pp. 367–413. Dani VS , Chang Q , Maffei A , Turrigiano GG , Jaenisch R , Nelson SB (2005 ) Reduced cortical activity due to a shift in the balance between excitation and inhibition in a mouse model of Rett syndrome. Proc Natl Acad Sci U S A 102 :12560 -12565 . 10.1073/pnas.0506071102 16116096 David MD , Yeramian A , Duñach M , Llovera M , Cantí C , de Herreros AG , Comella JX , Herreros J (2008 ) Signalling by neurotrophins and hepatocyte growth factor regulates axon morphogenesis by differential beta-catenin phosphorylation. J Cell Sci 121 :2718 -2730 . 10.1242/jcs.029660 18664491 DeAngelis T , Morrione A , Baserga R (2010 ) Mutual interaction and reciprocal down-regulation between c-met and insulin receptor substrate-1. J Cell Physiol 224 :658 -663 . 10.1002/jcp.22164 20578241 De Rubeis S , He X , Goldberg AP , Poultney CS , Samocha K , Cicek AE , Kou Y , Liu L , Fromer M , Walker S , Singh T , Klei L , Kosmicki J , Shih-Chen F , Aleksic B , Biscaldi M , Bolton PF , Brownfeld JM , Cai J , Campbell NG . (2014 ) Synaptic, transcriptional and chromatin genes disrupted in autism . Nature 515 :209 -215 .25363760 Eagleson KL , Milner TA , Xie Z , Levitt P (2013 ) Synaptic and extrasynaptic location of the receptor tyrosine kinase met during postnatal development in the mouse neocortex and hippocampus. J Comp Neurol 521 :3241 -3259 . 10.1002/cne.23343 23787772 Eagleson KL , Lane CJ , McFadyen-Ketchum L , Solak S , Wu HH , Levitt P (2016 ) Distinct intracellular signaling mediates C-MET regulation of dendritic growth and synaptogenesis . Dev Neurobiol . Advance online publication. Retrieved August 14, 2016 10.1002/dneu.22382 . Finsterwald C , Martin JL (2011 ) Cellular mechanisms underlying the regulation of dendritic development by hepatocyte growth factor. Eur J Neurosci 34 :1053 -1061 . 10.1111/j.1460-9568.2011.07839.x 21895802 Gao X , Arlotta P , Macklis JD , Chen J (2007 ) Conditional knock-out of β-catenin in postnatal-born dentate gyrus granule neurons results in dendritic malformation. J Neurosci 27 :14317 -14325 . 10.1523/JNEUROSCI.3206-07.2007 18160639 Garber K (2007 ) Neuroscience. Autism's cause may reside in abnormalities at the synapse. Science 317 :190 -191 . 10.1126/science.317.5835.190 17626859 Gatto CL , Broadie K (2010 ) Genetic controls balancing excitatory and inhibitory synaptogenesis in neurodevelopmental disorder models. Front Synaptic Neurosci 2 :4 . 10.3389/fnsyn.2010.00004 21423490 Gogolla N , Leblanc JJ , Quast KB , Südhof TC , Fagiolini M , Hensch TK (2009 ) Common circuit defect of excitatory-inhibitory balance in mouse models of autism. J Neurodev Disord 1 :172 -181 . 10.1007/s11689-009-9023-x 20664807 Gutierrez H , Dolcet X , Tolcos M , Davies A (2004 ) HGF regulates the development of cortical pyramidal dendrites. Development 131 :3717 -3726 . 10.1242/dev.01209 15229174 Hedrick A , Lee Y , Wallace GL , Greenstein D , Clasen L , Giedd JN , Raznahan A (2012 ) Autism risk gene MET variation and cortical thickness in typically developing children and adolescents . Autism Res 5 :434 -439 . 10.1002/aur.1256 23097380 Herynk MH , Tsan R , Radinsky R , Gallick GE (2003 ) Activation of c-Met in colorectal carcinoma cells leads to constitutive association of tyrosine-phosphorylated beta-catenin. Clin Exp Metastasis 20 :291 -300 . 12856716 Jackson PB , Boccuto L , Skinner C , Collins JS , Neri G , Gurrieri F , Schwartz CE (2009 ) Further evidence that the rs1858830 C variant in the promoter region of the MET gene is associated with autistic disorder. Autism Res 2 :232 -236 . 10.1002/aur.87 19681062 Judson MC , Eagleson KL , Wang L , Levitt P (2010 ) Evidence of cell-nonautonomous changes in dendrite and dendritic spine morphology in the met-signaling-deficient mouse forebrain. J Comp Neurol 518 :4463 -4478 . 10.1002/cne.22467 20853516 Judson MC , Eagleson KL , Levitt P (2011 ) A new synaptic player leading to autism risk: Met receptor tyrosine kinase. J Neurodev Disord 3 :282 -292 . 10.1007/s11689-011-9081-8 21509596 Kawas LH , Benoist CC , Harding JW , Wayman GA , Abu-Lail NI (2013 ) Nanoscale mapping of the Met receptor on hippocampal neurons by AFM and confocal microscopy. Nanomedicine 9 :428 -438 . 10.1016/j.nano.2012.08.008 22960190 Lambert N , Wermenbol V , Pichon B , Acosta S , van den Ameele J , Perazzolo C , Messina D , Musumeci MF , Dessars B , De Leener A , Abramowicz M , Vilain C (2014 ) A familial heterozygous null mutation of MET in autism spectrum disorder. Autism Res 7 :617 -622 . 10.1002/aur.1396 24909855 Lee SH , Peng IF , Ng YG , Yanagisawa M , Bamji SX , Elia LP , Balsamo J , Lilien J , Anastasiadis PZ , Ullian EM , Reichardt LF (2008 ) Synapses are regulated by the cytoplasmic tyrosine kinase Fer in a pathway mediated by p120catenin, Fer, SHP-2, and beta-catenin. J Cell Biol 183 :893 -908 . 10.1083/jcb.200807188 19047464 Levitt P , Eagleson KL , Powell EM (2004 ) Regulation of neocortical interneuron development and the implications for neurodevelopmental disorders. Trends Neurosci 27 :400 -406 . 10.1016/j.tins.2004.05.008 15219739 Lim CS , Walikonis RS (2008 ) Hepatocyte growth factor and c-Met promote dendritic maturation during hippocampal neuron differentiation via the Akt pathway. Cell Signal 20 :825 -835 . 10.1016/j.cellsig.2007.12.013 18262389 Lu KV , Chang JP , Parachoniak CA , Pandika MM , Aghi MK , Meyronet D , Isachenko N , Fouse SD , Phillips JJ , Cheresh DA , Park M , Bergers G (2012 ) VEGF inhibits tumor cell invasion and mesenchymal transition through a MET/VEGFR2 complex. Cancer Cell 22 :21 -35 . 10.1016/j.ccr.2012.05.037 22789536 Monga SP , Mars WM , Pediaditakis P , Bell A , Mulé K , Bowen WC , Wang X , Zarnegar R , Michalopoulos GK (2002 ) Hepatocyte growth factor induces Wnt-independent nuclear translocation of beta-catenin after Met-beta-catenin dissociation in hepatocytes. Cancer Res 62 :2064 -2071 . 11929826 Mullins C , Fishell G , Tsien RW (2016 ) Unifying views of autism spectrum disorders: a consideration of autoregulatory feedback loops. Neuron 89 :1131 -1156 . 10.1016/j.neuron.2016.02.017 26985722 Murase S , Mosser E , Schuman EM (2002 ) Depolarization drives beta-catenin into neuronal spines promoting changes in synaptic structure and function. Neuron 35 :91 -105 . 12123611 Nakano M , Takagi N , Takagi K , Funakoshi H , Matsumoto K , Nakamura T , Takeo S (2007 ) Hepatocyte growth factor promotes the number of PSD-95 clusters in young hippocampal neurons. Exp Neurol 207 :195 -202 . 10.1016/j.expneurol.2007.06.007 17678646 Niland S , Ditkowski B , Parrandier D , Roth L , Augustin H , Eble JA (2013 ) Rhodocetin-αβ-induced neuropilin-1-cMet association triggers restructuring of matrix contacts in endothelial cells. Arterioscler Thromb Vasc Biol 33 :544 -554 . 10.1161/ATVBAHA.112.00006 23288161 O’Roak BJ , Vives L , Girirajan S , Karakoc E , Krumm N , Coe BP , Levy R , Ko A , Lee C , Smith JD , Turner EH , Stanaway IB , Vernot B , Malig M , Baker C , Reilly B , Akey JM , Borenstein E , Rieder MJ , Nickerson DA . (2012a ) Sporadic autism exomes reveal a highly interconnected protein network of de novo mutations . Nature 485 :246 -250 .22495309 O’Roak BJ , Vives L , Fu W , Egertson JD , Stanaway IB , Phelps IG , Carvill G , Kumar A , Lee C , Ankenman K , Munson J , Hiatt JB , Turner EH , Levy R , O'Day DR , Krumm N , Coe BP , Martin BK , Borenstein E , Nickerson DA . (2012b ) Multiplex targeted sequencing identifies recurrently mutated genes in autism spectrum disorders . Science 338 :1619 -1622 .23160955 Okuda T , Yu LM , Cingolani LA , Kemler R , Goda Y (2007 ) beta-Catenin regulates excitatory postsynaptic strength at hippocampal synapses. Proc Natl Acad Sci U S A 104 :13479 -13484 . 10.1073/pnas.0702334104 17679699 Peng Y , Huentelman M , Smith C , Qiu S (2013 ) MET receptor tyrosine kinase as an autism genetic risk factor. Int Rev Neurobiol 113 :135 -165 . 10.1016/B978-0-12-418700-9.00005-8 24290385 Peng Y , Lu Z , Li G , Piechowicz M , Anderson M , Uddin Y , Wu J , Qiu S (2016 ) The autism-associated MET receptor tyrosine kinase engages early neuronal growth mechanism and controls glutamatergic circuits development in the forebrain . Mol Psychiatry 21 :925 -935 .26728565 Phillips GR , Huang JK , Wang Y , Tanaka H , Shapiro L , Zhang W , Shan WS , Arndt K , Frank M , Gordon RE , Gawinowicz MA , Zhao Y , Colman DR (2001 ) The presynaptic particle web: ultrastructure, composition, dissolution, and reconstitution. Neuron 32 :63 -77 . 11604139 Piedra J , Miravet S , Castaño J , Pálmer HG , Heisterkamp N , García de Herreros A , Duñach M (2003 ) p120 Catenin-associated Fer and Fyn tyrosine kinases regulate beta-catenin Tyr-142 phosphorylation and beta-catenin-alpha-catenin Interaction. Mol Cell Biol 23 :2287 -2297 . 12640114 Pinto D , Delaby E , Merico D , Barbosa M , Merikangas A , Klei L , Thiruvahindrapuram B , Xu X , Ziman R , Wang Z , Vorstman JA , Thompson A , Regan R , Pilorge M , Pellecchia G , Pagnamenta AT , Oliveira B , Marshall CR , Magalhaes TR , Lowe JK . (2014 ) Convergence of genes and cellular pathways dysregulated in autism spectrum disorders. Am J Hum Genet 94 :677 -694 . 10.1016/j.ajhg.2014.03.018 24768552 Plummer JT , Evgrafov OV , Bergman MY , Friez M , Haiman CA , Levitt P , Aldinger KA (2013 ) Transcriptional regulation of the MET receptor tyrosine kinase gene by MeCP2 and sex-specific expression in autism and Rett syndrome. Transl Psychiatry 3 :e316 . 10.1038/tp.2013.91 24150225 Qiu S , Anderson CT , Levitt P , Shepherd GM (2011 ) Circuit-specific intracortical hyperconnectivity in mice with deletion of the autism-associated Met receptor tyrosine kinase. J Neurosci 31 :5855 -5864 . 10.1523/JNEUROSCI.6569-10.2011 21490227 Qiu S , Lu Z , Levitt P (2014 ) MET receptor tyrosine kinase controls dendritic complexity, spine morphogenesis, and glutamatergic synapse maturation in the hippocampus. J Neurosci 34 :16166 -16179 . 10.1523/JNEUROSCI.2580-14.2014 25471559 Rasola A , Fassetta M , De Bacco F , D'Alessandro L , Gramaglia D , Di Renzo MF , Comoglio PM (2007 ) A positive feedback loop between hepatocyte growth factor receptor and beta-catenin sustains colorectal cancer cell invasive growth. Oncogene 26 :1078 -1087 . 10.1038/sj.onc.1209859 16953230 Reshetnikova G , Troyanovsky S , Rimm DL (2007 ) Definition of a direct extracellular interaction between Met and E-cadherin. Cell Biol Int 31 :366 -373 . 10.1016/j.cellbi.2007.01.022 17336101 Roura S , Miravet S , Piedra J , García de Herreros A , Duñach M (1999 ) Regulation of E-cadherin/catenin association by tyrosine phosphorylation. J Biol Chem 274 :36734 -36740 . 10593980 Rubenstein JL , Merzenich MM (2003 ) Model of autism: increased ratio of excitation/inhibition in key neural systems. Genes Brain Behav 2 :255 -267 . 14606691 Rudie JD , Hernandez LM , Brown JA , Beck-Pancer D , Colich NL , Gorrindo P , Thompson PM , Geschwind DH , Bookheimer SY , Levitt P , Dapretto M (2012 ) Autism-associated promoter variant in MET impacts functional and structural brain networks. Neuron 75 :904 -915 . 10.1016/j.neuron.2012.07.010 22958829 Smyth LA , Brady HJ (2005 ) cMet and Fas receptor interaction inhibits death-inducing signaling complex formation in endothelial cells. Hypertension 46 :100 -106 . 10.1161/01.HYP.0000167991.82153.16 15911745 Strom A , Bonal C , Ashery-Padan R , Hashimoto N , Campos ML , Trumpp A , Noda T , Kido Y , Real FX , Thorel F , Herrera PL (2007 ) Unique mechanisms of growth regulation and tumor suppression upon Apc inactivation in the pancreas. Development 134 :2719 -2725 . 10.1242/dev.02875 17596282 Südhof TC (2008 ) Neuroligins and neurexins link synaptic function to cognitive disease. Nature 455 :903 -911 . 10.1038/nature07456 18923512 Sun Y , Bamji SX (2011 ) β-Pix modulates actin-mediated recruitment of synaptic vesicles to synapses. J Neurosci 31 :17123 -17133 . 10.1523/JNEUROSCI.2359-11.2011 22114281 Sun Y , Aiga M , Yoshida E , Humbert PO , Bamji SX (2009 ) Scribble interacts with beta-catenin to localize synaptic vesicles to synapses. Mol Biol Cell 20 :3390 -3400 . 10.1091/mbc.E08-12-1172 19458197 Tai CY , Mysore SP , Chiu C , Schuman EM (2007 ) Activity-regulated N-cadherin endocytosis. Neuron 54 :771 -785 . 10.1016/j.neuron.2007.05.013 17553425 Thanseem I , Nakamura K , Miyachi T , Toyota T , Yamada S , Tsujii M , Tsuchiya KJ , Anitha A , Iwayama Y , Yamada K , Hattori E , Matsuzaki H , Matsumoto K , Iwata Y , Suzuki K , Suda S , Kawai M , Sugihara G , Takebayashi K , Takei N . (2010 ) Further evidence for the role of MET in autism susceptibility. Neurosci Res 68 :137 -141 . 10.1016/j.neures.2010.06.014 20615438 Tyndall SJ , Walikonis RS (2006 ) The receptor tyrosine kinase Met and its ligand hepatocyte growth factor are clustered at excitatory synapses and can enhance clustering of synaptic proteins. Cell Cycle 5 :1560 -1568 . 10.4161/cc.5.14.2918 16861928 Uchida N , Honjo Y , Johnson KR , Wheelock MJ , Takeichi M (1996 ) The catenin/cadherin adhesion system is localized in synaptic junctions bordering transmitter release zones. J Cell Biol 135 :767 -779 . 8909549 Voineagu I , Wang X , Johnston P , Lowe JK , Tian Y , Horvath S , Mill J , Cantor RM , Blencowe BJ , Geschwind DH (2011 ) Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature 474 :380 -384 . 10.1038/nature10110 21614001 Wang D , Li Z , Messing EM , Wu G (2005 ) The SPRY domain-containing SOCS box protein 1 (SSB-1) interacts with MET and enhances the hepatocyte growth factor-induced Erk-Elk-1-serum response element pathway. J Biol Chem 280 :16393 -16401 . 10.1074/jbc.M413897200 15713673 Wang F , Eagleson KL , Levitt P (2015 ) Positive regulation of neocortical synapse formation by the Plexin-D1 receptor. Brain Res 1616 :157 -165 . 10.1016/j.brainres.2015.05.005 25976775 Xie Z , Li J , Baker J , Eagleson KL , Coba MP , Levitt P (2016 ) Receptor tyrosine kinase MET interactome and neurodevelopmental disorder partners at the developing synapse . Biol Psychiatry . Advance online publication. Retrieved August 15, 2016. doi:10.1016/j.biopsych.2016.02.022. Yu X , Malenka RC (2003 ) Beta-catenin is critical for dendritic morphogenesis. Nat Neurosci 6 :1169 -1177 . 10.1038/nn1132 14528308 Yu X , Malenka RC (2004 ) Multiple functions for the cadherin/catenin complex during neuronal development. Neuropharmacology 47 :779 -786 . 10.1016/j.neuropharm.2004.07.031 15458849 Zeng G , Apte U , Micsenyi A , Bell A , Monga SP (2006 ) Tyrosine residues 654 and 670 in beta-catenin are crucial in regulation of Met-beta-catenin interactions. Exp Cell Res 312 :3620 -3630 . 10.1016/j.yexcr.2006.08.003 16952352 Zheng L , Baumann U , Reymond JL (2004 ) An efficient one-step site-directed and site-saturation mutagenesis protocol. Nucleic Acids Res 32 :e115 . 10.1093/nar/gnh110 15304544 Zhou X , Xu Y , Wang J , Zhou H , Liu X , Ayub Q , Wang X , Tyler-Smith C , Wu L , Xue Y (2011 ) Replication of the association of a MET variant with autism in a Chinese Han population. PloS One 6 :e27428 . 10.1371/journal.pone.0027428 22110649 Zoghbi HY (2003 ) Postnatal neurodevelopmental disorders: meeting at the synapse? Science 302 :826 -830 . 10.1126/science.1089071 14593168 Zoghbi HY , Bear MF (2012 ) Synaptic dysfunction in neurodevelopmental disorders associated with autism and intellectual disabilities. Cold Spring Harb Perspect Biol 4 :a009886 . 10.1101/cshperspect.a009886 22258914

PMC005xxxxxx/PMC5002984.txt

==== Front eNeuroeNeuroeneuroeneuroeNeuroeNeuro2373-2822Society for Neuroscience 10.1523/ENEURO.0148-16.2016eN-NWR-0148-162New ResearchDevelopmentADP Ribosylation Factor 6 Regulates Neuronal Migration in the Developing Cerebral Cortex through FIP3/Arfophilin-1-dependent Endosomal Trafficking of N-cadherin The Arf6-FIP3 Pathway in Neuronal MigrationHara Yoshinobu 1Fukaya Masahiro 1http://orcid.org/0000-0002-2757-363XHayashi Kanehiro 2http://orcid.org/0000-0002-8650-4474Kawauchi Takeshi 34http://orcid.org/0000-0003-1864-9425Nakajima Kazunori 2Sakagami Hiroyuki 11 Department of Anatomy, Kitasato University School of Medicine, Sagamihara, Kanagawa 252-0374, Japan2 Department of Anatomy, Keio University School of Medicine, Tokyo 160-8582, Japan3 Laboratory of Molecular Life Science, Institute of Biomedical Research and Innovation, Kobe 650-0047, Japan4 Department of Physiology, Keio University School of Medicine, Tokyo 160-8582, JapanThe authors declare no competing financial interests. Author contributions: Y.H. and H.S. designed the study and wrote the paper; Y.H. performed experiments and analysed the data; M.F. raised antibodies and discussed the data; K.H., T.K., and K.N. provided technical advice on in utero electroporation and time-lapse imaging, and valuable comments. This work was supported by a Grant-in-Aid for Scientific Research (B) Grant Number 15H04672 from the Japan Society for the Promotion of Science (H.S.), Parents Association Grant of Kitasato University School of Medicine (Y.H.), and The Science Research Promotion Fund (Y.H.). Correspondence should be addressed to Hiroyuki Sakagami, Department of Anatomy, Kitasato University School of Medicine, 1-15-1, Kitasato, Sagamihara, Kanagawa 252-0374, Japan. E-mail: sakagami@med.kitasato-u.ac.jp4 8 2016 29 8 2016 Jul-Aug 2016 3 4 ENEURO.0148-16.20166 6 2016 29 7 2016 2 8 2016 Copyright © 2016 Hara et al.2016Hara et al.This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International, which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed.Abstract During neural development, endosomal trafficking controls cell shape and motility through the polarized transport of membrane proteins related to cell–cell and cell–extracellular matrix interactions. ADP ribosylation factor 6 (Arf6) is a critical small GTPase that regulates membrane trafficking between the plasma membrane and endosomes. We herein demonstrated that the knockdown of endogenous Arf6 in mouse cerebral cortices led to impaired neuronal migration in the intermediate zone and cytoplasmic retention of N-cadherin and syntaxin12 in migrating neurons. Rescue experiments with separation-of-function Arf6 mutants identified Rab11 family-interacting protein 3 (FIP3)/Arfophilin-1, a dual effector for Arf6 and Rab11, as a downstream effector of Arf6 in migrating neurons. The knockdown of FIP3 led to impaired neuronal migration in the intermediate zone and cytoplasmic retention of N-cadherin in migrating neurons, similar to that of Arf6, which could be rescued by the coexpression of wild-type FIP3 but not FIP3 mutants lacking the binding site for Arf6 or Rab11. These results suggest that Arf6 regulates cortical neuronal migration in the intermediate zone through the FIP3-dependent endosomal trafficking. Arf6layer formationmembrane traffickingmigrationN-cadherinreceptor recycling cover-dateJuly/August 2016 ==== Body Significance Statement Endosomal trafficking is a critical mechanism that regulates cell motility and shape. During cortical layer formation, migrating neurons undergo dynamic changes in cell motility and shape, but regulatory mechanisms that control endosomal trafficking in migrating neurons have just begun to be elucidated. Here, we demonstrate that ADP ribosylation factor 6 (Arf6) small GTPase regulates neuronal migration in the intermediate zone through FIP3/Arfophilin-1-dependent endosomal transport of N-cadherin. The present findings indicate the importance of the Arf6-FIP3 pathway in the cortical layer formation and the existence of multiple small GTPases that regulate the surface N-cadherin expression in migrating neurons. Introduction Neuronal migration is essential for the construction of brain structure and contributes to the evolutionary expansion of the number of neurons and complexity of neural circuits in the mammalian brain (Olson and Walsh, 2002; LoTurco and Bai, 2006; Rakic, 2007). Among various brain regions, the cerebral cortex has been extensively studied for neuronal migration due to its characteristic birthdate-dependent laminated structures and the feasibility of gene transfer by in utero electroporation (Inoue and Krumlauf, 2001; Saito and Nakatsuji, 2001; Tabata and Nakajima, 2001). Recent imaging analyses identified distinct migratory modes of radial migration during cortical development: multipolar migration, locomotion, and terminal translocation (Nadarajah et al., 2001; Tabata and Nakajima, 2003; Nishimura et al., 2010; Sekine et al., 2011; Ohshima, 2014). Among these, multipolar migration is highlighted by its vulnerability, which causes neurodevelopmental disorders, including periventricular nodular heterotopia, subcortical band heterotopia, and double cortex syndrome (Gressens, 2000; Kato and Dobyns, 2003; Lu and Sheen, 2005; LoTurco and Bai, 2006; Cooper, 2014). During multipolar migration, neurons unsteadily move in the subventricular zone (SVZ) and intermediate zone (IZ) with their processes repeatedly extending and retracting, and establish cell polarity by forming an axon and reorienting intracellular organelles, such as the centrosome and Golgi apparatus (de Anda et al., 2010; Jossin, 2011; Sakakibara et al., 2014). In the upper IZ, multipolar neurons initiate contact with radial glial fibers, transform into a bipolar shape, and enter the locomotion mode (Tabata and Nakajima, 2003; Nishimura et al., 2010). Therefore, to complete multipolar-to-bipolar transition, multipolar cells may sense some directional cues through cell–cell and cell–extracellular matrix interactions. Consistent with this idea, recent evidence suggests that the surface expression of N-cadherin, a neural transmembrane cell adhesion molecule, on multipolar cells at an appropriate level and location is required for the multipolar-to-bipolar transition and regulated by endosomal trafficking mediated by Rap1 and Rab small GTPases (Kawauchi et al., 2010; Jossin and Cooper, 2011). The ADP ribosylation factor (Arf) family is also a critical small GTPase for endosomal trafficking, and is grouped into three classes based on structural similarities: Arf1, Arf2, and Arf3 in class I; Arf4 and Arf5 in class II; and Arf6 in class III (D'Souza-Schorey and Chavrier, 2006; Gillingham and Munro, 2007; Donaldson and Jackson, 2011). Of these, Arf6 is present at the plasma membrane and a subpopulation of endosomes, where it regulates not only actin cytoskeleton remodeling but also endocytosis and/or the recycling of various receptors, including E-cadherin (Palacios et al., 2001, 2002), integrin (Powelka et al., 2004; Dunphy et al., 2006), transferrin receptor (D'Souza-Schorey et al., 1995), G-protein-coupled receptors (Claing et al., 2001; Houndolo et al., 2005; Macia et al., 2012), and major histocompatibility complex class I molecules (Klein et al., 2006). Accumulating evidence implicates Arf6 as a critical regulator of cell shape and motility in various cell types. For example, the activation of Arf6 leads to the disassembly of adherens junctions through the internalization of E-cadherin, leading to changes in cell shape and motility, a process referred to as epithelial–mesenchymal transition, during wound healing and cancer invasion (Palacios et al., 2001, 2002; Luton et al., 2004). Arf6 also regulates the cell motility of MDA-MB231 breast cancer cells through the recycling of integrin β to the cell surface (Powelka et al., 2004), suggesting the importance of the Arf6-mediated polarized transport of cell adhesion molecules, such as cadherin and integrin, during cell migration and cancer invasion. Regarding the role of Arf6 in the developing cerebral cortex, Falace et al. (2014) provided the first evidence for the functional involvement of Arf6 in cortical neuronal migration. However, our understanding of how Arf6 regulates neuronal migration is still incomplete. Here, we demonstrate that Arf6 regulates neuronal migration in the IZ through the interaction with Rab11 family-interacting protein 3 (FIP3), also known as Arfophilin-1 or Eferin, a dual effector of Arf6 and Rab11 for endosomal trafficking during cytokinesis (Fielding et al., 2005; Horgan and McCaffrey, 2009). We further demonstrate the functional involvement of the Arf6-FIP3 pathway in the trafficking of N-cadherin in cortical neurons. These results underscore the role of Arf6-FIP3-dependent endosomal trafficking in cortical neuronal migration. Materials and Methods Animals Pregnant ICR (Institute for Cancer Research) mice were purchased from Charles River Japan. All experimental procedures in this study were approved by the Animal Experimentation and Ethics Committee of Kitasato University School of Medicine. Plasmid construction The wild-type and mutant cDNAs for Arf6 (Arf6T44N and Arf6Q67L) were kind gifts from Dr. Kazuhisa Nakayama at Kyoto University. The cDNAs for FIP3 and Rab11A were amplified by a polymerase chain reaction (PCR) from the mouse embryonic day (E) 17 brain cDNA library using the following primers supplemented with the EcoRI restriction site (underlined): 5′-GAA TTC ATG GAG CTG TGC CAG CCG ACC TCC C-3′ and 5′-CTA CTT GAC CTC TAG GAT GGA TGG GTT GGT C-3′ for FIP3; 5′-GAA TTC ATG GGC ACC CGC GAC GAG TA-3′ and 5′-TTA GAT GTT CTG ACA GCA CTG CAC CTT TGG-3′ for Rab11A. The PCR fragments were subcloned into the pGEM-T easy vector (Promega). FIP4 cDNA was purchased from DNAFORM (clone ID: B530005J18). The shRNA-resistant Arf6 mutants (Arf6WT, Arf6iSW, Arf6W168L/L169V, and Arf6N48I), and FIP3 mutants (FIP3WT, FIP3ΔABD, and FIP3ΔRBD) were made using the PrimeSTAR mutagenesis basal kit (Takara Bio) according to the manufacturer’s protocols. The resultant cDNAs were digested with EcoRI, and subcloned into the pCAGGS expression vector with the FLAG-tag sequence (Niwa et al., 1991; Sakagami et al., 2005) or pNeuroD-IRES-GFP vector, which had been provided by Dr. Franck Polleux at Columbia University via Addgene (plasmid #61403; Guerrier et al., 2009). Expression vectors carrying membrane-targeted and nucleus-targeted EGFP (EGFP-Fyn and EGFP-NLS) were constructed by subcloning the cDNAs for EGFP, which had been supplemented at C termini with a membrane-targeted signal from Fyn (MGCVQCKDKEATKLTEF) or nuclear localization signal (PKKKRKVEDA), respectively, into pCAGGS-loxP-polyA-loxP (Floxp; Shitamukai et al., 2011). pCAGGS-Cre-recombinase was a kind gift from Dr. Fumio Matsuzaki at the RIKEN Center for Developmental Biology. To construct shRNA vectors for Arf6 and FIP3, oligonucleotides targeting the coding sequence [5′-GCA CCG CAT TAT CAA TGA CCG-3′ for Arf6 KD, 5′-TCA CCT CAG GAG CAA GTA CCT-3′ for Arf6 KD scramble (Scr; Control), 5′-AAG CAG CTA GAA CAT CTA C-3′ for FIP3 KD#1, 5′-AAC CGT AAC CTG AAG GAG C-3′ for FIP3 KD#2, 5′-AAG GCT AAG GTC AAC CAC G-3′ for FIP3 KD#2 Scr (Control)] and their respective complementary sequences were aligned in tandem with a hairpin loop sequence (5′-TTC AAG AGA-3′), and inserted into the pmU6pro vector containing the mouse U6 small nuclear RNA promoter (Yu et al., 2002; Kawauchi et al., 2006). The target sequences were designed using the GenScript siRNA target finder (GenScript USA). All constructs using in utero electroporation were purified with PureLink HiPure Plasmid Filter Purification Kits (Life Technologies). Regarding bacterial expression vectors for N-cadherin, the N-terminal extracellular region (amino acids 160–723) of mouse N-cadherin was amplified by PCR and subcloned into pGEX4T-2 (GE Healthcare) or pMAL-SXN, a modified pMAL-2c (New England Biolabs; Fukaya et al., 2011). In situ hybridization The riboprobes for in situ hybridization against Arf6, FIP3, and FIP4 were synthesized using the Digoxigenin RNA labeling Mix (Roche), and T7 and SP6 RNA polymerases (Promega). In situ hybridization was performed as described previously (Hara et al., 2013). Cell culture and transfection To examine the efficiency of shRNAs, cultured cortical neurons were prepared as described previously (Beaudoin et al., 2012) and transfected before plating using Amaxa Nucleofector 2D (Lonza Group) according to the manufacturer’s protocol. Three days after plating, neurons were subjected to an immunoblot analysis to assess shRNA efficiency and internalization/recycling assay for N-cadherin. Antibodies To produce an anti-Arf6 antibody, a 10 aa peptide (LTWLTSNYKS) corresponding to amino acids 166–175 of mouse Arf6 was conjugated to the Keyhole limpet hemocyanin (KLH) and used as an antigen for immunization and affinity purification. Regarding an anti-N-cadherin antibody, the N-terminal extracellular region (amino acids 160–723) of mouse N-cadherin was expressed as fusion proteins of glutathione S-transferase (GST) and maltose-binding protein (MBP) in Escherichia coli BL21 (DE3; Stratagene) in the presence of 0.3 mm isopropyl-β-D-thiogalactopyranoside at 25ºC overnight and purified with glutathione-Sepharose 4B (GE Healthcare) and amylose-resin (New England Biolabs), respectively. Rabbits and guinea pigs were subcutaneously immunized with the KLH-conjugated peptide or GST-N-cadherin fusion protein five times at 2 week intervals. Sera were affinity-purified using cyanogen bromide-activated Sepharose 4B (GE Healthcare) conjugated with the Arf6 peptide or MBP-N-cadherin fusion protein. The specificities of antibodies were characterized by an immunoblot analysis. Immunoblot analysis Lysates (10 μg) were subjected to an immunoblot analysis with the following primary antibodies at a final concentration of 0.5 μg/ml: guinea pig anti-Arf6 IgG, guinea pig anti-FIP3 IgG (Yazaki et al., 2014), mouse anti-N-cadherin IgG (32/N-cadherin, BD Transduction Laboratories), and mouse anti-α-tubulin IgG (DM1A, Sigma-Aldrich). Immunoreactive bands were detected using ECL Plus kit (Thermo Fisher Scientific) and LAS-4000 miniluminescent image analyzer (GE Healthcare). The quantification of intensities for immunoreactive bands was performed using ImageJ software (National Institutes of Health). In utero electroporation In utero electroporation was performed as described previously (Kawauchi et al., 2010; Hara et al., 2013). Plasmids were mixed under the following conditions in phosphate-buffered saline (PBS): overexpression experiments, 3 μg/μl of the plasmid; knockdown experiments, 1 μg/μl of the shRNA plasmid; rescue experiments, 1 μg/μl of the shRNA plasmid and 0.1 μg/μl of pCAGGD-Arf6-FLAG or 0.5 μg/μl of pCAGGS-FLAG-FIP3 in combination with 0.5 μg/μl of pCAGGS-mCherry or EGFP. Electroporation (35 V, 450 ms, three pulses) was performed for mouse embryos at E14.5 through the uterus with forceps-type electrodes using a square electroporator (CUY-21 Edit, Bex). The embryos were fixed at E15.5, E16.5, E17.5, or postnatal day (P) 0. To label proliferating cells, pregnant mice at E15.5 were intraperitoneally administrated 10 mg/ml of 5-bromo-2-deoxyuridine (BrdU, Roche) dissolved in 0.9% saline solution (140 mg/kg body weight) three times at 20 minute intervals before being killed. In time-lapse imaging, a mixture of the pCAGGS-Floxp plasmids encoding EGFP, EGFP-Fyn, and EGFP-NLS were electroporated together with shRNA and pCAGGS-Cre-recombinase. Immunohistology Immunohistochemistry and BrdU staining were performed as described previously (Hara et al., 2013). The following primary antibodies were used at a final concentration of 1 μg/ml: guinea pig anti-Arf6 IgG, goat anti-MAP2 IgG (MAP2-Go-Af, Frontier Institute), goat anti-BLBP IgG (Yamada et al., 2000), rabbit anti-EEA1 IgG (Fukaya et al., 2014), rabbit anti-syntaxin12 (STX12) IgG (Hara et al., 2013), goat anti-LAMP2 polyclonal IgG (sc-8100, Santa Cruz Biotechnology), guinea pig or rabbit anti-mCherry IgG (Hara et al., 2013), rabbit anti-EGFP IgG (Sakagami et al., 2005), mouse anti-polysialated-neural cell adhesion molecule (anti-PSA-NCAM) IgM (Seki, 2002), rabbit anti-Rab11 IgG (71-5300, Life Technologies), guinea pig anti-FIP3 IgG (Yazaki et al., 2014), and guinea pig anti-N-cadherin IgG. The immunoreaction was visualized using species-specific secondary antibodies conjugated with Alexa488, Alexa594, or Alexa647 (Invitrogen). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (Roche). Immunoreactions were examined with a confocal laser microscope (LSM 710, Carl Zeiss). Quantitative analysis The dorsolateral region of the cerebral cortex was analyzed for all electroporation experiments. One representative brain section was selected from each sample, and EGFP-positive or mCherry-positive cells in each cortical layer, which was identified based on cell density visualized by nuclear staining with DAPI, were counted in a blind manner using ImageJ software (National Institutes of Health). Data were collected from ≥3 animals for each condition as indicated in graphs. The fraction of EGFP-positive or mCherry-positive cells in each cortical layer of the brain transfected with the indicated plasmids was compared with that in the equivalent layer under the control conditions. Statistical analyses were performed using the paired t test for direct comparison, or one-way ANOVA with post hoc Tukey–Kramer’s test for multiple comparisons. Time-lapse imaging Time-lapse imaging was performed as described previously (Tabata and Nakajima, 2003). Embryos were electroporated at E14.5 and collected at E17.5. Brains were embedded with 3% low melting temperature agarose gel and sliced into 200-μm-thick sections with a microslicer (VT1000S, Leica). The slices were placed on an insert membrane (Millipore) and maintained in glass-bottom dishes (Iwaki) filled with culture medium. Recording was carried out using a confocal laser microscope (LSM 710, Carl Zeiss) and stage top incubator (40% O2, 5% CO2; ZILCS-H2; Tokai Hit). Images were collected every 15 min for 20 h and analyzed using ImageJ software (National Institutes of Health). Internalization and recycling assays Internalization and recycling assays were performed as described previously (Tai et al., 2007). Three days after transfection and plating, cultured cortical neurons were treated with 100 μg/ml leupeptin at 37°C for 1 h. Cell-surface proteins were labeled with 0.3 mg/ml of Sulfo-NHS-SS-Biotin (Thermo Fisher Scientific) in PBS containing Mg2+ and Ca2+ (PBS-MC) on ice for 12 min. Excess biotin was quenched by Tris buffer (25 mm TrisHCl, pH 8.0, 133 mm NaCl, 10 mm KCl). Neurons were then rinsed twice with maintenance medium (Neurobasal medium, 1× B27 supplement, 2 mm L-glutamine, penicillin/streptomycin) and incubated at 37°C for various periods as indicated in Figure 4D. At the end of each time point, neurons were chilled to 4°C by washing with ice-cold PBS-MC, and biotins exposed on the cell surface were removed by rinsing twice for 15 min with ice-cold 50 mm glutathione solution (50 mm reduced glutathione, 75 mm NaCl, 10 mm EDTA, 1% BSA, pH 8.0). Neurons were then solubilized with lysis buffer (0.1 m phosphate buffer, pH 7.5, 150 mm NaCl, 1% Triton X-100, 0.1% SDS) containing a mixture of protease inhibitors (Sigma-Aldrich) and phenylmethylsulfonyl fluoride. Internalized biotinylated membrane proteins were precipitated with NeutrAvidin resin (Thermo Fisher Scientific) overnight at 4°C and subjected to an immunoblot analysis with anti-N-cadherin IgG (BD Transduction Laboratories). In the recycling assay, neurons were biotinylated and incubated for 90 min at 37°C, and biotinylated proteins remaining on the cell surface were removed by rinsing with glutathione solution. Neurons were then incubated for the various periods indicated in Figure 4E. At the end of each time point, neurons were treated with glutathione solution and lysed. Internalized biotinylated membrane proteins were precipitated with NeutrAvidin resin (Thermo Fisher Scientific) and subjected to an immunoblot analysis. The significance of differences was determined by the Mann-Whitney U test. Results Arf6 is required for cortical layer formation In a previous study, overexpression of constitutively active Arf6 mutant Arf6Q67L in cortical progenitor cells under a ubiquitous promoter was shown to disturb cortical layer formation (Falace et al., 2014). However, it remains unclear whether this effect may be attributed to the overactivation of Arf6 in neurons, glial cells, or both. In order to understand the role of Arf6 in neuronal migration in more detail, we first examined the effects of overexpression of Arf6 mutants specifically in postmitotic neurons under the NeuroD promoter (Guerrier et al., 2009). At E17.5, three days after in utero electroporation, significantly more neurons overexpressing either Arf6Q67L or GDP-locked inactive mutant Arf6T44N were retained in the IZ than control neurons (Fig. 1A; Control, 41.2 ± 1.9%; Arf6T44N, 61.3 ± 1.5%, P < 0.01; Arf6Q67L, 61.6 ± 2.3%, P < 0.01), suggesting the importance of the proper functioning of the GDP/GTP cycling of Arf6 in migrating neurons for cortical layer formation. Figure 1. Arf6 is required for cortical neuronal migration. A, Representative micrographs of E17.5 cerebral cortices electroporated with the pNeuroD plasmids carrying Mock, Arf6Q67L, or Arf6T44N in combination with pCAGGS-EGFP at E14.5 (n = 5 embryos). B, Immunoblot analysis of cultured cortical neurons transfected with Arf6 shRNA (Arf6 KD) or its scramble shRNA (Control) for 3 d. The numbers indicate the percentage of endogenous Arf6 expression level relative to α-tubulin. C, Representative micrographs of E17.5 cerebral cortices electroporated with the pmU6pro plasmids carrying scramble sequence (Control) or Arf6 shRNA (Arf6 KD) in combination with pCAGGS-EGFP at E14.5 (n = 4 embryos). D, Representative micrographs of P0 cerebral cortices electroporated with Control, Arf6 KD, or Arf6 KD and Arf6WT plasmids in combination with pCAGGS-EGFP at E14.5 (n = 4 embryos). Bottom graphs in A, C, and D show the quantification of the distribution of EGFP-positive cells in cortical layers. Data were presented as mean ± SEM and statistically analyzed using one-way ANOVA followed by Tukey–Kramer’s tests in A and D (*P < 0.05, **P < 0.01), and by unpaired t test in C (*P < 0.05, **P < 0.005, ***P < 0.0005). n in the graph indicates the number of embryos examined. Scale bars: A, C, D, 200 μm. Next, we examined the effect of the knockdown of endogenous Arf6 on cortical layer formation using shRNA. In the first step to determine the efficiency of shRNA for Arf6, cultured cortical neurons from E14.5 cerebral cortices were transfected with shRNA and harvested 3 d after transfection. The immunoblot analysis revealed that the expression of Arf6 shRNA significantly decreased endogenous Arf6 in cultured cortical neurons by ∼40% compared with that of control scramble shRNA (Fig. 1B). Having established the efficiency of Arf6 shRNA, we introduced Arf6 shRNA together with EGFP into ventricular progenitor cells in the dorsal pallium at E14.5 by in utero electroporation and analyzed the distribution of EGFP-positive cells in the cerebral cortex 3 and 5 d later. At E17.5, 41.7 ± 3.7% of control neurons were migrating in the cortical plate (CP) with a bipolar shape, whereas 62.5 ± 0.7% of Arf6-knockdown neurons were still positioned in the IZ with only 21.2 ± 1.8% in the CP (Fig. 1C), suggesting that the knockdown of Arf6 delayed the exit of migrating neurons from the IZ to the CP. At P0, the knockdown of Arf6 led to a pronounced migration defect with a significantly smaller proportion of neurons reaching the superficial layers II–IV than control neurons (Fig. 1D; Control, 54.9 ± 1.2%; Arf6 knockdown, 38.3 ± 1.6%, P < 0.01, n = 4 embryos). In a rescue experiment, the coexpression of shRNA-resistant wild-type Arf6 (Arf6WT), in which four target nucleotides were mutated without amino acid substitutions, with Arf6 shRNA restored the proportion of cells that reached the superficial layers at P0 to an indistinguishable level from the control (Fig. 1D), suggesting the target specificity of Arf6 shRNA. Since Arf6 is known to regulate cytokinesis during mitosis (Schweitzer and D'Souza-Schorey, 2002, 2005), we examined the possibility that the knockdown of Arf6 could affect the cell proliferation in the ventricular zone (VZ)/SVZ. Pregnant mice, which had been subjected to in utero electroporation with pCAGGS-EGFP at E14.5, were intraperitoneally administrated BrdU 3 times at E15.5 and embryos were fixed 1 h later. BrdU immunostaining failed to detect any apparent differences in the number or position of BrdU-incorporated transfected cells between Arf6-knockdown and control cells in the VZ (Fig. 2A; Control, 36.9 ± 0.1%; Arf6 knockdown, 33.6 ± 0.1%, P = 0.61, n = 3 embryos). Further immunostaining of cerebral cortices, which had been electroporated with control or Arf6 shRNA, with neuronal differentiation markers failed to detect any apparent differences in the expression of PSA-NCAM, a marker for immature neurons, in the IZ at E17.5 (Fig. 2B; Control, 93.6 ± 0.01%; Arf6 knockdown, 89.5 ± 0.1%, P = 0.21, n = 3 embryos), or Cux1, a marker for cortical superficial layers, in the CP at P0 (Fig. 2C; Control, 90.7 ± 0.04%; Arf6 knockdown, 90.5 ± 0.1%, P = 0.95, n = 3 embryos) between control and Arf6-knockdown cells, excluding the possibility that the knockdown of Arf6 affected neuronal differentiation. Together, these results suggest that not only the proper GDP/GTP cycling but also the endogenous expression of Arf6 is required for the cortical layer formation. Figure 2. Effects of the knockdown of Arf6 on cell proliferation and neuronal differentiation. A, Representative micrographs showing the distribution of BrdU-positive cells in the VZ. Embryos were electroporated with the indicated shRNAs and EGFP at E14.5 and killed at E15.5 1 h after the intraperitoneal administration of BrdU to pregnant mice. Arrowheads show BrdU-incorporated mCherry-positive cells in the VZ. Note no apparent differences in the cell number or position of BrdU-positive transfected cells between Control and Arf6 knockdown (Arf6 KD). B, Representative micrographs showing the expression of PSA-NCAM in control and Arf6-KD neurons in the IZ at E17.5. C, Representative micrographs showing that the expression of Cux1 in control and Arf6-KD neurons in the superficial cortical layers at P0. Scale bars: A, 25 μm; B, 50 μm; C, 100 μm. Arf6 regulates neuronal migration in the IZ To determine how Arf6 regulates the properties of neuronal migration, we performed the time-lapse imaging analysis of neuronal migration using acute cortical slices at E17.5, 3 d after in utero electroporation. To monitor the cell behavior at the single-cell level, shRNA-transfected neurons were sparsely labeled by a mixture of cytoplasm-targeted, nucleus-targeted, and membrane-targeted EGFP using a Cre-loxP clonal expression system (Shitamukai et al., 2011). In the CP, control and Arf6-knockdown neurons radially migrated toward the pial surface with a typical bipolar shape, as shown in Figure 3A. The morphology and migration speed of radially migrating neurons were indistinguishable between control and Arf6-knockdown neurons (Fig. 3A,B ; Control, 15.6 ± 0.4 μm/h; Arf6 KD, 15.7 ± 0.5 μm/h, P = 0.9203). By contrast, in the IZ, the migration speed toward the CP was significantly slower in Arf6-knockdown neurons than in the control neurons (Fig. 3C,D ; Control, 8.5 ± 0.3 μm/h; Arf6 KD, 6.8 ± 0.3 μm/h, P < 0.0005). Furthermore, a single-cell tracking analysis revealed that directional migration of Arf6-knockdown neurons toward the CP in the IZ was disorganized compared with that of the control neurons (Fig. 3E). These results suggest that Arf6 regulates neuronal migration in the IZ but not in the CP during cortical layer formation. Figure 3. Arf6 is important for neuronal migration in the IZ. A, C, Representative time-lapse images of transfected migrating neurons in the CP and IZ. Embryos were electroporated with the indicated shRNA in combination with pCAGGS-Cre recombinase and pCAGGS-Floxp carrying EGFP, EGFP-NLS, and EGFP-Fyn at E14.5, and brains were subjected to slice culture and time-lapse observations at E17.5. Arrowheads indicate transfected neurons. B, D, Quantification of the migration speed of transfected neurons in the CP (B, n = 115 cells from 3 embryos) and IZ [D, Control, n = 41 cells from 3 embryos; Arf6 knockdown (Arf6 KD), n = 43 cells from 3 embryos]. E, Tracking analysis of migration of transfected neurons in the IZ. The graphs show the trajectory of individual migrating neurons (Control, 41 cells; Arf6 KD, 43 cells from 3 embryos) for 10–13 h. F, The orientation of the Golgi apparatus in migrating neurons in the IZ. Arrowheads indicate the Golgi apparatus labeled by GM130 in transfected neurons in the IZ. G, Quantification of the proportion of cells in the IZ with the Golgi apparatus facing the CP. Note the disorientation of Golgi apparatus in Arf6-KD neurons (n = 3 embryos). H, Representative micrographs showing the morphology of transfected neurons in the IZ at E16.5. Brains were electroporated with the indicated shRNA plasmids in combination with pCAGGS-EGFP at E14.5 and immunostained with an anti-EGFP antibody. I, Quantification of the number of primary processes. Note the increase in primary processes in migrating neurons transfected with Arf6-KD plasmid (Control, n = 195 cells from 3 embryos; Arf6 KD, n = 208 cells from 3 embryos). Data were presented as mean ± SD, and statistically analyzed by unpaired t test (*P < 0.05, ***P < 0.0005). Scale bars: A, C, 50 μm; F, H, 20 μm. Since the position of intracellular organelles, such as the centrosome and Golgi apparatus is reoriented during the multipolar-to-bipolar transition in the IZ (Jossin, 2011; Sakakibara et al., 2014), we investigated the position of the Golgi apparatus in multipolar neurons in the IZ at E16.5 by immunostaining with an anti-GM130 antibody, a marker for the cis-Golgi. The knockdown of Arf6 significantly decreased the proportion of multipolar neurons with the Golgi apparatus oriented toward the CP compared with the control scramble shRNA (Fig. 3F,G ; Control, 54.1 ± 4.2%; Arf6 knockdown, 37.7 ± 1.4%, P < 0.05). Furthermore, the number of primary processes increased in Arf6-knockdown neurons (Fig. 3H,I ; Control, 6.7 ± 0.2; Arf6 knockdown, 9.0 ± 0.6, P < 0.0005). These results suggest that Arf6 regulates the establishment of cell polarity in migrating neurons during the multipolar-to-bipolar transition in the IZ. Arf6 regulates the transport of recycling endosomes in migrating neurons To obtain a mechanistic insight into the role of Arf6 in neuronal migration, we first examined the subcellular localization of Arf6 in migrating neurons using a novel anti-Arf6 antibody raised against the C-terminal 10 aa sequence, which is divergent among the Arf family. In the immunoblot analysis, the anti-Arf6 antibody specifically detected a 17-kDa immunoreactive band corresponding to endogenous human Arf6 in HeLa cell lysates and an additional band in the lysates of HeLa cells expressing FLAG-tagged mouse Arf6, but not other Arf isoforms (Fig. 4A). Having confirmed the specificity of the antibody, we performed double immunofluorescent staining of migrating neurons at E17.5, which had been transfected with mCherry at E14.5, with antibodies against Arf6 and various endosomal markers: EEA1 for early endosomes, STX12 and Rab11 for recycling endosomes, and LAMP2 for late endosomes and lysosomes. The immunofluorescene for Arf6 appeared as numerous cytoplasmic fine puncta in migrating neurons and partially colocalized with EEA1, STX12, and Rab11, but not with LAMP2 (Fig. 4B). These results suggest the occurrence of Arf6 at subpopulations of early and recycling endosomes in migrating neurons in the IZ. Figure 4. Arf6 localizes at endosomes and regulates the distribution of syntaxin12-positive endosomes in migrating neurons. A, Immunoblot analysis showing the specificity of rabbit anti-Arf6 antibody. The lysates of the mouse cerebral cortex (Cx) at E14.5 (10 μg) and HeLa cells transfected with the indicated FLAG-tagged Arf plasmids were immunoblotted with antibodies against Arf6 or FLAG. The positions and sizes of the molecular weight markers are indicated on the left. B, Representative micrographs showing endosomal localization of Arf6 in migrating neurons in the IZ at E17.5. Coronal sections of the E17.5 cerebral cortex, which had been electroporated with pCAGGS-mCherry at E14.5, were subjected to immunofluorescent staining with antibodies against Arf6 (magenta), the indicated markers (green), and mCherry (red). Nuclei were counterstained with DAPI (blue). Insets show the high-magnification views of boxed areas. C, D, Representative micrographs showing the subcellular localization of EEA1-positive (C) and syntaxin12 (STX12)-positive (D) endosomes in migrating neurons transfected with Control or Arf6 knockdown (Arf6 KD) plasmid in the IZ at E17.5. Arrowheads in D indicate STX12-positive puncta in transfected neurons. E, Quantification of the relative immunofluorescence intensity for STX12 in the perinuclear region. The relative immunofluorescence intensity was calculated by normalizing the immunofluorescence intensity for perinuclear STX12 in transfected neurons to that in the surrounding untransfected neurons. Note the significant increase in perinuclear STX12 in Arf6-KD neurons (Control, n = 155 cells; Arf6 KD, n = 239 cells). Data collected from three embryos in each condition were presented as mean ± SD and statistically analyzed with unpaired t test (*** P < 0.0001). Scale bars: B–D, 10 μm. This subcellular localization of Arf6 prompted us to examine the effect of the knockdown of Arf6 on the endosomal system in migrating neurons by immunostaining with endosomal markers. Arf6-knockdown and control neurons in the IZ at E17.5 exhibited indistinguishable intracellular distribution of EEA1-positive early endosomes (Fig. 4C). On the other hand, Arf6-knockdown neurons, compared with control neurons, had more STX12-positive endosomes accumulated in the perinuclear region (Fig. 4D). A quantitative analysis revealed that the ratio of the fluorescence intensity for perinuclear STX12 in transfected cells to that in the surrounding untransfected cells in the same sections was significantly higher in Arf6-knockdown neurons than in control neurons (Fig. 4E; Control, 1.0 ± 0.2; Arf6 KD, 1.3 ± 0.2, P < 0.0001). These results suggest that the knockdown of Arf6 disturbs the transport of a subpopulation of recycling endosomes in migrating neurons in the IZ. Arf6 regulates the recycling of N-cadherin in cortical neurons Recent evidence highlighted the functional importance of endosomal trafficking of cell adhesion molecules, such as N-cadherin and integrin β, in neuronal migration (Jossin, 2011; Kawauchi, 2012; Sekine et al., 2014). Particularly, Rap1 and Rab small GTPases were shown to regulate cortical neuronal migration through the endosomal trafficking of N-cadherin (Kawauchi et al., 2010; Jossin and Cooper, 2011). In addition, Arf6 is implicated in cell shape and motility during cancer invasion and wound healing through the regulation of surface E-cadherin expression (Palacios et al., 2001, 2002). Therefore, we hypothesized that Arf6 may be functionally linked to the endosomal trafficking of N-cadherin in migrating neurons. To test this hypothesis, we examined the subcellular localization of N-cadherin in migrating neurons by immunostaining with an anti-N-cadherin antibody raised against its extracellular domain (Fig. 5A). In the control neurons migrating in the IZ, the immunofluorescence for N-cadherin was largely distributed along the plasma membrane (Fig. 5B). By contrast, the knockdown of Arf6 resulted in the accumulation of the immunofluorescence for N-cadherin in the cell bodies of transfected neurons (Fig. 5B). A quantitative analysis confirmed that the immunofluorescence for cytoplasmic N-cadherin was significantly higher in Arf6-knockdown neurons than in control neurons (Fig. 5C; Control, 1.1 ± 0.2; Arf6 KD, 1.3 ± 0.2, P < 0.0001). Figure 5. Arf6 regulates endosomal trafficking of N-cadherin in cortical neurons. A, Immunoblot analysis showing the specificity of guinea pig anti-N-cadherin antibody. The lysates of the adult mouse brain (10 µg) and HeLa cells transfected with or without C-terminally FLAG-tagged N-cadherin were immunoblotted with antibodies against N-cadherin or FLAG. B, Representative micrographs showing the subcellular localization of N-cadherin in migrating neurons transfected with Control or Arf6 knockdown (Arf6 KD)in the IZ at E17.5. Arrowheads indicate the localization of N-cadherin in transfected neurons. C, Quantification of the relative immunofluorescence intensity for cytoplasmic N-cadherin. The relative intensity was calculated by normalizing the immunofluorescence intensity for cytoplasmic N-cadherin in transfected neurons to that in the surrounding untransfected neurons (Control, n = 137 cells from 3 embryos; Arf6 KD, 174 cells from 3 embryos). Note the significant increase in cytoplasmic N-cadherin in Arf6-KD neurons. D, Internalization assay for N-cadherin in cultured cortical neurons. A bottom graph shows the quantification of biotinylated N-cadherin internalized in neurons transfected with Control (open bars) or Arf6 KD(black bars; Control, n = 5 independent experiments; Arf6 KD, n = 5). E, Recycling assay for N-cadherin in cultured cortical neurons. A bottom graph shows the quantification of biotinylated N-cadherin retained inside neurons transfected with Control (open bars) or Arf6 KD (black bars; Control, n = 5 independent experiments; Arf6 KD, n = 6). Data were presented as mean ± SD in C and mean ± SEM in D and E, and statistically analyzed by unpaired t test in C (***P < 0.0001), and Mann-Whitney U test in D and E (*P < 0.05). GSH, Glutathione. Scale bar: B, 20 μm. To further verify the involvement of Arf6 in the trafficking of N-cadherin in cortical neurons, we investigated the effects of the knockdown of Arf6 on the time course of the internalization and recycling of N-cadherin in cultured cortical neurons from E14.5 mouse embryos. In the internalization assay, cultured cortical neurons transfected with control or Arf6 shRNAs were subjected to surface labeling with cleavable biotin, followed by internalization in the presence of leupeptin to inhibit lysosomal proteolysis. At various time points, the remaining surface biotin was removed and internalized biotinylated proteins were precipitated by avidin-conjugated resin. The immunoblot analysis with an anti-N-cadherin antibody revealed that the time course and ratio of N-cadherin internalization were indistinguishable between control and Arf6-knockdown neurons (Fig. 5D). In the recycling assay, after biotinylated neurons were subjected to internalization at 37ºC for 90 min, the biotin recycled to the cell surface was removed at various time points, and the biotinylated proteins retained in the cell body were precipitated by avidin-conjugated resin and subjected to an immunoblot analysis. In control neurons, internalized biotinylated N-cadherin was recycled back to the plasma membrane at 45 min, whereas ∼20% of internalized N-cadherin was still retained in the cell bodies of Arf6-knockdown neurons, even at 60 min (Fig. 5E; 45 min: Control, 2.0 ± 1.7%; Arf6 knockdown, 20.6 ± 3.4%, P = 0.011; 60 min: Control, 2.3 ± 1.7%; Arf6 knockdown, 16.5 ± 4.8%, P = 0.028). These results suggest that Arf6 regulates the recycling of N-cadherin in cortical neurons. FIP3 is a candidate for Arf6 downstream effector in neuronal migration. To elucidate the Arf6 downstream pathways that regulate neuronal migration in more detail, we examined whether the impaired neuronal migration phenotype caused by Arf6 shRNA could be rescued by the coexpression of separation-of-function Arf6 mutants that interfere with the interaction with specific downstream effectors: Arf6iSW, which blocks the interaction with JIP3/4 and vezatin by mutating Asn-45, Thr-53, Val-57, Lys-58, and Asn-60 residues in the interswitch region (Montagnac et al., 2009; Sanda et al., 2010; Suzuki et al., 2010); Arf6N48I, which blocks the Arf6-dependent activation of phospholipase D by mutating Asn-48 to Ile (Vitale et al., 2005); Arf6W168L/L169V, which blocks the interaction with class II FIPs by mutating Trp-168 and Leu-169 to Leu and Val, respectively (Schonteich et al., 2007). At P0, the cotransfection of either Arf6iSW or Arf6N48I with Arf6 shRNA significantly increased the proportion of EGFP-positive cells in the superficial cortical layer compared with Arf6 knockdown (Fig. 6A,B ; Arf6 knockdown and Mock, 34.3 ± 5.0%; Arf6 knockdown and Arf6iSW, 66.7 ± 2.8%, P < 0.05; Arf6 knockdown and Arf6N48I, 68.8 ± 5.8%, P < 0.05). On the other hand, the cotransfection of Arf6W168L/L169V with Arf6 shRNA did not restore the delay in neuronal migration with only 38.4 ± 3.1% of EGFP-positive neurons in the superficial cortical layers (Fig. 6A,B ). These results suggest that Arf6 may regulate neuronal migration through the interaction with class II FIPs. Figure 6. Arf6 regulates neuronal migration through the interaction with class II FIPs. A, Representative micrographs showing rescue experiments of impaired neuronal migration caused by the knockdown of Arf6 with separation-of-function Arf6 mutants. Brains were electroporated with the indicated pCAGGS plasmids carrying Mock, Arf6iSW, Arf6N48I, Arf6W168L/L169V in combination with Arf6 shRNA and pCAGGS-EGFP at E14.5, and fixed at P0. Bottom graphs show the quantification of the distribution of EGFP-positive cells in cortical layers. B, Comparison of the percentage of EGFP-positive cells transfected with the indicated plasmids in the superficial layers II–IV at P0. Data were presented as mean ± SEM and statistically analyzed using one-way ANOVA followed by post hoc Tukey-Kramer’s test (*P < 0.05). n in the graph indicates the number of embryos examined. Scale bar, 200 μm. The class II FIP family includes two structurally related isoforms, FIP3 and FIP4, and is known to function as an effector of Arf6-dependent and Rab11-dependent membrane trafficking during cytokinesis (Fielding et al., 2005; Horgan and McCaffrey, 2009). To understand the role of class II FIPs in the cortical development, we first examined the gene expression of FIP3, FIP4, and Arf6 in the developing cerebral cortex by a nonradioactive in situ hybridization analysis. In the cerebral cortex at E14.5 and E16.5, FIP3 and Arf6 mRNAs were already detectable in the VZ and maintained throughout the cerebral cortex, while FIP4 mRNA was expressed preferentially in the CP with a negligible level in the VZ, SVZ, and IZ (Fig. 7A). Since the expression pattern of FIP3 traced well with that of Arf6, we hereafter focused on FIP3 as a candidate for Arf6 downstream effector in neuronal migration. In the immunoblot analysis of the developing cerebral cortex, a single immunoreactive band for FIP3 was already detectable at E14.5 and the expression level was gradually increased after birth (Fig. 7B). Immunofluorescent staining of the cerebral cortex at E17.5, which had been transfected with mCherry at E14.5 by in utero electroporation, revealed that the immunofluorescence for FIP3 appeared as numerous puncta that partially colocalized with Arf6, EEA1, and STX12 in migrating neurons in the IZ (Fig. 7C). These results suggest the occurrence of FIP3 at subpopulations of early and recycling endosomes with Arf6 in migrating neurons in the IZ. Figure 7. Expression of FIP3 in the developing cerebral cortex. A, Representative micrographs showing the gene expression of Arf6, FIP3, and FIP4 in the developing cerebral cortex. Coronal sections of mouse cerebral cortices at E14.5 and E16.5 were subjected to nonradioactive in situ hybridization analysis for Arf6, FIP3, and FIP4. Right, Magnified views of the boxed region in the middle panel. Note the widespread gene expression of FIP3 and Arf6 in the VZ, IZ, and CP in contrast to the preferential gene expression of FIP4 in the CP. B, Immunoblot analysis of developing cerebral cortices with anti-FIP3 and anti-α-tubulin antibodies. Note the gradual increase in FIP3 during cortical development. The positions and sizes of the molecular weight markers are indicated on the right. C, Representative micrographs showing the subcellular localization of FIP3 in migrating neurons in the IZ at E17.5. Coronal sections of the cerebral cortex, which had been electroporated with pCAGGS-mCherry at E14.5, were subjected to immunofluorescent staining with antibodies against FIP3 (magenta), the indicated markers (green), and mCherry (red). Nuclei were counterstained with DAPI (blue). Insets show the high-magnification views of boxed areas. Scale bars: A, left, middle, 500 μm; A, right, 10 μm; C, 20 μm. The Arf6-FIP3 pathway is required for neuronal migration in the IZ. To determine the functional involvement of FIP3 in neuronal migration, we designed two shRNAs for FIP3 (FIP3 KD#1 and KD#2; Yazaki et al., 2014). The immunoblot analysis revealed that the expression of each of these FIP3 shRNAs effectively decreased endogenous FIP3 in cultured cortical neurons (Fig. 8A). Having established the efficiency of FIP3 shRNAs, mouse embryos were subjected to in utero electroporation with FIP3 shRNAs at E14.5. Three days later, the proportion of FIP3 KD#2-transfected neurons in the IZ was significantly higher than that of control neurons (Control, 34.7 ± 1.9%; FIP3 KD#2, 57.6 ± 4.0%, P < 0.01), and only 14.2 ± 1.0% of FIP3 KD#2-transfected neurons migrated into the CP (Fig. 8B,F ). Disturbed cortical layer formation by the knockdown of FIP3 was still evident at P0 (Fig. 8C). FIP3 KD#1 also had the same effect on cortical layer formation as FIP3 KD#2 (Fig. 8C). In addition, the rescue experiment revealed that the cotransfection of shRNA-resistant wild-type FIP3 (FIP3WT) with FIP3 KD#2 partially improved the cortical migration phenotype caused by FIP3 KD#2 (Fig. 8E,F ). Figure 8. FIP3 regulates neuronal migration in the IZ through the interaction with Arf6 and Rab11. A, Immunoblot analysis showing the efficiency of FIP3 shRNAs (FIP3 KD#1 and #2) in primary cortical neurons. The numbers indicate the percentage of endogenous FIP3 expression level relative to α-tubulin. Note that the expression of FIP3 KD#1 and #2 decreased endogenous FIP3 by 37.0 and 65.6%, respectively. B, C, E, Representative images of coronal sections of E17.5 (Band E) and P0 (C) cerebral cortices electroporated with Control, FIP3 knockdown (B and C), or FIP3 KD#2 and FIP3WT, FIP3ΔABD, or FIP3ΔRBD (E) in combination with pCAGGS-EGFP at E14.5. The bottom graphs show the quantification of the distribution of EGFP-positive neurons in each layer. n in the graph indicates the number of embryos examined. D, Immunoprecipitation. HeLa cells were transfected with the indicated combinations of plasmids and subjected to immunoprecipitation with anti-FLAG affinity gel. Immunoprecipitates and total lysates were subjected to an immunoblot analysis with anti-HA or anti-FLAG antibodies. Note the lack of the ability of FIP3ΔABD and FIP3ΔRBD to interact with Arf6 and Rab11, respectively. WB, Western blot. F, Comparison of the percentage of EGFP-positive cells transfected with the indicated plasmids in the CP at E17.5. Note the ability of FIP3WT but not FIP3ΔABD or FIP3ΔRBD to partially rescue the disturbed cortical layer formation caused by the knockdown of FIP3. G, Representative time-lapse images of transfected neurons in the IZ. Embryos were electroporated with the indicated shRNAs, pCAGGS-Cre recombinase, and pCAGGS-Floxp carrying EGFP, EGFP-NLS, and EGFP-Fyn at E14.5, and subjected to time-lapse observations at E17.5. Arrowheads indicate transfected neurons. H, Quantification of the migration speed of transfected neurons in the IZ (n = 40 cells from 3 embryos). Note the decrease in the migration speed of FIP3 KD#2-transfected neurons compared with that of control neurons. I, Tracking analysis of migration of transfected neurons in the IZ. The graphs show the trajectory of migrating neurons for 10–13 h (n = 40 cells from 3 embryos). Data were presented as mean ± SEM (B, C, E, F) and mean ± SD (H), and statistically analyzed using one-way ANOVA followed by post hoc Tukey-Kramer’s test in B (vs control) and E (vs FIP3 KD#2 in B; *P < 0.05, ** P < 0.01), and unpaired t test in C and H(vs control; *P < 0.05, ** P < 0.005, ***P < 0.0005). The total numbers of examined animals were indicated in the graphs in B, C,and E. Scale bars: B, C, E, 200 μm; F, 50 μm. Since human FIP3 was shown to interact simultaneously with Arf6 and Rab11 through distinct binding regions in the C-terminal region (Fielding et al., 2005; Shiba et al., 2006), we prepared mouse FIP3 mutants that lacked the ability to interact with Arf6 or Rab11 by deleting the respective binding regions [amino acids 943–1048 for Arf6 (FIP3ΔABD); 1069–1092 for Rab11 (FIP3ΔRBD)]. A coimmunoprecipitation analysis confirmed that FIP3WT and FIP3ΔRBD, but not FIP3ΔABD, coprecipitated with Arf6WT, whereas FIP3WT and FIP3ΔABD, but not FIP3ΔRBD, coprecipitated with Rab11A in a GTP-dependent manner (Fig. 8D). The cotransfection of either FIP3ΔABD or FIP3ΔRBD mutants with FIP3 KD#2 did not restore the retention of neurons transfected with FIP3 KD#2 in the IZ (Fig. 8E,F ). The time-lapse imaging analysis of the acute cortical slice revealed that the migration speed of FIP3-knockdown neurons in the IZ was significantly slower than that of control neurons (Fig. 8G,H ; Control, 10.3 ± 0.3 μm/h; FIP3 KD#2, 6.7 ± 0.2 μm/h, P < 0.0001). Furthermore, a single-cell tracking analysis revealed that the directional migration of FIP3 KD#2-transfected neurons toward the CP in the IZ was disorganized compared with that of the control neurons (Fig. 8I). These results suggest that FIP3 regulates the neuronal migration in the IZ through the interaction with Arf6 and Rab11 during cortical layer formation. FIP3 regulates the trafficking N-cadherin through the interaction with Arf6 and Rab11 in migrating neurons Finally, to determine whether FIP3 regulates the trafficking of N-cadherin downstream of Arf6 in migrating neurons, we immunohistochemically examined the subcellular localization of N-cadherin in the migrating neurons, which had been transfected with FIP3 KD#2 alone or in combination with FIP3WT, FIP3ΔABD, or FIP3ΔRBD. As observed in Arf6-knockdown neurons, FIP3-knockdown neurons exhibited significantly more cytoplasmic N-cadherin-immunoreactive puncta than control neurons (Fig. 9A; Control, 1.1 ± 0.2; FIP3 KD#2, 1.2 ± 0.2, P < 0.01). Furthermore, the cytoplasmic accumulation of N-cadherin caused by the knockdown of FIP3 was partially rescued by the cotransfection of FIP3WT, but not FIP3ΔABD or FIP3ΔRBD (Fig. 9A,B; FIP3 KD#2 and FIP3WT, 1.1 ± 0.1, P < 0.01 vs FIP3 KD#2; FIP3 KD#2 and FIP3ΔABD, 1.2 ± 0.2, P < 0.01 vs Control; FIP3 KD#2 and FIP3ΔRBD, 1.5 ± 0.2, P < 0.01 vs Control). These results suggest that FIP3 mediates the Arf6-dependent and Rab11-dependent trafficking of N-cadherin in migrating neurons. Figure 9. FIP3 regulates the trafficking of N-cadherin in migrating neurons through the interaction with Arf6 and Rab11. A, Representative micrographs showing the effect of the expression of FIP3 shRNA with or without the indicated FIP3 mutants on the subcellular localization of N-cadherin in migrating neurons in the IZ at E17.5. Arrowheads indicate the localization of N-cadherin in transfected neurons. B, Quantification of the relative immunofluorescence intensity for cytoplasmic N-cadherin. The relative intensity was calculated by normalizing the immunofluorescence intensity for cytoplasmic N-cadherin in transfected neurons to that in the surrounding untransfected neurons. Note the cytoplasmic accumulation of N-cadherin in neurons transfected with FIP3 KD#2, which was rescued by the coexpression of FIP3WT, but not FIP3ΔABD or FIP3ΔRBD (Control, n = 62 cells; FIP3 KD#2, n = 136 cells; FIP3 KD#2 and FIP3WT, n = 96 cells; FIP3 KD#2 and FIP3ΔABD, n = 114 cells; FIP3 KD#2 and FIP3ΔRBD, n = 117 cells). Data collected from three embryos in each condition were presented as mean ± SD and statistically analyzed using one-way ANOVA followed by post hoc Tukey–Kramer’s test (*P < 0.05, **P < 0.01). Scale bar, 20 μm. Discussion Recent evidence implicates endosomal trafficking mediated by small GTPases, such as Rap1 and Rabs, as a critical mechanism regulating neuronal migration through the polarized delivery of membrane lipids and proteins, including N-cadherin (Kawauchi et al., 2010; Franco et al., 2011; Jossin and Cooper, 2011) and integrin (Marchetti et al., 2010; Sekine et al., 2012). Arf6 is a small GTPase that regulates a variety of neuronal functions, including neuronal migration (Falace et al., 2014), the formation of axons (Hernández-Deviez et al., 2002), dendrites, and spines (Miyazaki et al., 2005; Raemaekers et al., 2012; Kim et al., 2015), and synaptic plasticity (Scholz et al., 2010) through the endosomal trafficking and actin cytoskeleton reorganization. In this study, we provide evidence suggesting that the Arf6-FIP3 pathway regulates neuronal migration in the IZ through the endosomal trafficking of N-cadherin. Falace et al. (2014) recently provided the first evidence for the functional involvement of Arf6 in neuronal migration and proposed that the suppression of Arf6 activation through the interaction with TBC1D24, a protein encoded by the gene associated with familiar infantile myoclonic epilepsy, is important for proper cortical neuronal migration. This conclusion was based on their findings that overexpression of a constitutively active Arf6 mutant (Arf6Q67L) but not a constitutively inactive mutant (Arf6T27N) disturbed the cortical layer formation and that the impaired neuronal migration phenotype caused by the knockdown of TBC1D24 could be rescued by the coexpression of Arf6T27N. The present study extended the previous findings by showing that overexpression of not only Arf6Q67L but also another constitutively inactive mutant (Arf6T44N) in postmitotic neurons under the NeuroD promoter disturbed cortical layer formation. The discrepancy of phenotypes caused by inactive Arf6 mutants between previous and present studies is likely to stem from differences in experimental conditions, including inactive Arf6 mutants (Arf6T27N vs Arf6T44N), promoters of expression vectors (chick β-actin vs NeuroD), and animal species (rat vs mouse). Concerning the biochemical properties of inactive Arf6 mutants, Marcia et al. (2004) reported that Arf6T44N is stably bound to GDP and targeted to physiologically proper membrane compartments, whereas Arf6T27N tends to form aggregates in a nucleotide-free form when exogenously overexpressed in vivo. Thus, it is possible that Arf6T27N may have given a limited dominant-negative effect on the Arf6 pathway in migrating neurons due to its improper subcellular targeting. Therefore, we propose that proper GDP/GTP cycling of Arf6 at appropriate locations and timings in migrating neurons is required for neuronal migration. N-cadherin is known to regulate various steps in cortical development, including the maintenance of adherens junctions and polarity of radial glial progenitors in the VZ (Kadowaki et al., 2007; Gil-Sanz et al., 2014), radial glia-independent somal migration (Gil-Sanz et al., 2013), multipolar migration (Jossin and Cooper, 2011), radial glia-dependent locomotion (Kawauchi et al., 2010; Shikanai et al., 2011), neurite extension, and synaptogenesis (Bamji, 2005; Tai et al., 2007; Takeichi, 2007; Ito and Takeichi, 2009; Brigidi and Bamji, 2011). In the present study, we provided two lines of evidence suggesting the functional involvement of Arf6 in the endosomal recycling of N-cadherin to the plasma membrane in multipolar neurons. First, the knockdown of Arf6 resulted in the cytoplasmic accumulation of N-cadherin and STX12-positive recycling endosomes in migrating neurons. Second, the knockdown of Arf6 impaired the recycling of N-cadherin to the plasma membrane, but not internalization from the plasma membrane, in cultured cortical neurons. In polarized epithelial cells, there is unambiguous evidence for a functional relationship between Arf6 and E-cadherin: the activation of Arf6 facilitates the internalization of E-cadherin from adherens junctions, thereby triggering epithelial–mesenchymal transition, during wound healing and cancer invasion (Palacios et al., 2001, 2002; Luton et al., 2004). Although the directions of Arf6-mediated membrane transport are reversed between E-cadherin in epithelial cells and N-cadherin in multipolar neurons, Arf6 is suggested to affect cell motility and shape by regulating either the inward or outward transport of cadherin through the interaction with various effectors, depending on the cell type and context. Small GTPases, such as Rap1 and Rab, are highlighted as critical regulatory pathways for surface N-cadherin expression through endosomal trafficking in neuronal migration (Kawauchi et al., 2010; Franco et al., 2011; Jossin and Cooper, 2011; Ye et al., 2014). During multipolar migration, Rap1 promotes surface N-cadherin expression in migrating neurons and their establishment of radial orientation upstream of Rac1, Cdc42, and RalA/B (Jossin and Cooper, 2011). On the other hand, Rab5 and Rab11 regulate neuronal migration through N-cadherin endocytosis and recycling (Kawauchi et al., 2010), respectively. The migratory phenotype caused by the knockdown of Arf6 is quite similar to those caused by the inhibition of Rap1 and knockdown of Rab11. Therefore, it is plausible that N-cadherin trafficking may be fine-tuned by multiple small GTPases, including Arf6, Rabs, and Rap1, in multipolar neurons along a sequential and/or parallel cascade. However, it should be noted that we failed to rescue the Arf6-knockdown phenotype of cortical layer formation by the coexpression of N-cadherin (our unpublished observations). Since either overexpression or knockdown of N-cadherin has an adverse effect on neuronal migration (Shikanai et al., 2011), our failure to rescue the Arf6-knockdown phenotype by N-cadherin may have been due to the narrow window of the optimal expression level of N-cadherin to rescue the phenotype. Alternatively, it is also possible that other cargo molecules transported with N-cadherin in endosomes are primarily responsible for Arf6-dependent neuronal migration. Therefore, a proteomic analysis to identify other cargo molecules in Arf6-positive endosomes in migrating neurons will be necessary to comprehensively understand how Arf6-dependent endosomal trafficking contributes to neuronal motility and morphology during neuronal migration. By rescue experiments with separation-of-function Arf6 mutants that interfere with specific downstream pathways, we identified the class II FIP family, particularly FIP3, as a possible Arf6 effector for neuronal migration in the IZ. This result was further strengthened by the following lines of evidence. First, in situ hybridization analysis demonstrated that the spatiotemporal expression pattern of FIP3 rather than FIP4 paralleled well with that of Arf6 in the developing cerebral cortex. Further immunofluorescent analysis revealed that FIP3 partially colocalized at a subpopulation of endosomes with Arf6 and STX12 in migrating neurons. Second, the knockdown of FIP3 impaired neuronal migration in the IZ with reduced speed and disorganized directionality, which is consistent with the Arf6-knockdown phenotypes. Furthermore, the disturbed cortical layer formation caused by the knockdown of FIP3 was restored by the coexpression of shRNA-resistant wild-type FIP3, but not FIP3ΔABD or FIP3ΔRBD. Finally, the knockdown of FIP3 also resulted in the cytoplasmic accumulation of N-cadherin in migrating neurons, similar to that of Arf6, which was restored by the coexpression of wild-type FIP3, but not FIP3ΔABD or FIP3ΔRBD. Taken together, it is reasonable to conclude that FIP3 regulates neuronal migration downstream of Arf6 and Rab11. Our knowledge on the physiological roles of class II FIPs largely comes from previous studies on their functional role in cytokinesis, the final stage of cell division (Fielding et al., 2005; Wilson et al., 2005; Takahashi et al., 2011; Schiel et al., 2012). FIP3 is required for the completion of cytokinesis through Arf6-dependent and Rab11-dependent endosomal transport of cargo molecules, such as p50RhoGAP and SCAMP2/3, to the midbody (Schiel et al., 2012), where FIP3-positive endosomes are tethered and fused by the interaction with the Exocyst complex via the Arf6-FIP3-Rab11 ternary complex and with the Centralspindlin complex via FIP3 (Simon et al., 2008; Takahashi et al., 2011). Arf6 also functions as an acceptor for endosomes transporting to the cleavage furrow and midbody during cytokinesis through the interaction with various components of cytokinesis, such as FIP3/4 (Fielding et al., 2005), the Sec10 subunit of the Exocyst complex (Fielding et al., 2005), mitotic kinesis-like protein MKLP1 (also known as kinesin-6 or KIF23) of the Centralspindlin complex (Makyio et al., 2012), JIP3/4 (Montagnac et al., 2009), and the serologically defined colon antigen-3 (Sakagami et al., 2016). Interestingly, recent evidence suggests that various molecular components are conserved between cytokinesis and neuronal migration. For example, the knockdown of MKLP1 or blocking of Exo70, a subunit of the Exocyst complex, was shown to result in similar disturbance of cortical layer formation to the knockdown of Arf6 (Letinic et al., 2009; Falnikar et al., 2013). Therefore, by analogy to the molecular mechanism by which the Arf6-FIP3 pathway regulates cytokinesis, we hypothesize that the activation of Arf6 may facilitate the tethering and docking of Rab11/FIP3 endosomes containing N-cadherin at the leading edge of multipolar neurons through the interaction with the Exocyst complex and MKLP1, thereby establishing the neuronal polarity and interaction with radial glial processes during the multipolar-to-bipolar transition. In summary, the present study demonstrates that Arf6 regulates neuronal migration in the IZ through the FIP3-dependent trafficking of endosomes containing N-cadherin during cortical layer formation. The switching of Arf6 between GDP-bound and GTP-bound states is precisely regulated by multiple guanine nucleotide exchange factors and GTPase-activating proteins, which are abundantly expressed in the brain and possess multiple functional domains to interact with various proteins and lipids (Gillingham and Munro, 2007; Sakagami et al., 2008; Donaldson and Jackson, 2011). Further studies will be required to determine how the GDP/GTP cycling of Arf6 is fine-tuned in migrating neurons by these upstream regulatory factors in response to external and internal stimuli. We thank Drs. Kazuhisa Nakayama (Kyoto University), Fumio Matsuzaki (RIKEN Center for Developmental Biology), Jun-ichi Miyazaki (Osaka University), David L. Turner (University of Michigan), and Franck Polleux (Columbia University) for plasmids; and Dr. Tatsunori Seki (Tokyo Medical University) for anti-PSA-NCAM IgM. Synthesis The decision was a result of the Reviewing Editor Darcy Kelley and the peer reviewers coming together and discussing their recommendations until a consensus was reached. A fact-based synthesis statement explaining their decision and outlining what is needed to prepare a revision is listed below. The following reviewers agreed to reveal their identity: Jonathan Cooper, Jean-Bernard Manent. Through utilizing a combination of knockdown and overexpression constructs, and a variety of technical approaches, the authors quite convincingly show that Arf6 regulates multipolar migration via FIP3-dependent endosome trafficking of N-Cadherin. These are interesting novel findings that further expanded the current knowledge on the molecular pathways controlling neocortical histogenesis. Both reviewers agree that the manuscript is now acceptable for publication, pending some very minor revision (Reviewer #2). Reviewer #1: I thank the authors for the revisions and am satisfied that the results are statistically significant. For the authors' information, regarding Fig 5 D, E, I questioned the significance of the internalization/recycling assay not because of the number of sample groups (n=2) but because of the number of time points (n=4). Unless there is a clear hypothesis before an experiment is done, there may be an unconscious bias to select time points at which there is a statistically significant effect and ignore the time points at which there is no effect, essentially "cherry picking" the data to get P(more than symbol) 0.05. However, since FIP3 has a known function in the exocyst it is reasonable to focus on the 45' and 60' time points. For this reason I agree with the authors that the results show a role for Arf6 in recycling internalized N cadherin back to the surface. Reviewer #2 In this revised version, the authors have addressed most of the concerns, and their manuscript is very nicely improved. However, one minor concern regarding cell proliferation and differentiation analyses remains to be addressed. Although the authors claim that they found no difference in terms of number or position of BrdU+ cells (pg 17, ln20), it appears that this is only 'qualitative' since no quantification is provided. PSA-NCAM and Cux1 immunostainings for investigating any differentiation defects were not quantified either. The authors may wish to include quantifications, or at least to mention how many sections and animals showing no 'qualitative' difference were examined. Also, minor cosmetic changes for improving figure clarity could be made: 1) Tracing cell contours in the highlighted cell in Fig 4B, 7C ; 2) Improving dot plots in Fig9B in order to make dots smaller and lighter (mean and SEM could hardly be seen in the present version of the fig). Last, a few typos could be found: 'Samll GTPases' pg29, ln13; 'knockdwon of FIP3' pg31, ln2; 'ADDIN EN.CITE' pg32, ln17. ==== Refs References Bamji SX (2005 ) Cadherins: actin with the cytoskeleton to form synapses. Neuron 47 :175 –178 . 10.1016/j.neuron.2005.06.024 16039559 Beaudoin GM , 3rd, Lee SH , Singh D , Yuan Y , Ng YG , Reichardt LF , Arikkath J (2012 ) Culturing pyramidal neurons from the early postnatal mouse hippocampus and cortex. Nat Protoc 7 :1741 –1754 . 10.1038/nprot.2012.099 22936216 Brigidi GS , Bamji SX (2011 ) Cadherin-catenin adhesion complexes at the synapse. Curr Opin Neurobiol 21 :208 –214 . 10.1016/j.conb.2010.12.004 21255999 Claing A , Chen W , Miller WE , Vitale N , Moss J , Premont RT , Lefkowitz RJ (2001 ) beta-Arrestin-mediated ADP-ribosylation factor 6 activation and beta 2-adrenergic receptor endocytosis. J Biol Chem 276 :42509 –42513 . 10.1074/jbc.M108399200 11533043 Cooper JA (2014 ) Molecules and mechanisms that regulate multipolar migration in the intermediate zone. Front Cell Neurosci 8 :386 . 10.3389/fncel.2014.00386 25452716 D'Souza-Schorey C , Chavrier P (2006 ) ARF proteins: roles in membrane traffic and beyond . Nat Rev Mol Cell Biol 7 :347 –358 . 10.1038/nrm1910 16633337 D'Souza-Schorey C , Li G , Colombo MI , Stahl PD (1995 ) A regulatory role for ARF6 in receptor-mediated endocytosis . Science 267 :1175 –1178 . 10.1126/science.7855600 7855600 de Anda FC , Meletis K , Ge X , Rei D , Tsai LH (2010 ) Centrosome motility is essential for initial axon formation in the neocortex. J Neurosci 30 :10391 –10406 . 10.1523/JNEUROSCI.0381-10.2010 20685982 Donaldson JG , Jackson CL (2011 ) ARF family G proteins and their regulators: roles in membrane transport, development and disease. Nat Rev Mol Cell Biol 12 :362 –375 . 10.1038/nrm3117 21587297 Dunphy JL , Moravec R , Ly K , Lasell TK , Melancon P , Casanova JE (2006 ) The Arf6 GEF GEP100/BRAG2 regulates cell adhesion by controlling endocytosis of beta1 integrins. Curr Biol 16 :315 –320 . 10.1016/j.cub.2005.12.032 16461286 Falace A , Buhler E , Fadda M , Watrin F , Lippiello P , Pallesi-Pocachard E , Baldelli P , Benfenati F , Zara F , Represa A , Fassio A , Cardoso C (2014 ) TBC1D24 regulates neuronal migration and maturation through modulation of the ARF6-dependent pathway . Proc Natl Acad Sci USA 111 :2337 –2342 . 10.1073/pnas.1316294111 24469796 Falnikar A , Tole S , Liu M , Liu JS , Baas PW (2013 ) Polarity in migrating neurons is related to a mechanism analogous to cytokinesis. Curr Biol 23 :1215 –1220 . 10.1016/j.cub.2013.05.027 23791725 Fielding AB , Schonteich E , Matheson J , Wilson G , Yu X , Hickson GR , Srivastava S , Baldwin SA , Prekeris R , Gould GW (2005 ) Rab11-FIP3 and FIP4 interact with Arf6 and the exocyst to control membrane traffic in cytokinesis. EMBO J 24 :3389 –3399 . 10.1038/sj.emboj.7600803 16148947 Franco SJ , Martinez-Garay I , Gil-Sanz C , Harkins-Perry SR , Müller U (2011 ) Reelin regulates cadherin function via Dab1/Rap1 to control neuronal migration and lamination in the neocortex. Neuron 69 :482 –497 . 10.1016/j.neuron.2011.01.003 21315259 Fukaya M , Kamata A , Hara Y , Tamaki H , Katsumata O , Ito N , Takeda S , Hata Y , Suzuki T , Watanabe M , Harvey RJ , Sakagami H (2011 ) SynArfGEF is a guanine nucleotide exchange factor for Arf6 and localizes preferentially at post-synaptic specializations of inhibitory synapses. J Neurochem 116 :1122 –1137 . 10.1111/j.1471-4159.2010.07167.x 21198641 Fukaya M , Fukushima D , Hara Y , Sakagami H (2014 ) EFA6A, a guanine nucleotide exchange factor for Arf6, interacts with sorting nexin-1 and regulates neurite outgrowth. J Neurochem 129 :21 –36 . 10.1111/jnc.12524 24261326 Gil-Sanz C , Franco SJ , Martinez-Garay I , Espinosa A , Harkins-Perry S , Müller U (2013 ) Cajal-Retzius cells instruct neuronal migration by coincidence signaling between secreted and contact-dependent guidance cues. Neuron 79 :461 –477 . 10.1016/j.neuron.2013.06.040 23931996 Gil-Sanz C , Landeira B , Ramos C , Costa MR , Müller U (2014 ) Proliferative defects and formation of a double cortex in mice lacking Mltt4 and Cdh2 in the dorsal telencephalon. J Neurosci 34 :10475 –10487 . 10.1523/JNEUROSCI.1793-14.2014 25100583 Gillingham AK , Munro S (2007 ) The small G proteins of the Arf family and their regulators. Annu Rev Cell Dev Biol 23 :579 –611 . 10.1146/annurev.cellbio.23.090506.123209 17506703 Gressens P (2000 ) Mechanisms and disturbances of neuronal migration. Pediatr Res 48 :725 –730 . 10.1203/00006450-200012000-00004 11102537 Guerrier S , Coutinho-Budd J , Sassa T , Gresset A , Jordan NV , Chen K , Jin WL , Frost A , Polleux F (2009 ) The F-BAR domain of srGAP2 induces membrane protrusions required for neuronal migration and morphogenesis. Cell 138 :990 –1004 . 10.1016/j.cell.2009.06.047 19737524 Hara Y , Fukaya M , Tamaki H , Sakagami H (2013 ) Type I phosphatidylinositol 4-phosphate 5-kinase γ is required for neuronal migration in the mouse developing cerebral cortex. Eur J Neurosci 38 :2659 –2671 . 10.1111/ejn.12286 23802628 Hernández-Deviez DJ , Casanova JE , Wilson JM (2002 ) Regulation of dendritic development by the ARF exchange factor ARNO. Nat Neurosci 5 :623 –624 . 10.1038/nn865 12032543 Horgan CP , McCaffrey MW (2009 ) The dynamic Rab11-FIPs. Biochem Soc Trans 37 :1032 –1036 . 10.1042/BST0371032 19754446 Houndolo T , Boulay PL , Claing A (2005 ) G protein-coupled receptor endocytosis in ADP-ribosylation factor 6-depleted cells. J Biol Chem 280 :5598 –5604 . 10.1074/jbc.M411456200 15590645 Inoue T , Krumlauf R (2001 ) An impulse to the brain—using in vivo electroporation. Nat Neurosci 4[Suppl] :1156 –1158 . 10.1038/nn1101-1156 11687822 Ito S , Takeichi M (2009 ) Dendrites of cerebellar granule cells correctly recognize their target axons for synaptogenesis in vitro. Proc Natl Acad Sci USA 106 :12782 –12787 . 10.1073/pnas.0906653106 19622730 Jossin Y (2011 ) Polarization of migrating cortical neurons by Rap1 and N-cadherin: revisiting the model for the Reelin signaling pathway. Small GTPases 2 :322 –328 . 10.4161/sgtp.18283 22545231 Jossin Y , Cooper JA (2011 ) Reelin, Rap1 and N-cadherin orient the migration of multipolar neurons in the developing neocortex. Nat Neurosci 14 :697 –703 . 10.1038/nn.2816 21516100 Kadowaki M , Nakamura S , Machon O , Krauss S , Radice GL , Takeichi M (2007 ) N-cadherin mediates cortical organization in the mouse brain. Dev Biol 304 :22 –33 . 10.1016/j.ydbio.2006.12.014 17222817 Kato M , Dobyns WB (2003 ) Lissencephaly and the molecular basis of neuronal migration. Hum Mol Genet 12SpecNo1 :R89 –R96 . 12668601 Kawauchi T (2012 ) Cell adhesion and its endocytic regulation in cell migration during neural development and cancer metastasis. Int J Mol Sci 13 :4564 –4590 . 10.3390/ijms13044564 22605996 Kawauchi T , Chihama K , Nabeshima Y , Hoshino M (2006 ) Cdk5 phosphorylates and stabilizes p27kip1 contributing to actin organization and cortical neuronal migration. Nat Cell Biol 8 :17 –26 . 10.1038/ncb1338 16341208 Kawauchi T , Sekine K , Shikanai M , Chihama K , Tomita K , Kubo K , Nakajima K , Nabeshima Y , Hoshino M (2010 ) Rab GTPases-dependent endocytic pathways regulate neuronal migration and maturation through N-cadherin trafficking. Neuron 67 :588 –602 . 10.1016/j.neuron.2010.07.007 20797536 Kim Y , Lee SE , Park J , Kim M , Lee B , Hwang D , Chang S (2015 ) ADP-ribosylation factor 6 (ARF6) bidirectionally regulates dendritic spine formation depending on neuronal maturation and activity. J Biol Chem 290 :7323 –7335 . 10.1074/jbc.M114.634527 25605715 Klein S , Franco M , Chardin P , Luton F (2006 ) Role of the Arf6 GDP/GTP cycle and Arf6 GTPase-activating proteins in actin remodeling and intracellular transport. J Biol Chem 281 :12352 –12361 . 10.1074/jbc.M601021200 16527809 Letinic K , Sebastian R , Toomre D , Rakic P (2009 ) Exocyst is involved in polarized cell migration and cerebral cortical development. Proc Natl Acad Sci USA 106 :11342 –11347 . 10.1073/pnas.0904244106 19549839 LoTurco JJ , Bai J (2006 ) The multipolar stage and disruptions in neuronal migration. Trends Neurosci 29 :407 –413 . 10.1016/j.tins.2006.05.006 16713637 Lu J , Sheen V (2005 ) Periventricular heterotopia. Epilepsy Behav 7 :143 –149 . 10.1016/j.yebeh.2005.05.001 15996530 Luton F , Klein S , Chauvin JP , Le Bivic A , Bourgoin S , Franco M , Chardin P (2004 ) EFA6, exchange factor for ARF6, regulates the actin cytoskeleton and associated tight junction in response to E-cadherin engagement. Mol Biol Cell 15 :1134 –1145 . 10.1091/mbc.E03-10-0751 14668475 Macia E , Luton F , Partisani M , Cherfils J , Chardin P , Franco M (2004 ) The GDP-bound form of Arf6 is located at the plasma membrane. J Cell Sci 117 :2389 –2398 . 10.1242/jcs.01090 15126638 Macia E , Partisani M , Paleotti O , Luton F , Franco M (2012 ) Arf6 negatively controls the rapid recycling of the β2 adrenergic receptor. J Cell Sci 125 :4026 –4035 . 10.1242/jcs.102343 22611259 Makyio H , Ohgi M , Takei T , Takahashi S , Takatsu H , Katoh Y , Hanai A , Ueda T , Kanaho Y , Xie Y , Shin HW , Kamikubo H , Kataoka M , Kawasaki M , Kato R , Wakatsuki S , Nakayama K (2012 ) Structural basis for Arf6-MKLP1 complex formation on the Flemming body responsible for cytokinesis. EMBO J 31 :2590 –2603 . 10.1038/emboj.2012.89 22522702 Marchetti G , Escuin S , van der Flier A , De Arcangelis A , Hynes RO , Georges-Labouesse E (2010 ) Integrin alpha5beta1 is necessary for regulation of radial migration of cortical neurons during mouse brain development. Eur J Neurosci 31 :399 –409 . 10.1111/j.1460-9568.2009.07072.x 20105241 Miyazaki H , Yamazaki M , Watanabe H , Maehama T , Yokozeki T , Kanaho Y (2005 ) The small GTPase ADP-ribosylation factor 6 negatively regulates dendritic spine formation. FEBS Lett 579 :6834 –6838 . 10.1016/j.febslet.2005.11.022 16325184 Montagnac G , Sibarita JB , Loubéry S , Daviet L , Romao M , Raposo G , Chavrier P (2009 ) ARF6 Interacts with JIP4 to control a motor switch mechanism regulating endosome traffic in cytokinesis. Curr Biol 19 :184 –195 . 10.1016/j.cub.2008.12.043 19211056 Nadarajah B , Brunstrom JE , Grutzendler J , Wong RO , Pearlman AL (2001 ) Two modes of radial migration in early development of the cerebral cortex. Nat Neurosci 4 :143 –150 . 10.1038/83967 11175874 Nishimura YV , Sekine K , Chihama K , Nakajima K , Hoshino M , Nabeshima Y , Kawauchi T (2010 ) Dissecting the factors involved in the locomotion mode of neuronal migration in the developing cerebral cortex. J Biol Chem 285 :5878 –5887 . 10.1074/jbc.M109.033761 20022952 Niwa H , Yamamura K , Miyazaki J (1991 ) Efficient selection for high-expression transfectants with a novel eukaryotic vector. Gene 108 :193 –199 . 1660837 Ohshima T (2014 ) Neuronal migration and protein kinases. Front Neurosci 8 :458 . 10.3389/fnins.2014.00458 25628530 Olson EC , Walsh CA (2002 ) Smooth, rough and upside-down neocortical development. Curr Opin Genet Dev 12 :320 –327 . 12076676 Palacios F , Price L , Schweitzer J , Collard JG , D'Souza-Schorey C (2001 ) An essential role for ARF6-regulated membrane traffic in adherens junction turnover and epithelial cell migration. EMBO J 20 :4973 –4986 . 10.1093/emboj/20.17.4973 11532961 Palacios F , Schweitzer JK , Boshans RL , D'Souza-Schorey C (2002 ) ARF6-GTP recruits Nm23-H1 to facilitate dynamin-mediated endocytosis during adherens junctions disassembly. Nat Cell Biol 4 :929 –936 . 10.1038/ncb881 12447393 Powelka AM , Sun J , Li J , Gao M , Shaw LM , Sonnenberg A , Hsu VW (2004 ) Stimulation-dependent recycling of integrin beta1 regulated by ARF6 and Rab11. Traffic 5 :20 –36 . 14675422 Raemaekers T , Peric A , Baatsen P , Sannerud R , Declerck I , Baert V , Michiels C , Annaert W (2012 ) ARF6-mediated endosomal transport of Telencephalin affects dendritic filopodia-to-spine maturation. EMBO J 31 :3252 –3269 . 10.1038/emboj.2012.182 22781129 Rakic P (2007 ) The radial edifice of cortical architecture: from neuronal silhouettes to genetic engineering. Brain Res Rev 55 :204 –219 . 10.1016/j.brainresrev.2007.02.010 17467805 Saito T , Nakatsuji N (2001 ) Efficient gene transfer into the embryonic mouse brain using in vivo electroporation. Dev Biol 240 :237 –246 . 10.1006/dbio.2001.0439 11784059 Sakagami H , Kamata A , Fukunaga K , Kondo H (2005 ) Functional assay of EFA6A, a guanine nucleotide exchange factor for ADP-ribosylation factor 6 (ARF6), in dendritic formation of hippocampal neurons. Methods Enzymol 404 :232 –242 . 10.1016/S0076-6879(05)04022-X 16413273 Sakagami H , Sanda M , Fukaya M , Miyazaki T , Sukegawa J , Yanagisawa T , Suzuki T , Fukunaga K , Watanabe M , Kondo H (2008 ) IQ-ArfGEF/BRAG1 is a guanine nucleotide exchange factor for Arf6 that interacts with PSD-95 at postsynaptic density of excitatory synapses. Neurosci Res 60 :199 –212 . 10.1016/j.neures.2007.10.013 18164504 Sakagami H , Hara Y , Fukaya M (2016 ) Interaction of serologically defined colon cancer antigen-3 with Arf6 and its predominant expression in the mouse testis . Biochem Biophys Res Commun 477 :868 –873 . 10.1016/j.bbrc.2016.06.150 27373827 Sakakibara A , Sato T , Ando R , Noguchi N , Masaoka M , Miyata T (2014 ) Dynamics of centrosome translocation and microtubule organization in neocortical neurons during distinct modes of polarization. Cereb Cortex 24 :1301 –1310 . 10.1093/cercor/bhs411 23307632 Sanda M , Ohara N , Kamata A , Hara Y , Tamaki H , Sukegawa J , Yanagisawa T , Fukunaga K , Kondo H , Sakagami H (2010 ) Vezatin, a potential target for ADP-ribosylation factor 6, regulates the dendritic formation of hippocampal neurons. Neurosci Res 67 :126 –136 . 10.1016/j.neures.2010.02.008 20188128 Schiel JA , Simon GC , Zaharris C , Weisz J , Castle D , Wu CC , Prekeris R (2012 ) FIP3-endosome-dependent formation of the secondary ingression mediates ESCRT-III recruitment during cytokinesis. Nat Cell Biol 14 :1068 –1078 . 10.1038/ncb2577 23000966 Scholz R , Berberich S , Rathgeber L , Kolleker A , Köhr G , Kornau HC (2010 ) AMPA receptor signaling through BRAG2 and Arf6 critical for long-term synaptic depression. Neuron 66 :768 –780 . 10.1016/j.neuron.2010.05.003 20547133 Schonteich E , Pilli M , Simon GC , Matern HT , Junutula JR , Sentz D , Holmes RK , Prekeris R (2007 ) Molecular characterization of Rab11-FIP3 binding to ARF GTPases. Eur J Cell Biol 86 :417 –431 . 10.1016/j.ejcb.2007.05.004 17628206 Schweitzer JK , D'Souza-Schorey C (2002 ) Localization and activation of the ARF6 GTPase during cleavage furrow ingression and cytokinesis. J Biol Chem 277 :27210 –27216 . 10.1074/jbc.M201569200 12016212 Schweitzer JK , D'Souza-Schorey C (2005 ) A requirement for ARF6 during the completion of cytokinesis. Exp Cell Res 311 :74 –83 . 10.1016/j.yexcr.2005.07.033 16181626 Seki T (2002 ) Expression patterns of immature neuronal markers PSA-NCAM, CRMP-4 and NeuroD in the hippocampus of young adult and aged rodents. J Neurosci Res 70 :327 –334 . 10.1002/jnr.10387 12391592 Sekine K , Honda T , Kawauchi T , Kubo K , Nakajima K (2011 ) The outermost region of the developing cortical plate is crucial for both the switch of the radial migration mode and the Dab1-dependent “inside-out” lamination in the neocortex. J Neurosci 31 :9426 –9439 . 10.1523/JNEUROSCI.0650-11.2011 21697392 Sekine K , Kawauchi T , Kubo K , Honda T , Herz J , Hattori M , Kinashi T , Nakajima K (2012 ) Reelin controls neuronal positioning by promoting cell-matrix adhesion via inside-out activation of integrin α5β1. Neuron 76 :353 –369 . 10.1016/j.neuron.2012.07.020 23083738 Sekine K , Kubo K , Nakajima K (2014 ) How does Reelin control neuronal migration and layer formation in the developing mammalian neocortex? Neurosci Res 86 :50 –58 . 10.1016/j.neures.2014.06.004 24969097 Shiba T , Koga H , Shin HW , Kawasaki M , Kato R , Nakayama K , Wakatsuki S (2006 ) Structural basis for Rab11-dependent membrane recruitment of a family of Rab11-interacting protein 3 (FIP3)/Arfophilin-1. Proc Natl Acad Sci USA 103 :15416 –15421 . 10.1073/pnas.0605357103 17030804 Shikanai M , Nakajima K , Kawauchi T (2011 ) N-cadherin regulates radial glial fiber-dependent migration of cortical locomoting neurons. Commun Integr Biol 4 :326 –330 . 10.4161/cib.4.3.14886 21980571 Shitamukai A , Konno D , Matsuzaki F (2011 ) Oblique radial glial divisions in the developing mouse neocortex induce self-renewing progenitors outside the germinal zone that resemble primate outer subventricular zone progenitors. J Neurosci 31 :3683 –3695 . 10.1523/JNEUROSCI.4773-10.2011 21389223 Simon GC , Schonteich E , Wu CC , Piekny A , Ekiert D , Yu X , Gould GW , Glotzer M , Prekeris R (2008 ) Sequential Cyk-4 binding to ECT2 and FIP3 regulates cleavage furrow ingression and abscission during cytokinesis. EMBO J 27 :1791 –1803 . 10.1038/emboj.2008.112 18511905 Suzuki A , Arikawa C , Kuwahara Y , Itoh K , Watanabe M , Watanabe H , Suzuki T , Funakoshi Y , Hasegawa H , Kanaho Y (2010 ) The scaffold protein JIP3 functions as a downstream effector of the small GTPase ARF6 to regulate neurite morphogenesis of cortical neurons. FEBS Lett 584 :2801 –2806 . 10.1016/j.febslet.2010.05.020 20493856 Tabata H , Nakajima K (2001 ) Efficient in utero gene transfer system to the developing mouse brain using electroporation: visualization of neuronal migration in the developing cortex . Neuroscience 103 :865 –872 . 10.1016/S0306-4522(01)00016-1 11301197 Tabata H, Nakajima K (2003 ) Multipolar migration: the third mode of radial neuronal migration in the developing cerebral cortex . J Neurosci 23 :9996 –10001 . 14602813 Tai CY , Mysore SP , Chiu C , Schuman EM (2007 ) Activity-regulated N-cadherin endocytosis. Neuron 54 :771 –785 . 10.1016/j.neuron.2007.05.013 17553425 Takahashi S , Takei T , Koga H , Takatsu H , Shin HW , Nakayama K (2011 ) Distinct roles of Rab11 and Arf6 in the regulation of Rab11-FIP3/arfophilin-1 localization in mitotic cells. Genes Cells 16 :938 –950 . 10.1111/j.1365-2443.2011.01538.x 21790911 Takeichi M (2007 ) The cadherin superfamily in neuronal connections and interactions. Nat Rev Neurosci 8 :11 –20 . 10.1038/nrn2043 17133224 Vitale N , Mawet J , Camonis J , Regazzi R , Bader MF , Chasserot-Golaz S (2005 ) The small GTPase RalA controls exocytosis of large dense core secretory granules by interacting with ARF6-dependent phospholipase D1. J Biol Chem 280 :29921 –29928 . 10.1074/jbc.M413748200 15980073 Wilson GM , Fielding AB , Simon GC , Yu X , Andrews PD , Hames RS , Frey AM , Peden AA , Gould GW , Prekeris R (2005 ) The FIP3-Rab11 protein complex regulates recycling endosome targeting to the cleavage furrow during late cytokinesis. Mol Biol Cell 16 :849 –860 . 10.1091/mbc.E04-10-0927 15601896 Yamada K , Fukaya M , Shibata T , Kurihara H , Tanaka K , Inoue Y , Watanabe M (2000 ) Dynamic transformation of Bergmann glial fibers proceeds in correlation with dendritic outgrowth and synapse formation of cerebellar Purkinje cells. J Comp Neurol 418 :106 –120 . 10701759 Yazaki Y , Hara Y , Tamaki H , Fukaya M , Sakagami H (2014 ) Endosomal localization of FIP3/Arfophilin-1 and its involvement in dendritic formation of mouse hippocampal neurons. Brain Res 1557 :55 –65 . 10.1016/j.brainres.2014.02.018 24576489 Ye T , Ip JP , Fu AK , Ip NY (2014 ) Cdk5-mediated phosphorylation of RapGEF2 controls neuronal migration in the developing cerebral cortex. Nat Commun 5 :4826 . 10.1038/ncomms5826 25189171 Yu JY , DeRuiter SL , Turner DL (2002 ) RNA interference by expression of short-interfering RNAs and hairpin RNAs in mammalian cells. Proc Natl Acad Sci USA 99 :6047 –6052 . 10.1073/pnas.092143499 11972060

PMC005xxxxxx/PMC5002985.txt

==== Front eNeuroeNeuroeneuroeneuroeNeuroeNeuro2373-2822Society for Neuroscience 10.1523/ENEURO.0188-16.2016eN-NWR-0188-161New ResearchCognition and BehaviorThe Neural System of Postdecision Evaluation in Rostral Frontal Cortex during Problem-solving Tasks Postdecision evaluation in rostral frontal cortexWan Xiaohong 12Cheng Kang 13Tanaka Keiji 11 Cognitive Brain Mapping Laboratory, RIKEN Brain Science Institute, Wako, Saitama 351-0198, Japan2 State Key Laboratory of Cognitive Neuroscience and Learning and IDG/McGovern Institute for Brain Research, Beijing Normal University, Beijing, 100875, China, and3 Support Unit for Functional Magnetic Resonance Imaging, RIKEN Brain Science Institute, Wako, Saitama 351-0198, JapanThe authors declare no competing financial interests. Author Contributions: X.W., K.C., and K.T. designed the experiments, analyzed the results and wrote the manuscript. X.W. conducted the experiments. This work was supported in part by Fujitsu Laboratories, Japan, National Natural Science Foundation of China 31471068 to X.W., and 111 Project, China B07008 to X.W. We thank Drs. H. Nakatani, T. Asamizuya, K. Ueno, and C. Suzuki for technical assistance, and Japan Shogi Association for help in recruiting professional shogi players as subjects and advice on the task. Correspondence should be addressed to Dr. Keiji Tanaka, Cognitive Brain Mapping Laboratory, RIKEN Brain Science Institute, Wako, Saitama 351-0198, Japan. E-mail: keiji@riken.jp.9 8 2016 29 8 2016 Jul-Aug 2016 3 4 ENEURO.0188-16.20161 7 2016 29 7 2016 4 8 2016 Copyright © 2016 Wan et al.2016Wan et al.This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International, which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed.Visual Abstract Little attention has been paid to the postdecision processing in fMRI studies with task paradigms in which there was no explicit feedback. Although late-onset BOLD responses were previously observed in the lateral frontopolar cortex after the familiar-novel decision on visually presented words, the nature of neural activations that caused the late-onset BOLD responses remained elusive. We here found, in human experts conducting complicated problem-solving tasks in their expertise domain, that the rostral frontal cortex, including the lateral frontopolar cortex, along with the anterior inferior parietal lobule, was activated only during the postdecision period, although there was no feedback. That is, these areas showed late-onset BOLD responses, and fitting of the BOLD responses with different models indicates that they were caused by neural activations that occurred after the decision. However, there was no response after performing a sensory-motor control task, and the magnitude of postdecision activations was correlated with the degree of uncertainty about the preceding decision, which suggests that the postdecision neural activations were associated with the preceding decision procedure. Furthermore, the same set of areas was more strongly activated when the subject explicitly rethought the preceding problem. These results suggest that the rostral frontal cortex, together with anterior inferior parietal lobule, comprises a network for uncertainty monitoring and exploration of alternative resolutions in postdecision evaluation. The present results thus introduce a new aspect of the functional gradient along the rostrocaudal axis in the frontal cortex. dorsal anterior cingulate cortexexpertfMRIfrontopolar cortexmetacognitionshogiThe Fujitsu LaboratoriesThe National Natural Science Foundation of China31471068The 111 Project, ChinaB07008 cover-dateJuly/August 2016 ==== Body Significance Statement After generating and selecting a solution for a given problem, we often evaluate the solution, even without explicit feedback. This may be to change the solution when there is an opportunity for change, and, more generally, to deepen our understanding of similar problems. By using checkmate problems in Japanese chess, shogi, without giving any explicit feedback, we here found that the postdecision evaluation is mainly conducted by a frontoparietal network involving rostral frontal areas. These findings also introduce a new aspect of functional gradient in frontal cortex: postdecision evaluation in rostral areas and online task execution in caudal areas. Introduction While there is a consensus that the frontal cortical areas anterior to the primary motor cortex play essential roles in cognitive control, our knowledge about how these areas are functionally organized remains limited. Consistent with the rostrocaudal gradient in their intrinsic and external anatomical connections (Barbas and Pandia, 1991; Carmichael and Price, 1996; Fuster, 1997; Petrides, 2005; Saleem et al., 2014), neuroimaging and neuron-recording studies have found evidence suggesting a functional gradient in this region. Rostral frontal areas are more involved in domain-general processing with longer time span and execution of tasks with higher-order structure, whereas caudal frontal areas are more involved in domain-specific processing with shorter time span and execution of simpler tasks (Fuster, 1997; Ramnani and Owen, 2004; Botvinick, 2007; Koechlin and Summerfield, 2007; Badre and D’Esposito, 2009). The interpretation of these previous results, however, requires caution, as most of previous studies used deterministic tasks with limited problem space. The problems in real life are more complex and are usually accompanied with uncertainty; thus, exploration is required to resolve the uncertainty for taking the currently optimal action (Yoshida and Ishii, 2006; Badre et al., 2012). The rostrocaudal gradient may be associated with the control distribution between exploration and exploitation. Indeed, recent neuroimaging studies have suggested specific involvement of rostral frontal areas in exploration of nondefault options (Daw et al., 2006; Boorman et al., 2009, 2013; Kolling et al., 2012, 2014). In the present study, we explored the possibility that the rostrocaudal gradient in frontal cortex is differentially associated with online control of task execution and postdecision process of high-level monitoring as well as exploration of alternatives. Although the classical cognitive theory of human problem-solving proposes that decision-making is generally followed by evaluation (Newell and Simon, 1972; Engel et al., 1993; Zelazo et al., 1997), neural correlates of the postdecision evaluation have been little examined, except for the cases in which decisions were followed by explicit feedback. Although late-onset BOLD signals are observed in the lateral frontopolar cortex (lFPC) in the tasks requiring familiar-versus-novel judgment on visually presented words (Schacter et al., 1997; Buckner et al, 1998; Reynolds et al., 2006), the association of the late-onset BOLD signals with postdecision neural activities remains elusive. We measured brain activities of expert players of shogi, Japanese chess, while they generated an idea of the best next-move in a given board position. No feedback was given after decisions were made. We found that caudal frontal areas were activated only during the online processing of the generation task. In contrast, specific activations were observed after decisions in rostral frontal areas, including the lFPC, dorsal anterior cingulate cortex (dACC), and middle dorsolateral prefrontal cortex (mDLPFC), and in the anterior inferior parietal lobule (aIPL). We also found that postdecision activities tended to be larger when the subject was more uncertain about the decision that the subject had just made, and the same group of areas was activated when the subjects rethought the problem. It is reasonable to assume that uncertainty about the preceding decision drove people to explore other possible solutions in rethinking. The present results suggest that postdecision evaluation, which may be driven by decision uncertainty and be associated with exploration, takes place primarily in the frontoparietal network, including rostral frontal areas. Materials and Methods Subjects All the subjects were right-handed Japanese males. Experiments 1 and 2 were conducted on 17 professional players (30.2 ± 1.5 years old) and 17 high-rank amateur players (proficient level: 2-4 dan, 32.5 ± 2.3 years old). Another group of 17 high-rank amateur players (3-4 dan, 31.4 ± 2.7 years old) participated in Experiment 3. Nineteen novice subjects (20.3 ± 0.2 years old) participated in Experiment 4. Informed consent was obtained from each subject in accordance with protocols approved by the Institute Research Ethics Committee of RIKEN. Tasks Experiments 1 and 2 were originally designed to reveal neural substrates of quick next-move generation in expert players. Brain activities specifically associated with quick next-move generation, in contrast to those associated with deliberative search, had been reported (Wan et al., 2011). We made unexpected findings in these experiments, that is, the post-task activations in the post-task network common to the quick generation and deliberative search, which is the main subject of the present paper. To further examine properties of the post-task activations, we designed Experiment 3. Experiment 4 was originally designed to study the development of the capability of quick next-move generation and of associated brain activities along a long-term training of the game skill. The main results of Experiment 4 have been previously reported (Wan et al., 2012). The development of the post-task activations along the training is reported in the present paper. Subjects viewed the images for the tasks through an optic-fiber goggle system (resolution 800 × 600). All visual stimuli (200 × 200 pixels) were restricted to within 3 degrees of visual angle. Experiment 1. Trials of quick generation task were intermingled with those of sensory-motor control task (see Fig. 1A ). The subject was first presented with a board pattern in both tasks. In the generation task, the board pattern provided a checkmate problem, for which the subject generated the idea of the first move of the move sequences that would reach the final checkmate (capturing the opponent’s king) against the optimum counter moves of the opponent. In the control task, the board pattern was composed only of opponent’s pieces, among which the subject had to find the king piece. As there were no pieces of the subject’s side, the subject could not think about the next-move. For both next-move generation and control tasks, after selecting the answer from four options, the trial proceeded to an intertrial interval (ITI) period, during which the subject answered two simple questionnaires and then performed a distractor task. Figure 1. Tasks and activation patterns in Experiment 1. A, Sequence of main task events in each trial. Red and yellow shadings represent the on-task and ITI periods, respectively. B, The on-task activation (red) determined by comparing BOLD signal changes during the generation task with those during the ITI period after the generation task, and the off-task activation (yellow) determined by comparing signal changes during the ITI period after the generation task with those during the ITI period after the control task. The results shown in Figures 1B, 2, and 3 were obtained using half of the data from Experiment 1. A, lFPC; B, mDLPFC; C, dACC; D, aIPL; X, pDLPFC (right); Y, pre-SMA; O and O’, PMd; P and P’, pDLPFC (left); Q and Q’, precuneus. In detail, each trial started with the appearance of a fixation point. After a 1 s fixation, the board pattern of a checkmate problem (in the generation task) or that composed only of the opponent’s pieces (in the control task) appeared for 1 s. Four choice options were then presented, and the subject selected the one that matched his idea of the first move (in the generation task) or of the king’s position (in the control task) within 2 s. While general checkmate problems require reports of the sequence of moves that would reach the final checkmate together with the optimum counter moves of the opponent, we asked our subjects just to report their ideas of the first move of the sequence. In the questionnaires given at the beginning of the ITI period, the subject reported his confidence in the previous choice and then whether he made the choice by recalling the memory of the problem (see the next paragraph). In the following distractor task during ITI, the subject was presented with shogi pieces one at a time at a rate of four pieces per second (150 ms each followed by a mask for 100 ms) and reported the appearance of a “Gold” piece by pressing a button. We intended to stop the thinking of the previous problem by the distractor task. The total duration of a trial was fixed to 11 s. Because the trial proceeded to the next phase at the subject’s button press in the main task and questionnaires, the period of the distractor task varied from 3 to 8 s. We gave 180 trials of the generation task and 60 trials of the sensory-motor control task to each subject. The board pattern was trial unique. The checkmate problems were newly created by professional players belonging to the Japan Shogi Association. Their difficulty varied with 7-15 moves required to reach the final checkmate, including the opponent’s counter moves, so that they were challenging, and not too difficult, for both professional and high-level amateur players. As the checkmate problems were newly created, the subject rarely reported the problems’ memory in the second questionnaire (8% for the professional players and 3% for the amateur players). Some more details of the task have been described previously (Wan et al., 2011). Experiment 2.To compare brain activities associated with quick generation of the best next-move with those associated with deliberative search of the best next-move, we conducted Experiment 2 (with deliberative search) following Experiment 1 (with quick generation) in each fMRI session. We randomly selected 30 checkmate problems from the problems that the subject failed to give correct answers in Experiment 1. Each problem was presented for up to 8 s, during which the subject was instructed to search deliberately to find the best next-move (see Fig. 5A ). When the subject pressed a button within 8 s, the trial proceeded to the answer-selection phase. Otherwise, the task entered automatically to the answer selection phase at the end of 8 s. After the subject chose the answer from the four options within 2 s, the trial proceeded to the ITI period occupied with the distractor task (the “Gold” piece detection). Unlike Experiment 1, there was no questionnaire for confidence or memory report or sensory-motor control trial. The length of each trial was fixed to 16 s. The period of ITI varied more (between 5 and 13 s) than the variation in Experiment 1. Experiment 3. The basic structure of the task paradigm was the same as that in Experiment 1, but to further examine properties of the post-task activations found in Experiments 1 and 2, three conditions were introduced during ITI. After the quick next-move generation task, the subject was engaged in (1) the “Gold” piece detection, which was used in Experiments 1 and 2; (2) fixation only (“rest” condition); or (3) rethinking the preceding problem (“rethink” condition) (see Fig. 8A ). After the sensory-motor control trials, only the first two conditions (“Gold” piece detection and “rest” conditions) were provided. The task sequence was similar to that used in Experiment 1. After a 0.5 s fixation period, a checkmate problem or a board pattern for the detection of the king was presented for 2 s, and the subject was then instructed to choose one from four options within 2 s. Unlike Experiments 1 and 2, the screen for selection remained until the end of the 2 s period, even after the subject had pressed the button. Subsequently, an instruction indicating the condition during the upcoming ITI was briefly shown for 0.5 s, followed by ITI for 4 s. At the end of the ITI period, the subject either chose, within 2 s, his answer again from four options in the “rethink” condition or simply pressed the button marked in red in the other two conditions. The length of each trial was 11 s, and there was no questionnaire for confidence or memory report. There were 60 trials for each combination of on-tasks and ITI conditions, and a total of 300 trials were given to each subject. The order of the five types of trials was random, except that a control trial was always followed by a generation trial. Of 180 generation trials, 120 were preceded by a control trial and 60 were preceded by a generation trial. Experiment 4. The subjects who had had no prior experience of shogi were daily trained for playing games of a simple shogi (“gogo” shogi) for 15 weeks, and the fMRI experiments were conducted twice on each subject: at the early (the 2-3 weeks) and end (the 14-15 weeks) phases of the training. The subject practiced the exercise, on average, 40 min per day. Gogo-shogi uses a 5 × 5 board, in place of a 9 × 9 board in original shogi, and fewer types of pieces. A game of gogo-shogi is completed with fewer moves (typically ∼30, including both sides) than moves for a typical game of original shogi (∼120). The basic structure of the task paradigm used in the fMRI experiments was the same as that in Experiment 1: trials of the correct next-move generation task were intermingled with those of the sensory-motor control task, and the main task was followed by the ITI period occupied with the distractor task (“Gold” piece detection). Unlike Experiment 1, however, checkmate problems of gogo-shogi were used for the correct next-move generation task, the board pattern was presented for 2 s, the subject chose the answer within 3 s, all the four choice options were concrete moves or positions, and there were no questionnaires. The length of each trial was 11 s, and 180 next-move generation trials together with 60 sensory-motor control trials were given to each subject, as in Experiment 1. The next-move generation trials were randomly intermingled with the control trials. MRI specifications All fMRI experiments were conducted using a 4 T MRI system with a head gradient coil (Agilent). A birdcage radiofrequency transmit-receive coil was used in Experiments 1 and 2. A combination of a birdcage radiofrequency transmit coil and 4 phased-array receive surface coils was used in Experiments 3 and 4. Experiments 1 and 2. Functional images were acquired using a two-segment center-ordered gradient echo T2∗ EPI sequence with volume TR of 2 s, TE of 15 ms, slice thickness of 5.5 mm, and in-plane resolution of 3.75 × 3.75 mm2 (FOV: 24 × 24 cm2; flip angle: 40 degrees). Twenty-one axial slices, parallel to the anterior commissure-posterior commissure line, were acquired with an interleaved acquisition procedure. Experiments 3 and 4. Functional images were acquired using a TSENSE technique and a two-segment center-ordered gradient echo T2∗ EPI sequence with volume TR of 2 s, TE of 15 ms, slice thickness of 4.0 mm, and in-plane resolution of 3.0 × 3.0 mm2 (FOV: 19.2 × 19.2 cm2; flip angle: 40 degrees). Thirty oblique slices, oriented 15 degrees from the anterior commissure-posterior commissure line, were collected with an interleaved acquisition procedure. fMRI data analyses Preprocessing.After reconstruction of EPI images, data analyses were performed using BrainVoyager (Brain Innovation). To correct for the rigid head motion, all EPI images were realigned to the first volume of the first scan. Datasets in which translation motions were >1.0 mm or rotation motions were >1.0 degree were discarded. Functional EPI images were then transformed into the Talairach space by resampling the data with a resolution of 2 × 2 × 2 mm3. A spatial smoothing with a 4 mm Gaussian kernel (FWHM) and a high-pass temporal filtering with a cutoff of 0.005 Hz were applied to all fMRI data. Determination of post-task period and associated regions of interest (ROIs). GLM regression analyses were used to determine ROIs of activation and for several other analyses. All regression analyses used two regressors: one obtained by convolving the on-task period with a canonical hemodynamic function (HRF) and the other by convolving the ITI period or post-task period with the canonical HRF. The on-task period started at the onset of the problem presentation; and in Experiments 1, 3, and 4, it included the entire time for problem presentation and the difference in response time, obtained by subtracting the mean response time of the subject in the sensory-motor control task from the response time in individual trials. The mean response time in the control task was subtracted from the response time in individual trials because the former time was assumed to be used for perceiving the options and pushing a button. The remaining time was likely used for thinking the problem. For Experiment 2, the on-task period was fixed to the problem presentation time. As the subjects voluntarily terminated the problem presentation to move on to the option selection, we assumed that they did not continue to think the problem after the termination of the problem. The ITI-period regressor was used only for the initial analysis of the activation in Experiment 1. The ITI period started at the time when the subject made the choice, which initiated two questionnaires, and covered the periods for the questionnaires and for the distractor task (“Gold” piece detection). Half of the data in Experiment 1 was used to determine the activation (see Fig. 1B ); and then, by deconvolving the mean time course of the obtained BOLD signal change in each ROI for this off-task activation, the onset and duration of the neural activation that evoked BOLD signal changes were estimated (see Fig. 3). The initial and end positions were averaged among the four ROIs to determine a common window for analyses of the post-task activation (post-task window). The estimated initial and end positions of the post-task were 0.3 and 3.8 s, respectively, after the onset of the first questionnaire. A regressor obtained by convolving this post-task activation period with the canonical HRF was used for the second analysis of the activation in Experiment 1 and for Experiments 2-4. The second analysis of the activation in Experiment 1 used the remaining half of the data to redefine the ROIs for on-task and post-task activations (see Fig. 4A ). These ROIs were subsequently used for all ROI analyses presented in this paper. For Experiments 2 and 4, we assumed that the post-task period started at the time when the subject made the choice, which initiated the distractor task in these experiments. For Experiment 3, we assumed that the post-task period started at the end of instruction for the ITI task (0.5 s after the subject made the choice). Each of the on-task-period, ITI-period, and post-task-period regressors was differentiated into two: one for trials of quick next-move generation task and the other for trials of sensory-motor control task. For Experiment 3, each of the two post-task-period regressors was further differentiated into three according to the ITI conditions. The variation in response time across trials was taken into consideration in the regression analysis, as the length of the on-task period was changed depending the differences in the response time. To determine activated voxels, regression beta coefficients calculated for each individual subject were used for a group random-effect ANOVA. For Experiments 1 and 2, in which both amateur and professional players participated, after a group random-effect analysis within each subject group, a conjunction analysis across subject groups was also performed. Multiple comparison corrections were performed by calculating the false discovery rate (FDR, p < 0.05 after correction) throughout the whole brain. Unless noted otherwise, ROI analyses were based on data from both hemispheres. Analysis of functional connectivity between the on-task and post-task networks. The correlation of trial-by-trial variances in activities was examined by using the data obtained in Experiment 1 as follows (see Fig. 10). (1) The response at each time was averaged across voxels within each ROI of each subject. (2) The mean response was integrated over time with weights of the function obtained by convolving the canonical HRF with the on-task period (for the on-task network) or post-task period (for the post-task network). (3) Deviations of the trial-by-trial responses from the mean averaged over all the trials were calculated. (4) The coefficient was calculated for the correlation of the deviations for each pair of ROIs in the on-task and post-task networks (one ROI from the on-task network and the other ROI from the post-task network). (5) The coefficient was averaged over all the ROI pairs. (6) The mean coefficients in individual subjects of each subject group (professional or amateur) were converted by Fisher’s z-transformation. (7) The significance of the correlation between post-task activation and preceding on-task activation, between post-task activation and succeeding on-task activation, and between successive on-task activations was statistically examined by applying a two-tailed, one-sample t test to the distributions of converted coefficients across subjects within each subject group. As individual time points of BOLD signals in each trial are not independent, the degrees of freedom were modified by the Bartlett correction factor. Adjusting time courses of BOLD signal changes in Experiments 1 and 3.The ITI periods used in the present study were not long enough for the BOLD signal to completely return to the baseline level within each trial. To remove the general initial declining trends of BOLD signal changes caused by neural activations in the previous generation-task trial (for Experiment 1, see Fig. 2A ), we calculated differences in the mean time course between generation-task trials preceded by a generation-task trial (g-G trials) and generation-task trials preceded by a control-task trial (c-G trials) (for Experiment 1, see Fig. 2B ) and subtracted the mean differential time course (g-G – g-C trials) from the time courses in individual trials preceded by a generation-task trial (g-G and g-C trials). The declining trends were estimated separately for Experiments 1 and 3. As there were no control-task trials in Experiment 2, original response time courses were used for Experiment 2 (see Figs. 6,7). Figure 2. Initial decreasing trends in BOLD signal changes that commonly occurred in generation and control trials. BOLD signal changes were aligned to the onset of problem presentation (sti_on). A, The original BOLD signal changes in the generation trials preceded by a generation trial (black lines) and those in the control trials preceded by a generation trial (gray lines), respectively, in Experiment 1. Error bars indicate SEM across subjects. Red shadows represent the board position presentation period. Yellow shadows represent the ITI period. Mean signal changes were vertically shifted so that the value at time zero became zero. B, The initial decreasing trends in Experiment 1, extracted by calculating BOLD signal differences between the generation trials preceded by a control trial and the generation trials preceded by a generation trial. Fitting of fMRI signal change time courses in Experiment 2. We used the data obtained in Experiment 2 (deliberate search) to determine the task event to which neural activations that caused BOLD signal changes were locked. We divided the trials into three groups according to the subject’s responses: the trials that the subject terminated within 4 s (2.8 ± 0.3 s, mean ± SE), which are referred to as quick-search trials; those that the subject terminated between 4 and 8 s (6.3 ± 0.3 s, slow-search trials); and those during which the subject did not press the button to terminate. We did not include the last group of trials in the main analysis because the subjects likely continued to think the given problem in these trials even after the problem presentation was already terminated. We determined a model that consistently explained the time courses of BOLD signal changes in quick-search and slow-search trials by using the following formulas: (1) BOLDfit=a ∗ conv(HRF,Na)+b where (2) HRF=(tτ)n−1e−tττΓ(n−1) and a and b are the linear parameters to adjust the magnitude and bias of BOLD signal changes. conv represents the convolution, HRF a hemodynamic response function, n a parameter to adjust the shape of HRF, and τ a parameter to adjust the width of HRF. Na represents the position and duration of neural activation that evoked BOLD signal changes. We examined the performance of two models. Model 1 assumed that Na started at the problem onset and lasted for the problem presentation period, which varied across trials. Model 2 assumed that Na started at the time when the subject made the choice (i.e., the ITI onset) and lasted for 3.5 s (see above). Because there were no sensory-motor control trials in Experiment 2, we used original BOLD signals. Although the deconvolution method was used to reduce the contamination of responses from the preceding trial, it was only partly successful: a decreasing phase of BOLD response from the previous trial remained (see Fig. 6). The mean time courses aligned to the onset of the problem presentation were fitted with Model 1 (see Fig. 6A ), and those aligned to the subject’s option selection (i.e., the ITI onset) were fitted with Model 2 (see Fig. 6B ). The Levenberg-Marquardt optimization algorithm (“fminsearch” in MATLAB 7.7, The MathWorks) was used to determine the set of a, b, n, and τ that gave the best fit for each model. Finally, the mean square error (MSE) was compared between the two models (see Table 2). Results We used the checkmate problem in the game of shogi in the present study. The checkmate problem is a well-designed puzzle, a type of rule-based problem-solving task. For each problem (or board position), there exists only one correct solution, which is a sequence of moves that reaches the final checkmate (capture of the opponent’s king), even with optimal counter moves by the opponent. The number of moves required before reaching the final checkmate varied from 7 to 15 (including the opponent’s moves). Although checkmate problems usually require the player to report the entire sequence of moves that reaches the final checkmate, we asked our subjects just to report their ideas of the first move, so as to emphasize the rapid generation of the best next-move. Players with high proficient levels can quickly generate, for most of problems, an idea of the first move of the sequence that may reach the final checkmate. Experiments 1 and 2 were originally designed to identify neural substrates of the process in expert players to quickly generate the best next-move (Wan et al., 2011). Experiment 4 was designed as a follow-up study to examine the development of the capability along a long-term training and of associated brain activities in subjects who had been novices before the training (Wan et al., 2012). The present paper mainly reports the unexpected brain activities that occurred after the subject completed the decision. Experiment 3 was newly designed to further examine properties of these brain activities. fMRI responses during the off-task or post-task period In Experiment 1, trials of quick next-move generation task (to report the best next-move to the given checkmate problem) were intermingled with those of sensory-motor control task (to report the position of the “King” piece in the given board, which was exclusively composed of the opponent’s pieces) (Fig. 1A ; see Experiment 1). In both generation and control trials, after the subject selected the answer, the trial proceeded to the ITI period that was occupied with two simple questionnaires and a distractor task (to detect the appearance of “Gold” piece in sequentially presented pieces). Although the length of a trial was fixed to 11 s, because the times used by the subject for making the choice and answering the two questionnaires in each trial varied (1.26 ± 0.35 s and 1.31 ± 0.59 s, respectively, mean ± SD across trials), the length of the distractor task also varied across trials (3-8 s, 6.41 ± 0.75 s, mean ± SD across trials). BOLD signal changes were analyzed with two regressors: (1) the on-task (i.e., online processing) period covering the problem presentation (1 s) and the response times in individual trials subtracted by the mean response time of the subject in the sensory-motor control task (0.40 ± 0.33 s, mean ± SD across trials), convolved with the canonical HRF; and (2) the ITI period occupied with the questionnaires and distractor task, convolved with the canonical HRF. Several cortical areas, including the pDLPFC (or inferior frontal junction; BA 8/9), pre-SMA, dorsal premotor cortex (PMd, BA 6), and posterior precuneus (BA 7), were activated during the on-task period of the next-move generation task compared with the off-task period after the next-move generation task (p < 0.05, FDR corrected; Fig. 1B , red patches, Figs. 2,3, right; Table 1), as we had reported previously (Wan et al., 2011). Activations in these areas during the sensory-motor task were as strong as those during the next-move generation task in Experiment 1 (Figs. 2,3, right), whereas their activations during the sensory-motor task were much smaller than those during the next-move generation task in Experiment 3 (see Fig. 9, right), where the board positions were presented for a longer period (2 s) and the proportion of control trials was higher. Cortical areas in the rostral frontal cortex, including FPC, mDLPFC, and dACC, which have been repeatedly reported to be associated with cognitive control, were not activated during online processing of quick next-move generation (Figs. 2,3, left). Figure 3. Time courses of on-task and post-task (off-task) activations in Experiment 1. A, BOLD signal changes aligned to the onset of problem presentation (sti_on). Mean signal changes were vertically shifted so that the value at time zero became zero. The effects from the previous generation trial were removed by subtracting the differences between mean signals in generation trials preceded by a generation trial and those in generation trials preceded by a control trial (shown in Fig. 2B) from the time courses of signal changes in individual trials. Other conventions are the same as in Figure 2A. Horizontal red open bars represent the position and duration of neural activations determined by deconvolving mean BOLD signal changes in generation trials with a canonical HRF. B, The same as in A, but BOLD signal changes were aligned to the subject’s option selection (selection), which was also the onset of questionnaires. Mean signal changes were vertically shifted so that the value at 2 s before time zero became zero. Table 1. The brain areas in which fMRI activities were higher during the on-task period compared with those during the post-task period in the quick next-move generation task (on-task activation), and the brain areas in which fMRI activities were higher during the post-task period after the quick next-move generation task compared with those during the post-task period after the sensory-motor control task (post-task activation) in Experiment 1 (p < 0.05, FDR corrected) Brain region Talairach coordinates No. of voxels x y z Post-task activation lFPC L -27 53 11 123 R 23 55 10 43 mDLPFC L -40 29 33 214 R 35 31 33 125 dACCa -2 24 44 79 Left pDLPFC L -44 8 20 97 PMd L -22 2 54 422 R 20 2 58 296 aIPL L -51 -29 44 943 R 35 -35 39 268 Precuneus L -8 -64 48 452 R 4 -60 48 496 On-task activation pDLPFC L -43 6 28 104 R 39 5 27 92 Pre-SMA L -7 8 47 114 R 5 7 48 45 PMd L -22 -7 51 216 R 24 -10 53 228 Precuneus L -16 -60 44 162 R 15 -59 47 176 aThis cluster included voxels in both hemispheres. Table 2. The parameters that provided the optimal fitting of the mean fMRI time courses in each ROI obtained from Experiment 2 with two modelsa Quick search Slow search Subject group Model n τ a b a b MSE On-task network Right pDLPFC All 1 5 1.52 1.07 0.03 0.85 -0.01 0.20 2 4 0.83 0.48 -0.12 0.72 -0.13 0.30 Pre-SMA All 1 5 1.39 1.32 0.00 0.99 -0.07 0.15 2 4 0.70 0.70 -0.12 0.88 -0.17 0.44 Post-task network lFPC All 1 7 1.32 0.82 -0.20 0.45 -0.24 0.40 2 4 1.63 1.03 -0.05 1.25 -0.04 0.07 1 1 7 1.38 0.79 -0.23 0.58 -0.18 0.42 2 4 1.67 1.02 -0.08 1.28 -0.01 0.09 2 1 7 1.36 0.84 -0.18 0.54 -0.20 0.44 2 4 1.68 1.05 -0.07 1.26 -0.02 0.11 mDLPFC All 1 7 1.25 1.17 -0.26 0.72 -0.27 0.79 2 4 1.26 0.86 0.04 1.17 -0.03 0.22 1 1 7 1.30 1.09 -0.22 0.94 -0.23 0.74 2 4 1.29 0.90 0.01 1.21 0.02 0.24 2 1 7 1.32 1.22 -0.19 1.06 -0.21 0.85 2 4 1.31 095 0.03 1.24 0.04 0.26 dACC All 1 7 1.04 1.50 -0.08 0.89 -0.22 0.34 2 5 0.82 1.15 0.01 1.23 -0.09 0.14 1 1 7 1.08 1.03 0.04 1.12 -0.28 0.39 2 5 0.88 1.12 0.08 1.25 -0.04 0.16 2 1 7 1.25 1.42 -0.06 0.94 -0.23 0.43 2 5 0.90 1.18 0.05 1.21 -0.02 0.19 aIPL All 1 7 1.28 1.64 -0.23 0.82 -0.33 0.54 2 4 1.34 1.27 -0.04 1.32 -0.14 0.24 1 1 7 1.35 1.58 -0.28 0.89 -0.37 0.59 2 4 1.36 1.29 -0.02 1.28 -0.10 0.29 2 1 7 1.32 1.62 -0.26 0.88 -0.32 0.56 2 4 1.32 1.32 -0.06 1.29 -0.07 0.26 aMSE indicates the goodness of the fitting. The 32 subjects who participated in Experiment 2 were divided into two groups: Group 1, with more quick-search trials than slow-search trials (n = 16); and Group 2, with more slow-search trials than quick-search trials (n = 16). In contrast, when we compared activities during the ITI period after the next-move generation task with those during the ITI period after the sensory-motor control task, widespread cortical areas were activated (p < 0.05, FDR corrected, Fig. 1B , yellow patches, and Figs. 2,3, left), although the subjects were engaged in the identical “Gold” piece detection task in both conditions. The areas that exhibited such an off-task activation comprised lFPC (BA 10), mDLPFC (BA 9/46), dACC (BA 8/32), left pDLPFC, PMd, and aIPL (BA 5/7), and the posterior precuneus (Table 1). In some areas, the region that showed off-task activations after the next-move generation task partially overlapped with the region activated during the online processing of the next-move generation task (left pDLPFC, bilateral PMd, and bilateral posterior precuneus) (Fig. 1B ). The selectivity for the on-task or off-task activation was weaker in these regions (Fig. 4B ), as can be expected from the partial overlapping between on-task and off-task activation regions. To further examine the properties of the on-task and off-task activation, we focused on the regions that showed only the on-task or off-task activation. We refer to the regions activated only during online processing of next-move generation (right pDLPFC and bilateral pre-SMA) as the on-task network, and the regions activated only during the ITI period after the quick next-move generation (bilateral lFPC, bilateral mDLPFC, dACC, and bilateral aIPL) as the off-task network. The off-task network largely coincided with the “frontoparietal control network” (Vincent et al., 2008), but not with the “default-mode network” activated during rest (Buckner et al., 2008). Figure 4. Recalculated activation patterns and selectivity in Experiment 1. A, Similar to those in Figure 1B, but obtained by using the second half of the data from Experiment 1 and the post-task period. Conventions are the same as in Figure 1B. B, Magnitudes of on-task and post-task activations in the post-task network areas, on-task network areas, and overlapping areas, obtained by using the second half of the data from Experiment 1. Red bars represent differences between β values for the on-task period regressor in generation-task trials and those for the post-task period regressor in generation-task trials. Yellow bars represent differences between β values for the post-task period regressor in generation-task trials and those for the post-task period regressor in control-task trials. Error bars indicate SEM across subjects. The mean time courses of BOLD signal changes are shown in Figure 2A . Among the 180 trials of the next-move generation task, we focused on the 120 trials that were preceded by a next-move generation trial. Activities in these generation trials were contrasted with those in the 60 trials of the sensory-motor control task, which were all preceded by a next-move generation trial. There were initial decreasing trends commonly in both groups of trials (Fig. 2A ), which likely represented the late part of signal changes caused by neural activations in the preceding generation trial. They were estimated by taking the differences between the mean signals in generation trials preceded by a generation trial and those in generation trials preceded by a control trial (Fig. 2B ). The rectified mean time courses of BOLD signals obtained by subtracting the estimated initial decreasing trends from signal changes in individual trials are shown in Figure 3. Signal changes in individual trials were aligned to the beginning of the on-task period (i.e., the onset of problem presentation) in Figure 3A , as in Figure 2, and to the beginning of the ITI period (i.e., the onset of questionnaire period) in Figure 3B . In the regions of the off-task network, the BOLD signals in generation trials (black traces) started to increase and deviate from those in control trials (gray traces) 3-5 s after the task onset, and the signals in generation trials peaked toward the end of the trial. The time difference between this peak in BOLD signals in the off-task network (on average, 9 s after the onset of the board position presentation) and the peak in the BOLD signals in the on-task network (on average, 6 s after the onset of the board position presentation) was ∼3 s, which largely matched the mean time difference between the onset of the board position presentation and the onset of the ITI period (3.54 ± 0.01 s, mean ± SEM across subjects). By deconvolving these BOLD signals with the canonical HRF, we estimated the position and duration of the neural activations that led to BOLD signal changes in the generation trials (Fig. 3A,B , red horizontal bars). Whereas neural-activation periods estimated for the on-task network matched the on-task period relatively well, those estimated for the off-task network started approximately at the beginning of the ITI period. However, estimated neural-activation periods in the off-task network were much shorter than the off-task period: they all ended in the middle of the ITI period. Thus, we decided to use the estimated period of neural activation, averaged across lFPC, dACC, mDLPFC, and aIPL, as the second regressor. This period started 0.3 s after the subject’s option selection, or the onset of questionnaires, and ended 3.8 s after the subject’s option selection. This 3.5 s period was termed the “post-task period.” Statistical activation maps were generated again using one regressor associated with this post-task period and the other with the on-task period (Fig. 4A ). To avoid circularity of analysis, the initial determination of activated voxels and the estimation of position and period of neural activations that led to BOLD signal changes were made using a half of the data (odd runs), and activated voxels were determined again with the estimated neural-activation period using the remaining half of the data (even runs). Although activation maps calculated by the post-task-period regressor were almost identical to those calculated by the off-task-period regressor, we hereafter refer to the activation determined by the post-task-period regressor as the “post-task activation” and to the network of regions that showed significant post-task activation as the “post-task network,” for clarity. Late-onset BOLD signal changes can be explained by late neural activations, but not by slow hemodynamic responses In Experiment 2, the subjects were given longer time (8 s) to search for the best next-move while they were allowed to move to answer anytime within 8 s by pressing a button (Fig. 5A ; see Experiment 2). The time was long enough for the subjects, who were proficient in shogi, to make deliberative search. Similar on-task and post-task activation patterns as those in Experiment 1 were observed in Experiment 2 (Fig. 5B,C ). The presence of post-task activations after deliberative search indicated that the post-task activations occurred commonly after the quick generation and deliberative search of the best next-move to a given checkmate problem. Figure 5. Tasks and activation pattern in Experiment 2. A, Sequence of main task events in each trial. B, Statistical parametric maps for the on-task activation (the next-move generation period contrasted with the post-task period). C, β values determined by GLM analyses for the on-task and post-task period regressors. Error bars indicate SEM across subjects. Experiment 2 also provided a good opportunity to confirm that the post-task activations in the post-task network were caused by neural activations occurring after the completion of the decision in the next-move generation task, as the trial-by-trial variance of the interval between the onset of board position presentation and beginning of the ITI period was larger in Experiment 2 (SD, 3.6 s) than that in Experiment 1 (0.4 s). We focused on two groups of trials selected based on the task duration: the trials in which the subject moved to answer within 4 s (quick-search trials), and those in which the subject moved to answer between 4 and 8 s (slow-search trials). Trials in which the subjects completed the 8 s period without pressing the button were not included in the analyses below because the way by which the subject finished the search in these trials was different from that in the other trials (passive vs active). The numbers of quick- and slow-search trials were equal (25% and 25%), and the mean search durations in quick- and slow-search trials were 2.8 and 6.3 s, respectively. BOLD signals were aligned to the onset of the search period in Figure 6A and to the option selection (i.e., ITI onset) in Figure 6B , for quick-search trials (black traces) and slow-search trials (red traces). To examine whether late-onset BOLD signal changes in the post-task network were caused by neural activations during the generation task or those after the termination of the task, we adjusted shape parameters of the HRF (a gamma function) to fit BOLD signal changes: the HRF was convolved with the period of generation task in Model 1 and with the 3.5-s off-task period, starting at the option selection, in Model 2 (see Fitting of fMRI signal change time courses in Experiment 2). Shape parameters of the HRF in each area were adjusted separately in the two models but common to quick-search and slow-search trials. Figure 6A, B (dashed traces) indicates the time courses of the models that fitted the data optimally. Model 2 fitted BOLD signal changes better in all the regions of the post-task network, whereas Model 1 fitted BOLD signal changes better in all the regions of the on-task network (Table 2). Figure 6. Activation time courses in Experiment 2. A, Mean BOLD signal changes aligned to the onset of the problem presentation (sti_on), obtained by the deconvolution method for the trials with problem presentation shorter than 4 s (on average, 2.8 s, quick-search trials, solid black lines), and those obtained for the trials with problem presentation between 4 and 8 s (on average, 6.3 s, slow-search trials, solid red lines). Horizontal black and red filled bars represent mean problem-presentation periods, in quick-search and slow-search trials, respectively. B, Mean BOLD signal changes aligned to the subject’s option selection (selection), which was also the onset of the ITI period. Horizontal black and red filled bars represent the post-task period, starting at the onset of the ITI onset. A, B, Mean signal changes were vertically shifted so that the value at the problem-presentation onset was zero. Error bars indicate SEM across subjects. Horizontal open black and red bars represent the estimated neural activation period in quick-search and slow-search trials, respectively. Broken black and red lines indicate the time courses determined by optimally fitting HRF parameters with the neural activation coincided, presumably, with the on-task period in A or the post-task period in B. Blue line segments overlaid on each horizontal axis indicate the period during which the MSE was calculated for the fitting. The ratio of the number of quick-search trials to that of slow-search trials varied across subjects. Thus, it was possible that the differences in BOLD time courses between quick- and slow-search trials, as shown in Figure 6, merely reflected the differences between subjects, but not between trials. To rule out this possibility, we divided the subjects who participated in Experiment 2 into two groups: one with more quick- than slow-search trials and the other with more slow-search trials. We found that the time courses of BOLD signal changes in the post-task network were similar between the two groups of subjects (Fig. 7) and that Model 2 consistently better fitted BOLD signal changes in the post-task network in both groups of subjects (Table 2). Figure 7. Activation time courses in two groups of subjects in Experiment 2: Group 1 in which the quick-search trials dominated and Group 2 in which the slow-search trials dominated. BOLD signal changes were aligned to the onset of the problem presentation. Conventions are the same as Figure 6A. The time course of the slow-search trials (red lines) was delayed compared with that of the quick trials (black lines) in each region commonly in both subject groups. Robust post-task activation regardless of conditions during ITI According to the multiple-task switching hypothesis (Braver et al., 2003) for the function of the rostral frontal cortex, the post-task activation might be associated with the task switching between the next-move generation task and subsequent “Gold” piece detection task. To examine this possibility, we included the trials in which the subject was only required to maintain eye fixation during the ITI period (“rest” condition) in Experiment 3 (Fig. 8A ; see Experiment 3). The board position was presented for 2 s, instead of 1 s, because only (high-level) amateur players were recruited in this experiment. There was no questionnaire, and we therefore commenced the 3.5 s post-task period at the offset of the 0.5 s instruction for the ITI condition in the GLM analysis. Figure 8. Tasks and activation patterns in Experiment 3. A, Main task events in each trial. The ITI period was filled with “Gold” piece detection task, “rest,” or “rethinking” after the generation task, and with “Gold” piece detection task or “rest” after the control task. B, Statistical parametric maps for the on-task activation (the on-task period contrasted with the post-task period in generation-task trials) and for the post-task activation (the post-task period after the generation task contrasted with the post-task period after the control task). The trials with “Gold” piece detection and with “rest” were combined. Those with “rethinking” were not included. Other conventions are the same as in Figure 1B. The same sets of cortical areas, respectively, exhibited on-task and post-task activations in Experiment 3 as those in Experiment 1 (Fig. 8B ). Figure 9 shows mean time courses of BOLD signal changes. The initial declining trends in the trials following a generation trial had been removed as in Experiment 1 (see Adjusting time courses of BOLD signal changes in Experiments 1 and 3). The GLM analysis revealed that BOLD signal changes in the ROIs of the post-task network in the “rest” condition following the next-move generation task were significantly larger than those in the “rest” condition following the sensory-motor control task (two-tailed paired t test with 17 subjects, lFPC: t(16) = 3.9, p = 0.0006; mDLPFC: t(16) = 3.6, p = 0.001; dACC: t(16) = 3.2, p = 0.003; aIPL: t(16) = 3.3, p = 0.002; Fig. 9, left). They were also larger than those in the “Gold” piece detection condition following the generation task (two-tailed paired t test with 17 subjects, lFPC: t(16) = 2.4, p = 0.015; mDLPFC: t(16) = 2.2, p = 0.021; dACC: t(16) = 2.0, p = 0.029; aIPL: t(16) = 2.1, p = 0.026; Fig. 9, left). As expected, there were no significant differences between the two conditions in the two ROIs of the on-task network (two-tailed paired t test with 17 subjects, right pDLPFC: t(16) = 1.1, p = 0.30; pre-SMA: t(16) = 1.0, p = 0.36; Fig. 9, right). Figure 9. Activation time courses in Experiment 3. Signal changes were aligned to the onset of problem presentation. Error bars indicate SEM across trials. Red and yellow shadows represent the problem presentation and ITI periods, respectively. The differential effects from the previous generation and control trials were removed by subtracting the differences between mean signal changes in generation trials preceded by a generation trial and those in generation trials preceded by a control trial in Experiment 3 from the time courses of signal changes in individual trials preceded by a generation trial (see Adjusting time courses of BOLD signal changes in Experiments 1 and 3). These results suggest that the task switching was unlikely the cause for the post-task activation. They also demonstrated that the post-task activation did not crucially depend on the ongoing task during the ITI period, even though the engagement in the “Gold” piece detection partly reduced the post-task activation. In addition, because there was no questionnaire, either about confidence or memory at the beginning of the ITI period in Experiment 3, the cause of the post-task activation by these explicit evaluation processes is also excluded. Finally, the modulation of the post-task activation by the condition during the ITI period further supports our conclusion that the post-task activations were caused by neural processes that occurred after the preceding decision. If the neural processes that caused post-task activation had occurred before the onset of the ITI period, they could not have been modulated by the task condition during the ITI period. Thus, we contend that the post-task activations were caused by neural processes that occurred after, but in association with, the preceding decision. There was little on-task activation in sensory-motor control trials in Experiment 3 (Fig. 9, right). The difference between the results in Experiments 1 and 3 was likely due to the difference in the presentation time of board position (1 s in Experiment 1 and 2 s in Experiment 3) and in the proportion of control trials (40% in Experiment 3 and 25% in Experiment 1). The subjects might always prepare a task set for the generation task in Experiment 1, whereas they prepared the task set after detecting that the presented board position was specific for the generation task (i.e., including own pieces) in Experiment 3. Association of the post-task activation with the preceding on-task activation Multiple lines of evidence so far have suggested that the post-task activation was associated with the preceding generation task. To further clarify this association, we analyzed the correlation of trial-by-trial variations of the post-task activation in the post-task network with those of the preceding on-task activation in the on-task network (Fig. 10A , r1). Alternatively, the post-task activation might influence the performance of the generation task in the next trial. We thus also analyzed the correlation of the post-task activation in the current trial with the on-task activation in the next trial (Fig. 10A , r2). The data from Experiment 1 were used for these analyses. We used pairs of consecutive generation trials in the analysis of prospective association (r2). For a fair comparison, we also focused on the first trials of two consecutive generation trials in the analysis of retrospective association (r1). The coefficients of the correlation were averaged across all ROI pairs between the two networks for each subject (see Analysis of functional connectivity between the on-task and post-task networks). Figure 10. Correlation of trial-by-trial variations of BOLD signal changes. A, The correlation was calculated between post-task activation in the post-task network and on-task activation in the on-task network associated with the preceding generation process (r1), between the post-task activation and the on-task activation associated with the subsequent generation process (r2), and between the on-task activations in the consecutive generation trials (r3). B, Correlation coefficients in the three combinations. C, Results of similar analyses applied to the post-task activation after the control task. Error bars indicate SEM across subjects. The strength of the post-task activation was significantly correlated with that of the preceding on-task activation (Fig. 10B , r1) (two-tailed, one-sample t test with 17 subjects, mean r = 0.32, t(16) = 4.6, p = 1.5 × 10−4 for the professional group; two-tailed, one-sample t test with 17 subjects, mean r = 0.28, t(16) = 4.3, p = 2.7 × 10−4 for the amateur group), but not with that of the on-task activation in the succeeding generation trial (Fig. 10B , r2) (two-tailed, one-sample t test with 17 subjects, mean r = 0.09, t(16) = 1.2, p = 0.12 for the professional group; two-tailed, one-sample t test with 17 subjects, mean r = 0.04, t(16) = 0.6, p = 0.28 for the amateur group). The former correlation was significantly larger than the latter (two-tailed paired t test with 17 subjects, t(16) = 3.8, p = 7.3 × 10−4 for the professional group; t(16) = 3.4, p = 0.0017 for the amateur group). Meanwhile, there was no significant correlation between on-task activations in consecutive generation trials (Fig. 10B , r3) (two-tailed, one-sample t test with 17 subjects, mean r = 0.07, t(16) = 0.89, p = 0.20 for the professional group; two-tailed, one-sample t test with 17 subjects, mean r = 0.06, t(16) = 0.78, p = 0.22 for the amateur group). Because there was also no significant correlation in either the retrospective direction in control trials (Fig. 10C, r’1; two-tailed, one-sample t test with 17 subjects, mean r = 0.11, t(16) = 1.3, p = 0.10 for the professional group; two-tailed, one-sample t test with 17 subjects, mean r = 0.09, t(16) = 1.1, p = 0.14 for the amateur group) or the prospective direction in sequences of a control trial followed by a generation trial (Fig. 10C, r’2; two-tailed, one-sample t test with 17 subjects, mean r = 0.08, t(16) = 1.0, p = 0.17 for the professional group; two-tailed, one-sample t test with 17 subjects, mean r = 0.07, t(16) = 0.91, p = 0.19 for the amateur group), the difference between the retrospective and prospective directions in generation-generation trial sequences could not be due to the difference in ITI. These results demonstrated that post-task activations in the post-task network, after the option selection, were clearly influenced by the preceding problem-solving process, but they did not have impact on the subsequent problem-solving process. As a final note, behaviorally, post-task activations were not correlated with either the response accuracy or the response time of the next generation trial (two-tailed, one-sample t test with 34 subjects, r = -0.06, t(33) = 0.6, p = 0.27 for response accuracy; two-tailed, one-sample t test with 34 subjects, r = 0.05, t(33) = 0.3, p = 0.39 for response time). Correlation of post-task activation with subject’s uncertainty about the preceding decision Given the retrospective nature of the correlation between post-task and on-task activations, we posit that post-task activations may be related to the subject’s uncertainty about the preceding problem-solving process. Although we obtained a binary report of the subject’s confidence level (yes or no) after a decision was made for each trial in Experiment 1, we used the response time to quantify the uncertainty level because the criteria for binary confidence reports appeared to largely vary among subjects. The proportion of the trials in which each subject gave a confident report did not correlate with the percentage of correct responses of the subject (two-tailed t test, p = 0.12 in the amateur group and p = 0.58 in the professional group). The 180 trials in the next-move generation task in Experiment 1 were divided into four equally sized groups for each subject according to the response time. The mean response time in a trial group was negatively correlated with the proportion of correct responses in the trial group (Fig. 11A ): the regression coefficient determined for individual subjects was -0.42 ± 0.05 (mean ± SEM across subjects), which was significantly smaller than 0 (Model II regression, one-tailed one sample t test with 34 subjects, t(33) = 3.36, p = 0.0008). The mean degree of uncertainty (the proportion of trials with report of “no”) in a trial group, on the other hand, was negatively correlated with the proportion of correct trials in the trial group (Fig. 11B ): the regression coefficient determined for individual subjects was -0.48 ± 0.13 (mean ± SEM across subjects), which was significantly smaller than 0 (Model II regression, one-tailed one sample t test with 34 subjects, t(33) = 3.69, p = 0.0004). Thus, the subjective confidence report reliably reflected the actual performance. In all four ROIs of the post-task network, we found that the post-task activations in a trial group were positively correlated with the degree of uncertainty in the trial group (Model II regression, two-tailed t test with 34 subjects, p values < 0.01; Fig. 11C ). The association of the post-task activations with the uncertainty about the preceding decision was confirmed by the significantly larger post-task activations in the trials in which the subject gave an unconfident report than those in the trials in which the subject gave a confident report (two-tailed paired t test, p values < 0.01 in all four ROIs of the post-task network). Figure 11. Correlation of post-task activations with uncertainty about the preceding decision in Experiment 1. Trials were divided into quarters in each subject according to the response time for A-C. A, The mean response time in each trial group plotted against the accuracy in the trial group (proportion of correct trials). Data points indicate the means in each subject group (professional or amateur). Error bars indicate SEM across subjects. B, The mean degree of uncertainty in each trial group (proportion of trials with confidence report of “no”) plotted against the accuracy in the trial group. C, Off-task activations in each trial group plotted against the mean degree of uncertainty in the trial group. D, Overall degree of uncertainty of each subject plotted against the overall accuracy of the subject. E, Mean off-task activations in each subject plotted against the overall mean degree of uncertainty of the subject. Black circles represent data from professional players. Red circles represent amateur players. We also examined the across-subject correlation between mean post-task activations, averaged over all the trials for a subject, and the overall degree of uncertainty of the subject (proportion of the trials, in which the subject was not confident). Across subjects, the overall degree of uncertainty was highly correlated with the mean response accuracy (two-tailed, one-sample t test with 34 subjects, r = -0.42, p = 0.006; Fig. 11D ). In all four ROIs of the post-task network, we found that the mean BOLD signal change in a subject was positively correlated with the subject’s overall degree of uncertainty (two-tailed, one-sample t test with 34 subjects, p < 0.01; Fig. 11E ). When the same analyses were applied to on-task activations in the two areas of the on-task network, there was no significant correlation between on-task activations and the subject’s degree of uncertainty. Both the across-trial correlation within individual subjects and the across-subject correlation were not significant in either area (two-tailed t test with 34 subjects, p > 0.20 for the across-trial correlation and p > 0.26 for the across-subject correlation in both ROIs). However, there was a possible pitfall in this analysis; that is, the longer on-task-period regressor in the GLM may have diluted on-task activations associated with slow (and unconfident) responses. We thus conducted another GLM analysis, in which a fixed-duration on-task-period regressor, made by convolving the mean response time in the subject with the canonical HRF, was used to calculate the magnitude of on-task activation. This analysis allowed us to detect a marginal positive correlation between on-task activations and the uncertainty in individual subjects (pDLPFC, Model II regression, two-tailed, one-sample t test with 34 subjects, t(33) = 1.56, p = 0.064; pre-SMA, t(33) = 1.62, p = 0.057), suggesting that neural activations in the on-task network may also reflect the subject’s uncertainty. Involvement of the post-task network in decision adjustment Experiment 3 contained an additional condition for the ITI period, in which the subject was instructed to rethink the preceding next-move problem (Fig. 8A ). We contrasted BOLD signal changes when the subject was engaged in “rethinking” with those when the subject merely maintained fixation (i.e., “rest”) during the post-task period following the generation task. Activated cortical areas that were identified in this comparison coincided to a large extent with the areas activated during the post-task period after the generation task, but not with those activated during the generation task in Experiment 1 (Fig. 12A ). This specific augmentation of post-task activations by rethinking was also confirmed by ROI analyses for the ROIs determined in Experiment 1. Activations during rethinking were stronger than post-task activations in rest and “Gold” piece detection conditions in all four ROIs of the post-task network (two-tailed paired t test with 17 subjects, lFPC: t(16) = 2.8, p = 0.012; mDLPFC: t(16) = 2.6, p = 0.020; dACC: t(16) = 2.6, p = 0.020; aIPL: t(16) = 2.5, p = 0.024; Fig. 9, left). In contrast, rethinking did not activate the areas in the on-task network: β values for the post-task period regressor in the rethinking condition were not different from those in either the rest or the “Gold” piece detection condition (two-tailed, one-sample t test with 17 subjects, right pDLPFC: t(16) = 0.9, p = 0.38; pre-SMA: t(16) = 0.8, p = 0.44) and β values for the on-task period regressor in the rethinking condition were not different from those in either the rest or the “Gold” piece detection condition (two-tailed, one-sample t test with 17 subjects, right pDLPFC: t(16) = 0.8, p = 0.44; pre-SMA: t(16) = 0.7, p = 0.46) (Fig. 9, right). These results show that the post-task network rather than on-task network was recruited to rethink the preceding problem to which the subject had once responded. Figure 12. Activations associated with rethinking and correlation of activation magnitude with accuracy improvement in Experiment 3. A, Brain regions that increased activities during the post-task period by rethinking the preceding problem compared with the activities at rest during the post-task period after the generation task (green). Regions displayed in Figure 4A for on-task activation (red) and post-task activation (yellow) are shown again for comparison. Regions activated by rethinking largely coincided with post-task activations (p < 0.05, FDR corrected). B, The mean accuracy in the trials in which the subject did not alter the selection (left), the mean accuracy of the first answers (middle), and the mean accuracy after rethinking (right) in the trials in which the subject altered the selection. C, Rethinking-associated activations in unaltered trials (open bars) and altered trials (gray bars). B, C, *, p < 0.05. **, p < 0.01. ns, Not significant. Error bars indicate SEM across subjects. D, The magnitude of rethinking-associated activation in each subject plotted against the mean accuracy improvement of the subject by rethinking: r = 0.43 and p = 0.04 for lFPC, and r = 0.58 and p = 0.007 for dACC. The option selection was altered by rethinking in about half of the trials (52 ± 4%, mean ± SEM), and the response accuracy was improved by the alteration: the accuracy after alteration was significantly higher than those in the first thinking (two-tailed paired t test with 17 subjects, t(16) = 2.7, p = 0.0081) and in unaltered trials (two-tailed paired t test with 17 subjects, t(16) = 2.8, p = 0.0059; Fig. 12B ). Furthermore, when activations during rethinking in the post-task network were compared between altered and unaltered trials, those in altered trials were found to be stronger in all four ROIs of the post-task network (two-tailed paired t test with 17 subjects, p < 0.05 for all four areas; Fig. 12C ). We also found that the accuracy change by rethinking for each subject was significantly correlated with activations during rethinking in the subject’s lFPC (two-tailed, one-sample t test with 17 subjects, r = 0.43, p = 0.04; Fig. 12D , top) and dACC (two-tailed, one-sample t test with 17 subjects, r = 0.58, p = 0.007; Fig. 12D , bottom), and marginally with those in mDLPFC (two-tailed, one-sample t test with 17 subjects, r = 0.34, p = 0.08) and aIPL (two-tailed, one-sample t test with 17 subjects, r = 0.35, p = 0.08). In brief, these results demonstrated that activations during rethinking in the post-task network were correlated with the beneficial consequence of rethinking, both across trials in individual subjects and across subjects. Increased post-task activation as a result of training In Experiment 4, 19 subjects, who had had no prior experience of playing shogi, learned and practiced daily to play games of a simplified shogi (gogo-shogi) for 15 weeks (Tasks: Experiment 4). Brain activities associated with the quick-generation task (with 2 s board position presentation) were examined twice, at the early (the 2-3 weeks) and end (the 14-15 weeks) phases of the training. We found that post-task activations in the post-task network increased significantly from the first to the second measurement (two-tailed paired t test with 19 subjects, p < 0.05 in all areas; Fig. 13A ), whereas on-task activations in the on-task network did not change (Wan et al., 2012, their Fig. 4). Figure 13. Changes of post-task activations in originally novice subjects along a long-term training in Experiment 4. A, Bars represent differences between β values determined by GLM analyses for the post-task period regressor in generation-task trials and those determined for the post-task period regressor in control-task trials. Open and gray bars represent the results obtained in an early phase (second or third week) and at a late phase (the last week), respectively, of a long-term (15 weeks) training. Error bars indicate SEM across subjects. B, Response accuracy in each of the four trial groups plotted against the mean response time in the trial group at the early and late phases of training. C, Post-task activations in each of the four trial groups plotted against the mean response time in the trial. B, C, Trials were divided into quarters in each subject according to the response time. The data points indicate the mean values averaged over the 19 subjects. Error bars indicate SEM across subjects. Gray and black lines indicate the values at the early and late phases of training, respectively. When trials were divided into quarters in each subject according to the response time, the mean accuracy (percent of correct responses) in each trial group was negatively correlated with the mean response time of the trial group at the late phase of the training (regression coefficient was -0.66 ± 0.19, mean ± SEM across subjects, model II regression, which were significantly less than 0, two-tailed, one-sample t test, t(18) = 3.49, p = 0.001), but not at the early phase of the training (regression coefficient was -0.10 ± 0.30, mean ± SEM, model II regression, two-tailed, one-sample t test, t(18) = 0.34, p = 0.37; Fig. 13B ). Concurrently, the magnitude of post-task activations of each area of the post-task network in each trial group was correlated with the mean response time of the trial group only at the late phase of the training (p < 0.01 in any of the ROIs; Fig. 13C ), but not at the early phase of the training (p > 0.10 in any of the ROIs, Fig. 13C ). Discussion Late-onset BOLD responses caused by postdecision neural activations By measuring brain activities of experienced players while they were solving complex rule-based problems, the checkmate problems of shogi, we revealed that a frontoparietal network composed of rostral frontal cortical regions, including lFPC, mDLPFC, and dACC, along with aIPL, was activated only during the post-task period of a few seconds after the subjects generated the ideas of the best next-move. This post-task activation appeared after quick intuitive generation as well as after deliberate search, but not after performing a sensory-motor control task. By virtue of large variation of the on-task duration in Experiment 2, we confirmed that the responses were aligned to the end of the generation task, or the onset of ITI, but not to the beginning or the middle period of the generation task. That is, fitting of BOLD responses with different models demonstrated that the late-onset BOLD responses were caused by postdecision neural activations that occurred immediately after the preceding decision, but not by delayed hemodynamic responses coupled with neural activations that occurred during the on-task period. This inference was further supported by the observation that the magnitude of the post-task activation was modulated by the condition during the ITI period in Experiment 3. While the post-task activations were caused by neural activations that occurred after the completion of the preceding decision, several lines of evidence suggest that those neural activations were associated with the execution of the preceding generation task, but not the ongoing task during the post-task period. First, the post-task activation did not occur after performing the sensory-motor control task. Second, trial-by-trial variations of the post-task activation were correlated with those of the activation that occurred in another set of brain areas during the preceding on-task period. Third, trial-by-trial variations of the post-task activation were correlated with the subjects’ degree of uncertainty about the correctness of the preceding decision. On the other hand, the ongoing task during the post-task period had only a modulatory influence on the post-task activation. The “Gold” piece detection task, which was devised to interrupt the subject’s thinking about the problem given in the preceding generation task, reduced the post-task activation by only 20%-40% compared with that during fixation only. Little attention has been paid to the postdecision processing in fMRI task paradigms in which there was no explicit feedback, such as a reward or an error/correct signal. Although late-onset BOLD responses were previously observed in lFPC after the familiar-novel decision on visually presented words, the nature of neural activations that caused the late-onset BOLD responses was not determined in previous studies (Schacter et al., 1997; Buckner et al, 1998; Reynolds et al., 2006). Activation of the default-mode network during rest, compared with the activity during task periods, has been repeatedly demonstrated, but the default-mode activation is different from the post-task activation found in the present study, in that the default-mode activation does not depend on the preceding task (Raichle et al., 2001; Buckner et al., 2008). Indeed, there was little overlap between the default-mode and post-task networks. For example, within the frontopolar cortex, the medial part has been assigned as a part of the default-mode network, whereas the lateral part belonged to the post-task network. A network similar to the post-task network has been identified by analyzing the functional connectivity during rest (Vincent et al., 2008), but functions of the network have been little explored (but see Cole et al., 2013). By meta-analyzing a large set of human imaging studies, it has been shown that many different cognitive demands recruit three broad, yet confined, regions of the prefrontal cortex: the dorsal part of anterior cingulate region, the mid-dorsolateral region around the middle and caudal parts of the inferior frontal sulcus, and the mid-ventrolateral region extending from the frontal operculum to the anterior insula (Duncan and Owen, 2000). This set of regions was named as the multidemand system and discussed to be critical for the identification of subtasks and control of their sequential recruitment to achieve remote goals (Duncan, 2010). While each of the three regions was elongated widely in the rostrocaudal dimension, the functional gradient within the system has not been discussed. The regions of the post-task network and those of the on-task network were located in the rostral and caudal parts of the mid-dorsolateral and dorsal anterior cingulate regions of the multidemand system. Thus, our current study indicates a functional subdivision within the multidemand system. The post-task network partly overlapped with both the frontoparietal and cingulo-opercular networks of Dosenbach et al. (2006, 2007,2008). Properties of the postdecision activation As there was no explicit feedback after the preceding decision in the present study, the postdecision activations could not represent the outcome expectation error. The correlation of trial-by-trial variations of the postdecision activation with the subjective uncertainty about the preceding decision suggests that it is the uncertainty about the preceding decision that triggered the postdecision neural activations. Because explicit rethinking of the preceding problem activated the same network, the postdecision activations should contain functional components that overlap with those of rethinking. Thus, it is likely that the postdecision activations observed in our study represented the evaluation and adjustment procedures. Except for the rethinking condition, there was no explicit task requirement or merit for these procedures. The evaluation and adjustment procedures might automatically occur in experienced players as they help the players to better understand the game. Indeed, the postdecision activations were absent in the subjects who had just started to play the game of shogi in less than 3 weeks but emerged after the subjects underwent extensive daily training for 4 months (Experiment 4). Proposition of the post-task network’s general roles in strategy management Frontal areas in the post-task network, including lFPC, are activated while the subjects perform tasks of higher-order structure or abstract information processing (Baker et al., 1996; Koechlin et al., 1999; Christoff et al., 2001; Ramnani and Owen, 2004; Koechlin and Summerfield, 2007; Badre and D’Esposito, 2007, 2009). lFPC and dACC are activated during uncertainty-driven exploration (Daw et al., 2006; Boorman et al., 2009, 2013; Badre et al., 2012; Kolling et al., 2012, 2014), and by a metacognitive process to report the confidence level in visual memory (Yokoyama et al., 2010) and in perceptual judgment of noisy images as well (Fleming et al., 2012). The task of rethinking of the preceding problem, which activated the post-task network in the present experiment, and the tasks used in the previous studies described above have overlapping components. Rethinking of the preceding problem includes evaluation of the preceding decision and exploration of alternative moves. Exploration of alternatives and execution of tasks with higher-order structure require meta-level monitoring of multiple processes. Uncertainty monitoring and exploration of alternatives are the two key components of the metacognitive function (Nelson and Narens, 1990). Thus, here we propose that the post-task network, or the frontoparietal network, mediates a metacognitive control process for monitoring and adjusting decision-making and learning strategies. Generation of the best next-move for a given board position is thought to comprise a series of complex cognitive processes, including recognizing the position, selecting a particular problem-solving strategy, generating sequences of moves that reach the goal, and selecting the best move sequence (de Groot, 1965; Newell and Simon, 1972; Zelazo et al., 1997). Such a complicated process may recruit a metacognitive control process. However, the subjects who participated in the present study were either professional or experienced amateur players. As extensive training on the game of shogi makes the process automatic and the automated process may be implemented in the caudal frontal regions, the frontoparietal network is likely recruited only during the post-task evaluation and adjustment. In other words, whether the frontoparietal network works in the postdecision stage alone or in the on-task control as well may depend on the nature of the task to be performed and the subject’s experience with the task. In conclusion, our findings suggest that recruitment of cognitive control in the frontal cortex is subject to the strategy of task implementation. The caudal frontal areas mainly control the default strategy of exploiting routine processes, whereas the frontoparietal network, including the rostral frontal areas, mainly controls exploration of alternative processes. Our findings also indicate that the exploitation in the caudal frontal areas and the exploration in the frontoparietal network may work in the same task in a complementary manner: the exploitation works during online task execution, whereas the exploration works during postdecision evaluation and adjustment. Synthesis The decision was a result of the Reviewing Editor Philippe Tobler and the peer reviewers coming together and discussing their recommendations until a consensus was reached. A fact-based synthesis statement explaining their decision and outlining what is needed to prepare a revision is listed below. The following reviewers agreed to reveal their identity: Akira O'Connor, Sabina Gherman Dear Dr. Tanaka, Your manuscript has now been seen by two reviewers. Their comments are combined below. They found merit but make several constructive remarks that need to be addressed in a revision, which I look forward to receiving. Best regards, Philippe Tobler The authors report results from 4 experiments, aiming to demonstrate that post-decision evaluation following a task that requires making decisions on Japanese chess problems is supported by activation in a fronto-parietal network. They present evidence that activation in this network: 1) is unique to the post-task period (i.e., not present during task performance itself), yet is related to (i.e., influenced by) activation during the preceding task; 2) can be observed after the problem solving task but not after a sensori-motor control task; 3) is enhanced when subjects are prompted to "rethink" their choice during the post-task period, compared to when they perform a distraction task (additionally, changing choice after rethinking improves performance); 4) correlates with uncertainty of choice in the preceding problem solving task; 5) is enhanced in novice subjects after extensive training. They conclude that the identified network may play a role in post-decision uncertainty monitoring and/or exploration of alternatives. The tasks used are relatively unusual for the literature (shogi check-mate problems) but are far superior to many decision-making tasks in that they require participants to have undergone many hundreds of hours of training and are tasks in which the expert participants are extremely heavily invested. The manuscript has been revised to consider two previous reviewers' points and this appears to have been carried out as comprehensively as could be expected. I have a small number of additional minor points which the authors may wish to deal with in any future revision. Taken collectively the results are interesting and make for a compelling argument. Overall, the authors perform rigorous control analyses to rule out alternative explanations for their main findings. There are some methodological concerns that should be clarified and/or addressed: 1. Details of the GLM analysis on the fMRI data are at times unclear (or incomplete). In the Methods section, authors state that all regression analyses used only two regressors (one for the on-task period and one for the post-task period). Firstly, it is not clear how the different conditions (main task vs. control task, type of ITI etc.) were modelled and how follow-up comparisons between conditions or subsets of trials were conducted (for example between [ON-TASK main task trials vs. POST-TASK main task trials], or between [ITI control trials vs. ITI main task trials]* (Experiment 1, Fig. 4B). Secondly, authors do not mention whether they modelled out potential confounding task-related factors (e.g., visual stimulation, onset/duration of different ITI events, motor response etc.) in their GLM analyses. (*Note: I use "main task" to refer to the problem solving, i.e. next-move generation, condition) 2. Authors choose to use response times (RTs) as an index of uncertainty (despite collecting confidence reports in Experiment1), however they do not provide an explicit justification for this choice. RT is indeed known to correlate with confidence thus making it a good alternative measure when confidence reports are not available. Also, I can see how using a graded measure like RTs has an advantage over the binary confidence reports when demonstrating the association with post-task activation. However, it would be useful to confirm that post-task activation in the post-task ROIs is also modulated by subjective confidence (i.e., a High vs. Low contrast). Minor comments/suggestions 1.Not being familiar with the game, I would appreciate an additional sentence explaining whether check-mate problems are problems in which the participant is required to achieve a check-mate (which I assume from the move-limited way in which the check-mate is described) or avoid a check-mate. 2.The abstract may benefit from additional restructuring as it feels a little fragmented (for example the part "while the responses were caused by post-decision neural activations" can probably be omitted; also, it might help to move the conclusion that "post-decision neural activations were associated with the preceding decision procedure" to after all evidence has been mentioned (that is, after "absence of responses after performing a sensory-motor control task" and "the magnitude of post-decision activations was correlated with the degree of uncertainty"). 3.The authors write: "The on-task period started at the onset of the problem presentation, and in Experiments 1, 3 and 4, it included the entire time for problem presentation and the difference in response time, obtained by subtracting the mean response time of the subject in the sensory-motor control task from the response time in individual trials." (Methods, Determination of post-task period and associated ROIs) > From the text it is not clear why the on-task period was defined as it was (my only guess is that it was done in an effort to limit motor preparation activation, however this is not explained) 4."studies have found evidences" (Introduction, 1st paragraph) > evidence 5."the piece of King" (Methods, Experiment 1) > the "King" (piece) 6."As there was no piece of the own side" (Methods, Experiment 1) > As there were no pieces on the participant's side 7."Differed from Experiments 1 and 2" (Methods, Experiment 3) > Unlike Experiments 1 and 2 8."Fig. 1A yellow patches" (Results, fMRI responses during the off-task or post-task period, 3rd paragraph) > Fig. 1B 9."and then we commenced" (Results, Robust post-task activation regardless of conditions during ITI, 1st paragraph) > additionally, we commenced 10."but in associated with" (Results, Robust post-task activation regardless of conditions during ITI) > but were associated with 11.Table 2 > It would be helpful to add a notation on the vertical axis to mark the on-task vs. post-task networks. 12."All the subjects were right-handed Japanese male." p.4. Consider "males" 13."The main results of Experiment 4 had been reported (Wan et al., 2012)." p.4. Consider "have been previously reported" 14."proceeded to an intertrial-interval (ITI) period, which were occupied with two simple..." p.4. Consider "proceeded to an intertrial-interval (ITI) period, during which subjects completed two simple..." 15.The "interruption task" term could be replaced with the more standard "distracter task" terminology. 16."each pair of a ROI in" p.4. Consider "each pair of ROIs in" ==== Refs References Barbas H , Pandya , DN (1991 ) Patterns of connections of the prefrontal cortex in the rhesus monkey associated with cortical architecture In: Frontal lobe function and dysfunction , pp 35 –58 . Oxford : Oxford University . Badre D , D’Esposito M (2007 ) Functional magnetic resonance imaging evidence for a hierarchical organization of the prefrontal cortex. J Cogn Neurosci 19 :2082 –2099 . 10.1162/jocn.2007.19.12.2082 17892391 Badre D , D’Esposito M (2009 ) Is the rostro-caudal axis of the frontal lobe hierarchical? Nat Rev Neurosci 10 :659 –669 . 10.1038/nrn2667 19672274 Badre D , Doll BB , Long NM , Frank MJ (2012 ) Rostrolateral prefrontal cortex and individual differences in uncertainty-driven exploration. Neuron 73 :595 –607 . 10.1016/j.neuron.2011.12.025 22325209 Baker SC , Rogers RD , Owen AM , Frith CD , Dolan RJ , Frackowiak RS , Robbins TW (1996 ) Neural systems engaged in planning: a PET study of the Tower of London Task. Neuropsychologia 34 :515 –526 . 10.1016/0028-3932(95)00133-6 8736565 Boorman ED , Behrens TE , Woolrich MW , Rushworth MFS (2009 ) How green is the grass on the other side? Frontopolar cortex and the evidence in favor of alternative courses of action. Neuron 62 :733 –743 . 10.1016/j.neuron.2009.05.014 19524531 Boorman ED , Rushworth MF , Behrens TE (2013 ) Ventromedial prefrontal and anterior cingulate cortex adopt choice and default reference frames during sequential multi-alternative choice. J Neurosci 33 :2242 –2253 . 10.1523/JNEUROSCI.3022-12.2013 23392656 Botvinick MM (2007 ) Multilevel structure in behavior and in the brain: a model of Fuster’s hierarchy. Philos Trans R Soc Lond B Biol Sci 362 :1615 –1626 . 10.1098/rstb.2007.2056 17428777 Braver TD , Reynolds JR , Donaldson DI (2003 ) Neural mechanisms of transient and sustained cognitive control during task switching. Neuron 39 :713 –726 . 12925284 Buckner RL , Koutstaal W , Schacter DL , Dale AM , Rotte M , Rosen BR (1998 ) Functional-anatomic study of episodic retrieval: II. Selective averaging of event-related fMRI trials to test the retrieval success hypothesis. Neuroimage 7 :163 –175 . 10.1006/nimg.1998.0328 9597658 Buckner RL , Andrews-Hanna JR , Schacter DL (2008 ) The brain’s default network anatomy, function and relevance to disease. Ann N Y Acad Sci 1124 :1 –38 . 10.1196/annals.1440.011 18400922 Carmichael ST , Price JL (1996 ) Connectional networks within the orbital and medial prefrontal cortex of macaque monkeys. J Comp Neurol 371 :179 – 207 . 10.1002/(SICI)1096-9861(19960722)371:2&amp;lt;179::AID-CNE1&amp;gt;3.0.CO;2-# 8835726 Christoff K , Prabhakaran V , Dorfman J , Zhao Z , Kroger JK , Holyoak KJ , Gabrieli JD (2001 ) Rostrolateral prefrontal cortex involvement in relational integration during reasoning. Neuroimage 14 :1136 –1149 . [10.1006/nimg.20 ] []11697945 Cole MW , Reynolds JR , Power JD , Repoversus G , Anticevic A , Braver TS (2013 ) Multi-task connectivity reveals flexible hubs for adaptive task control. Nat Neurosci 16 :1348 –1355 . 10.1038/nn.3470 23892552 Daw ND , O’Doherty JP , Dayan P , Seymour B , Dolan RJ (2006 ) Cortical substrates for exploratory decisions in humans. Nature 441 :876 –879 . 10.1038/nature04766 16778890 de Groot AD (1965 ) Thought and choice in chess (translated from the Dutch original, 1946). The Hague : Mouton . Dosenbach NU , Visscher KM , Palmer ED , Miezin FM , Wenger KK , Kang HC , Burgund ED , Grimes AL , Schlaggar BL , Petersen SE (2006 ) A core system for the implementation of task sets. Neuron 50 :799 –812 . 10.1016/j.neuron.2006.04.031 16731517 Dosenbach NU , Fair DA , Miezin FM , Cohen AL , Wenger KK , Dosenbach RA , Fox MD , Snyder AZ , Vincent JL , Raichle ME , Schlaggar BL , Petersen SE (2007 ) Distinct brain networks for adaptive and stable task control in humans. Proc Natl Acad Sci U S A 104 :11073 –11078 . 10.1073/pnas.0704320104 17576922 Dosenbach NU , Fair DA , Cohen AL , Schlaggar BL , Petersen SE (2008 ) A dual-networks architecture of top-down control. Trends Cogn Sci 12 :99 –105 . 10.1016/j.tics.2008.01.001 18262825 Duncan J (2010 ) The multiple-demand (MD) system of the primate brain: mental programs for intelligent behaviour. Trends Cogn Sci 14 :172 –179 . 10.1016/j.tics.2010.01.004 20171926 Duncan J , Owen AM (2000 ) Common regions of the human frontal lobe recruited by diverse cognitive demands. Trends Cogn Sci 23 :475 –483 . 10.1016/S0166-2236(00)01633-7 11006464 Engel JF , Blackwell RD , Miniard PW (1993 ) Consumer behavior . Fort Worth, TX : Dryden . Fleming SM , Huigen J , Dolan R (2012 ) Prefrontal contributions to metacognition in perceptual decision-making. J Neurosci 32 :6117 –6125 . 10.1523/JNEUROSCI.6489-11.2012 22553018 Fuster J (1997 ) The prefrontal cortex . New York : Raven . Koechlin E , Basso G , Pietrini P , Grafman J (1999 ) The role of the anterior prefrontal cortex in human cognition. Nature 399 :148 –151 . 10.1038/20178 10335843 Koechlin E , Summerfield C (2007 ) An information theoretical approach to prefrontal executive function. Trends Cogn Sci 11 :229 –235 . 10.1016/j.tics.2007.04.005 17475536 Kolling N , Behrens TE , Mars RB , Rushworth MF (2012 ) Neural mechanisms of foraging. Science 336 :95 –98 . 10.1126/science.1216930 22491854 Kolling N , Wittmann M , Rushworth MF (2014 ) Multiple neural mechanisms of decision making and their competition under changing risk pressure. Neuron 81 :1190 –1202 . 10.1016/j.neuron.2014.01.033 24607236 Nelson TO , Narens L (1990 ) Metamemory: a theoretical framework and new findings In: The psychology of learning and motivation (Bower GH , ed), pp 125 –173 . San Diego : Academic . Newell A , Simon HA (1972 ) Human problem-solving . Englewood Cliffs, NJ : Prentice Hall . Petrides M (2005 ) Lateral prefrontal cortex: architectonic and functional organization. Philos Trans R Soc Lond B Biol Sci 360 :781 –795 . 10.1098/rstb.2005.1631 15937012 Raichle ME , MacLeod AM , Snyder AZ , Powers WJ , Gusnard DA , Shulmane GL (2001 ) A default mode of brain function. Proc Natl Acad Sci U S A 98 :676 –682 . 10.1073/pnas.98.2.676 11209064 Ramnani N , Owen AM (2004 ) Anterior prefrontal cortex: insights into function from anatomy and neuroimaging. Nat Rev Neurosci 5 :184 –194 . 10.1038/nrn1343 14976518 Reynolds JR , McDermott KB , Braver TS (2006 ) A direct comparison of anterior prefrontal cortex involvement in episodic retrieval and integration. Cereb Cortex 16 :519 –528 . 10.1093/cercor/bhi131 16049191 Saleem KS , Miller B , Price JL (2014 ) Subdivisions and connectional networks of the lateral prefrontal cortex in the macaque monkey. J Comp Neurol 522 :1641 –1690 . 10.1002/cne.23498 24214159 Schacter DL , Buckner RL , Koutstaal W , Dale AM , Rosen BR (1997 ) Late onset of anterior prefrontal activity during true and false recognition: an event-related fMRI study. Neuroimage 6 :259 –269 . 10.1006/nimg.1997.0305 9417969 Vincent JL , Kahn I , Snyder AZ , Raichle ME , Buckner RL (2008 ) Evidence for a frontoparietal control system revealed by intrinsic functional connectivity. J Neurophysiol 100 :3328 –3342 . 10.1152/jn.90355.2008 18799601 Wan X , Nakatani H , Asamizuya T , Ueno K , Cheng K , Tanaka K (2011 ) The neural basis of intuitive best next-move generation in board game experts. Science 331 :341 –346 . 10.1126/science.1194732 21252348 Wan X , Takano D , Asamizuya T , Suzuki C , Ueno K , Cheng K , Ito T , Tanaka K (2012 ) Developing intuition: neural correlates of cognitive-skill learning in caudate nucleus. J Neurosci 32 :17492 –17501 . 10.1523/JNEUROSCI.2312-12.2012 23197739 Wan X , Cheng K , Tanaka K (2015 ) Neural encoding of opposing strategy values in anterior and posterior cingulate cortex. Nat Neurosci 18 :752 –759 . 10.1038/nn.3999 25894290 Yokoyama O , Miura N , Watanabe J , Takemoto A , Uchida S , Sugiura M , Horie K , Sato S , Kawashima R , Nakamura K (2010 ) Right frontopolar cortex activity correlates with reliability of retrospective rating of confidence in short-term recognition memory performance. Neurosci Res 68 :199 –206 . 10.1016/j.neures.2010.07.2041 20688112 Yoshida W , Ishii S (2006 ) Resolution of uncertainty in prefrontal cortex. Neuron 50 :781 –789 . 10.1016/j.neuron.2006.05.006 16731515 Zelazo PD , Carter A , Reznick J , Frye D (1997 ) Early development of executive function: a problem-solving framework. Rev Gen Psychol 1 :198 –226 .

PMC005xxxxxx/PMC5002986.txt

==== Front Front OncolFront OncolFront. Oncol.Frontiers in Oncology2234-943XFrontiers Media S.A. 10.3389/fonc.2016.00185OncologyClinical TrialDose-Escalated Stereotactic Body Radiation Therapy for Prostate Cancer: Quality-of-Life Comparison of Two Prospective Trials Quon Harvey C. 12*Musunuru Hima Bindu 34Cheung Patrick 34Pang Geordi 4Mamedov Alexandre 4D’Alimonte Laura 4Deabreu Andrea 4Zhang Liying 4Loblaw Andrew 341University of Calgary, Calgary, AB, Canada2Tom Baker Cancer Centre, Calgary, AB, Canada3University of Toronto, Toronto, ON, Canada4Odette Cancer Centre, Sunnybrook Health Sciences Centre, Toronto, ON, CanadaEdited by: Fabio Grizzi, Istituto Clinico Humanitas IRCCS, Italy Reviewed by: Sanja Štifter, University of Rijeka, Croatia; Michael A. Feuerstein, Lenox Hill Hospital, USA *Correspondence: Harvey C. Quon, harvey.quon@albertahealthservices.caSpecialty section: This article was submitted to Genitourinary Oncology, a section of the journal Frontiers in Oncology 29 8 2016 2016 6 18528 6 2016 04 8 2016 Copyright © 2016 Quon, Musunuru, Cheung, Pang, Mamedov, D’Alimonte, Deabreu, Zhang and Loblaw.2016Quon, Musunuru, Cheung, Pang, Mamedov, D’Alimonte, Deabreu, Zhang and LoblawThis is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.Introduction The optimal prostate stereotactic body radiation therapy (SBRT) dose-fractionation scheme is controversial. This study compares long-term quality of life (QOL) from two prospective trials of prostate SBRT to investigate the effect of increasing dose (NCT01578902 and NCT01146340). Material and methods Patients with localized prostate cancer received SBRT 35 or 40 Gy delivered in five fractions, once per week. QOL was measured using the Expanded Prostate Cancer Index Composite at baseline and every 6 months. Fisher’s exact test and generalized estimating equations were used to analyze proportions of patients with clinically significant change and longitudinal changes in QOL. Results One hundred fourteen patients were included, 84 treated with 35 Gy and 30 treated with 40 Gy. Median QOL follow-up was 56 months [interquartile range (IQR) 46–60] and 38 months (IQR 32–42), respectively. The proportion of patients reporting clinically significant declines in average urinary, bowel, and sexual scores were not significantly different between dose levels, and were 20.5 vs. 24.1% (p = 0.60), 26.8 vs. 41.4% (p = 0.16), and 42.9 vs. 38.5% (p = 0.82), respectively. Similarly, longitudinal analysis did not identify significant differences in QOL between treatment groups. Conclusion Dose-escalated prostate SBRT from 35 to 40 Gy in five fractions was not associated with significant decline in long-term QOL. prostatic neoplasmsstereotactic body radiotherapyradiotherapyquality of lifeclinical trialradiation effects ==== Body Introduction There has been significant interest in stereotactic body radiation therapy (SBRT) for prostate cancer, given the potential for increased tumor control (1), patient convenience, and lower treatment costs (2). However, the optimal dose and fractionation remain controversial, with total doses ranging from 33.5 to 50 Gy delivered in five fractions (3–5). No randomized studies have been conducted to evaluate the impact of dose-escalated prostate SBRT. In the only prospective dose-escalation SBRT study to date, groups of 15 patients received 45, 47.5, and 50 Gy in five fractions (5). Grade 3+ toxicities were limited to one patient treated with 47.5 Gy and two patients who received 50 Gy. Bowel quality of life (QOL) was worse for patients on the 47.5 Gy arm, but no differences were found in urinary QOL. Quality-of-life outcomes have become increasingly important in patient counseling as patient-reported experiences can be discordant from physician-rated toxicities (6). There is limited data examining QOL outcomes with increasing prostate SBRT dose. Therefore, we conducted a comparative analysis of long-term QOL outcomes from two prospective clinical trials to evaluate the impact of increasing prostate SBRT dose from 35 to 40 Gy in five fractions. Analysis of biochemical control and toxicity outcomes has been reported separately (7). Materials and Methods This study is a secondary analysis of data from two prospective clinical trials (NCT01578902 and NCT01146340). Both trials were approved by the research ethics board at Sunnybrook Health Sciences Centre. Informed consent was provided by all patients. Study 1 Treatment details of study 1 have been previously published (3, 8). Men over 18 years of age with prostate adenocarcinoma and clinical stage T1-T2b (TNM 2002), Gleason sum ≤6, and Prostate-specific antigen (PSA) ≤ 10 ng/mL were eligible. Neoadjuvant androgen deprivation therapy (ADT) was permitted. Patients were excluded if they had received prior pelvic radiotherapy, a bleeding diathesis, which precluded gold fiducial marker insertion, hip prosthesis, pelvic girth > 40 cm, prostate > 90 cm3 on imaging, or International Prostate Symptom Score (IPSS) > 19. Patients received 35 Gy in five fractions delivered once per week over 29 days. Each patient underwent transperineal insertion of three gold fiducial markers followed by CT simulation. Patients were immobilized in a vacuum lock bag (Vac-Lock, MED-TEC Inc., Orange City, IA, USA) with a comfortably full bladder and empty rectum for simulation and treatment. The clinical target volume (CTV) consisted of the prostate only. Organs at risk (OAR) were contoured as solid organs and included the rectum, bladder, penile bulb, and femoral heads. The planning target volume (PTV) included the CTV plus an isotropic 4 mm margin. Planning objectives included the volume of CTV receiving 35 Gy (CTV V35 Gy) to receive >99%, PTV V33.25 Gy > 99%, PTV maximum dose (Dmax) ≤ 36.75 Gy. Normal tissue constraints were rectum V28 Gy ≤ 40%, V32 Gy ≤ 33%, bladder V32 Gy ≤ 40%, and penile bulb V20 ≤ 90%. Patients were treated on linear accelerators (Siemens Primus, Concord, CA, USA; Elekta Synergy, Stockhold, Sweden) using a “step-and-shoot” intensity-modulated radiotherapy (IMRT) technique with 6 MV photons. Daily image-guidance was performed using orthogonal portal images to identify the fiducial markers and apply any necessary table shifts before treatment. Study 2 The second study included men over 18 years of age with prostate adenocarcinoma and clinical stage T1-2b, Gleason sum ≤ 6, and PSA ≤ 15 ng/mL; or clinical T1-2b, Gleason 7, and PSA ≤ 10 ng/mL. Neoadjuvant ADT was also permitted. Exclusion criteria were the same as for study 1. Patients received 40 Gy in five fractions delivered once per week over 29 days. Patients underwent multiparametric MRI using cardiac surface coil with gadolinium infusion. After the MRI (to reduce artifacts), three gold fiducial markers were inserted transperineally. Patients, then, underwent CT simulation up to 1 week later. Immobilization and bladder and bowel filling were the same as for study 1. CT simulation images were fused with MRI for target delineation. The CTV consisted of the prostate only and OAR were the same as for study 1. The PTV included the CTV and an isotropic 5 mm margin. Planning objectives included a CTV V40 Gy > 99%, PTV V38 Gy > 99%, and PTV Dmax ≤ 42 Gy. Normal tissue constraints were rectum V28 Gy ≤ 20% (up to maximum 40%), V31.8 Gy ≤ 15% (up to maximum 33%), bladder V31.8 Gy ≤ 15% (up to maximum 40%), and penile bulb V20 ≤ 90% (maximum V22.2 Gy < 90%). Patients were treated on the same machines using the same image-guidance protocol as for study 1. Patient Assessments Quality of life was assessed using the Expanded Prostate Cancer Index Composite (EPIC) (9), a validated 50-item instrument that measures prostate cancer-specific QOL. It consists of four summary domains (urinary, bowel, sexual, and hormonal) with function and bother subscales for each domain. Scores were transformed to a 0–100 scale, with higher scores indicating better QOL. Day 0 was defined as the start of radiotherapy. QOL was assessed at baseline and every 6 months thereafter. In study 2, QOL was also assessed at week 5 and month 3. Because QOL was not assessed during the acute time period in study 1, QOL comparisons will be restricted to late effects. Genitourinary (GU) and gastrointestinal (GI) toxicities were measured using the Common Terminology Criteria for Adverse Events Version 3 (CTCAE v3) during the acute period (≤3 months) and the Radiation Therapy Oncology Group (RTOG) late radiation morbidity schema for late effects (>3 months). Toxicities were assessed at baseline, weekly during treatment (only weeks 3 and 5 in study 2), at months 3 and 6, and every 6 months thereafter. PSA was assessed at baseline, months 3 and 6, and every 6 months. Follow-up for all QOL and toxicity endpoints continued for 5 years; biochemical and survival outcomes continued annually until the patient was discharged from follow-up. Statistical Analysis Patient characteristics were summarized as median and interquartile range (IQR) for continuous variables, and proportions for categorical variables. The EPIC scores were calculated as mean ± SD and graphically presented as mean (with 95% confidence intervals) over time. A minimum clinically important change (MCIC) was defined as a decrease in QOL from baseline to follow-up, which exceeded half of the SD of that value at baseline (10). The “average EPIC change” was calculated by (mean of EPIC scores from month 6 to 60 – baseline score) while the “worst EPIC score” was calculated by (lowest of EPIC scores from month 6 to 60 – baseline score). These metrics were chosen to identify the average change in QOL as well as the maximum peak change in QOL, respectively. Actual, non-imputed data were used for analysis. Waterfall plots were created using changes in EPIC scores. In addition, to assess more severe changes in QOL, we compared the proportion of patients with 1 and 2 SD of change. Fisher’s exact test was used to compare categorical data between studies, including proportion of patients experiencing MCIC and proportion with moderate/big problems in specific EPIC items. Wilcoxon rank-sum test was used to compare continuous data. Two-sided p-value <0.05 was considered statistically significant. Generalized estimating equations methodology was used to account for repeated measurements and the impact of covariates on trends in EPIC scores over time (11), with binomial distribution and logit link function. All analyses were conducted using Statistical Analysis Software (SAS version 9.4 for Windows). Results A total of 114 patients were included. There were 84 patients from study 1 and 30 from study 2 with median QOL follow-up of 56 months (IQR 46–60) and 38 months (IQR 32–42), respectively. Overall completion of EPIC questionnaires at 36 months was 70/95 (73.7%). Patient characteristics are described in Table 1 and were balanced between the groups with respect to age, clinical stage, baseline PSA, prostate volume, and baseline IPSS. Neoadjuvant ADT was given to one patient in study 1 and none in study 2. However, owing to different study eligibility criteria, there was a greater proportion of patients with Gleason 6 adenocarcinoma (100 vs. 60%, p < 0.0001) and Prostate Cancer Risk Stratification (ProCaRS) (12) low-risk disease (100 vs. 60%, p < 0.0001) in study 1. Table 1 Patient characteristics. Study 1 (35 Gy/5 F) N = 84 Study 2 (40 Gy/5 F) N = 30 p-Value Median age (IQR), years 67 (61–71) 68 (65–73) 0.27 Clinical stage 0.78  T1a-c 78 (92.9%) 27 (90.0%)  T2a 6 (7.1%) 3 (10.0%) Median PSA at baseline (IQR), ng/mL 5.3 (4.2–7.3) 4.7 (3.5–7.5) 0.33 Gleason score <0.0001  6 84 (100.0%) 18 (60.0%)  7 0 (0%) 12 (40.0%) Risk group (ProCaRS) (12) <0.0001  Low risk 84 (100%) 18 (60.0%)  Low-intermediate risk 0 (0%) 11 (36.7%)  High-intermediate risk 0 (0%) 1 (3.3%) Median TRUS prostate volume (IQR), mL 37 (29–55) 40 (31–53) 0.62 Median IPSS score (IQR) 5 (3–9) 5 (3–11) 0.66 IQR, interquartile range; IPSS, International Prostate Symptom Score; TRUS, transrectal ultrasound. Baseline QOL data are shown in Data Sheet 1. There were no statistically significant differences in baseline EPIC summary domain scores between the two groups. However, there was a lower mean urinary function score in study 1 compared to study 2 (90.4 vs. 95.0%, p < 0.001). The proportion of patients with average and worst changes in EPIC QOL score, which were clinically significant from months 6–60, are shown in Table 2. No statistically significant differences were found between treatment groups. MCIC was reported in average EPIC scores in 19.5 vs. 24.1% for urinary QOL (p = 0.60), 26.8 vs. 41.4% for bowel QOL (p = 0.16), and 42.9 vs. 38.5% in sexual QOL (p = 0.82) for the 35 and 40 Gy groups, respectively. When examining the worst change in EPIC QOL score, 67.1 vs. 58.6% had a urinary MCIC (p = 0.50), 65.9 vs. 62.1% had a MCIC in bowel score (p = 0.82), and 64.9 vs. 50.0% (p = 0.24) had MCIC in sexual QOL for study 1 and 2. Using additional cutoff values of 1 and 2 SD of baseline value to compare the proportion of patients with moderate or large changes in average QOL did not identify any significant differences (Data Sheet 2). Table 2 Proportion of patients with minimum clinically important change (MCIC)a in EPIC quality-of-life scores by treatment group. Average change Worst change Summary domain Study 1 (35 Gy/5 F) Study 2 (40 Gy/5 F) p-Value Study 1 (35 Gy/5 F) Study 2 (40 Gy/5 F) p-Value Urinary 0.60 0.50  No MCIC 66 (80.5%) 22 (75.9%) 27 (32.9%) 12 (41.4%)  MCIC 16 (19.5%) 7 (24.1%) 55 (67.1%) 17 (58.6%) Bowel 0.16 0.82  No MCIC 60 (73.2%) 17 (58.6%) 28 (34.2%) 11 (37.9%)  MCIC 22 (26.8%) 12 (41.4%) 54 (65.9%) 18 (62.1%) Sexual 0.82 0.24  No MCIC 44 (57.1%) 16 (61.5%) 27 (35.1%) 13 (50.0%)  MCIC 33 (42.9%) 10 (38.5%) 50 (64.9%) 13 (50.0%) QOL, quality of life; EPIC, Expanded prostate cancer index composite. aMinimum clinically important change defined as a decrease in quality of life from baseline to follow-up which exceeds half of the standard deviation of the baseline value (10). Waterfall plots of the average change in EPIC urinary, bowel, and sexual scores are shown in Figure 1. Figure 1 Average change in EPIC (A) urinary (B) bowel and (C) sexual quality-of-life scores. Negative and positive changes reflect worse and better QOL after treatment, respectively. Horizontal dotted, dashed, and straight lines indicate 0.5, 1, and 2 SD of the baseline score. Analysis of QOL Over Time Mean EPIC urinary, bowel, and sexual scores over time are shown in Figure 2. The QOL scores are fairly stable over time, and no significant differences exist between treatment groups at any time point, with the exception of lower bowel QOL for patients in study 2 at 6 (p = 0.06) and 12 months (p = 0.04) after treatment. Figure 2 Mean scores (and 95% confidence intervals) for (A) urinary (B) bowel (C) and sexual EPIC quality-of-life domains over time. Longitudinal analysis of risk of MCIC in QOL by treatment group after adjustment for time is shown in Table 3. There was no significant time trend in MCIC of QOL scores with the exception of the sexual domain, in which there was an increasing risk of MCIC in sexual scores over time (OR 1.02, 95% CI 1.01–1.04, p < 0.001). This indicates a 2% increased risk of MCIC in sexual score per month. Treatment group was not significantly associated with MCIC in any of the QOL domains. Table 3 Longitudinal analysis of EPIC quality of life by treatment group. OR 95% CI p-Value Urinary MCIC  Time (months) 0.99 0.97–1.01 0.23  Treatment group (study 1 vs. 2) 0.98 0.54–1.79 0.95 Bowel MCIC  Time (months) 1.00 0.99–1.02 0.99  Treatment group (study 1 vs. 2) 0.61 0.31–1.20 0.15 Sexual MCIC  Time (months) 1.02 1.01–1.04 <0.001  Treatment group (study 1 vs. 2) 1.18 0.54–2.59 0.68 OR, odds ratio; CI, confidence interval; MCIC, minimum clinically important change. The impact of baseline covariates on QOL scores was conducted while adjusting for time and treatment group. Analysis of urinary MCIC included the following covariates: baseline urinary QOL score (p = 0.01), age (years) (p = 0.18), clinical stage (T1a-c vs. T2a) (p = 0.63), log prostate volume (p = 0.73), Gleason score (6 vs. 7) (p = 0.21), log baseline PSA (p = 0.18), and log IPSS (p = 0.49). Only baseline urinary QOL score had a statistically significant effect (OR 1.05, 95% CI 1.01–1.10, p = 0.01), indicating that patients with higher baseline urinary score were at increased risk for MCIC. There was no significant interaction among the covariates. Similarly, analysis of factors affecting bowel QOL included: baseline bowel QOL score (p = 0.18), age (years) (p = 0.84), clinical stage (T1a-c vs. T2a) (p = 0.13), log prostate volume (0.41), Gleason score (6 vs. 7) (p = 0.02), and log baseline PSA (p = 0.55). Only Gleason score was significant (OR 0.23, 95% CI 0.07–0.81, p = 0.02), with patients with Gleason 7 having lower risk of MCIC in bowel score compared to Gleason 6. Severity of QOL Impairment Patients were asked “Overall, how big a problem has your (urinary function/bowel habits/sexual function) been for you during the last 4 weeks?” with answers dichotomized into no problem/very small problem/small problem vs. moderate problem/big problem. Results are listed in Table 4. There were no significant differences between the two studies at baseline or last follow-up. Few (<4%) patients reported moderate/severe problems in overall urinary function or bowel habits at last follow-up in both studies. Table 4 Overall problematic nature of quality-of-life domains at baseline and last follow-up. Baseline N (%) Last follow-up N (%) Study 1 Study 2 p* Study 1 Study 2 p* Urinary function 0.39 0.98  No/very small/small problem 77 (95.1) 27 (90.0) 81 (96.4) 28 (96.6)  Moderate/severe problem 4 (4.9) 3 (10.0) 3 (3.6) 1 (3.5) Bowel habits 0.45 0.98  No/very small/small problem 80 (98.8) 28 (96.6) 81 (96.4) 28 (96.6)  Moderate/severe problem 1 (1.2) 1 (3.5) 3 (3.6) 1 (3.5) Sexual function 0.91 0.92  No/very small/small problem 60 (76.9) 22 (75.9) 60 (74.1) 21 (75.0)  Moderate/severe problem 18 (23.1) 7 (24.1) 21 (25.9) 7 (25.0) *Comparison between study 1 and study 2. Discussion This is one of the first studies to examine the effect of dose-escalation in prostate SBRT on QOL endpoints. The results show that when increasing the dose from 35 to 40 Gy, delivered in five fractions once per week, there is no significant difference in long-term QOL outcomes. Unique features of this analysis include prospective data collection and longer follow-up. The results of the present study should be compared to existing literature within the context of differences in treatment technique and total dose. There is conflicting data regarding the impact of conventionally fractionated, dose-escalated radiotherapy on QOL endpoints using older radiation techniques (13–16). In the Proton Radiation Oncology Group 9509 randomized study of 70.2 vs. 79.2 Gy, there were no differences in urinary, bowel, or sexual outcomes using the Prostate Cancer Symptom Indices assessed at a median 9.4 years (13). Similarly, in the GETUG-06 trial of 70 vs. 80 Gy, there were no differences in QOL using the European Organization for Research and Treatment of Cancer (EORTC) QOL Questionnaire and prostate-specific module (14). In contrast, other studies have shown detriment in specific patient-reported outcomes after dose-escalated radiotherapy. In the MD Anderson randomized trial of 70 vs. 78 Gy, there was a significant increase in frequency of bowel movements associated with high-dose radiotherapy at 3 years after treatment (p = 0.03) (15). No differences were found in the other bowel-related, urinary, or sexual QOL. Also, in a detailed analysis of GI outcomes from the UK RT01 randomized trial comparing 64 vs. 74 Gy, QOL assessment using the University of Los Angeles Prostate Cancer Index showed no differences in diarrhea, bowel distress, or bowel problems (16). However, there was an increased risk of moderate abdominal pain (HR 1.53, 95% CI 1.13–2.06) and severe rectal urgency (HR 1.64, 95% CI 1.11–2.42) observed with higher dose. Many studies have reported on QOL outcomes after prostate SBRT (5, 17), but few have compared such data with respect to the effect of increasing dose. The previously mentioned dose-escalation study of 45, 47.5, and 50 Gy in five fractions was not powered to compare QOL outcomes between treatment arms (5). However, comparison of QOL measured by the EPIC instrument did not identify any difference in bowel QOL between the 45 and 50 Gy arms (p = 0.5), but there was significantly worse bowel QOL for the 47.5 Gy dose level (p = 0.01). No differences were found in urinary QOL between the three dose levels. In a separate analysis of physician-rated toxicities from the present study, dose-escalated SBRT was associated with greater risk of grade 2+, but not grade 3+, toxicities (7). Other studies have found similar associations between higher SBRT dose and toxicities (18), and there are several possible explanations for why it did not result in worse patient-reported outcomes (19), in the present study. First, the increase in biologically effective dose to OAR is dependent upon the alpha–beta ratio. For the rectum, the alpha–beta ratio for grade≥2 late toxicity has been estimated to be 4.8 Gy, but with a wide 68% CI ranging from 0.6 to 46 Gy (20). The corresponding difference between dose levels in the present study in equivalent 2 Gy fractions would be 14.6 Gy (68% CI 6.4–30.0), but may be as low as 6.4 Gy, if the alpha–beta ratio for bowel QOL is similar to that for late toxicity. Second, the relationship between dose and QOL outcomes has not been well characterized, and differences in dose at 35 to 40 Gy may not be expected to result in significant changes in QOL (i.e., on the “flat” segment of the dose-complication curve). Third, although this study used the validated EPIC instrument and commonly reported threshold of one-half SD (10) to identify a MCIC, it may not have been sensitive enough to detect small changes in QOL. Finally, patients who received higher dose may have had worse treatment-related toxicity, but adapted to the effects of altered function, and not reported any impairment on the QOL questionnaires (21). The QOL outcomes reported in this study are similar to those reported by other groups. In a multi-institutional study of 864 patients treated with prostate SBRT to a median dose of 36.25 Gy in four or five fractions (17), EPIC scores were similar to the present study with long-term urinary, bowel, and sexual scores of 80–90, 85–95, and 20–50. Although 20–40% of patients in our study reported a change in long-term QOL, which was clinically detectable, the majority of changes were mild to moderate in severity. This is a limitation of the MCIC definition in that it does not distinguish minimal from severe changes in QOL. To assess severe problems in QOL, we also analyzed specific EPIC items, which asked about overall patient function. The proportion of patients who reported moderate or severe problems in overall urinary function or bowel habits was <4% at both dose levels, which is reassuring. Also while 25% of patients experienced moderate or severe problems in sexual function at last follow-up, this is similar to values observed at baseline. We analyzed the effect of patient, tumor, and treatment-related factors to identify variables associated with worse post-treatment QOL. Only a higher baseline urinary QOL score was associated with greater risk of clinically significant change in QOL after treatment. This finding is similar to other QOL studies in which patients with greater current function have more potential for loss of function, compared to patients with intermediate or poor baseline function (22). Gleason 7 was also found to be associated with greater risk of MCIC in bowel QOL. The reason for this is unclear and requires further investigation, but may be a type I error. Limitations of our study should be noted. As QOL was not assessed in study 1 during the first 3 months after treatment, acute changes in QOL could not be compared. Also, the smaller sample size of patients receiving 40 Gy limited the power to detect small differences between study groups. We have completed accrual to a multicentre, phase 2 randomized trial of 40 Gy in five fractions, delivered every other day vs. once weekly (23). One hundred fifty-two patients had acute and late QOL measured, and we will report these data as they mature. Finally, radiotherapy planning was different between the two cohorts. CTV volumes were based on CT and MRI for patients receiving 40 Gy, whereas it was based on CT only imaging for those treated with 35 Gy. Offset against this, the PTV margin was 5 mm for the 40 Gy study and 4 mm for the 35 Gy study. Compared to CT-based planning, MRI-defined prostate volumes are 21–30% (24–26) smaller, which has been shown to reduce rectal and bladder doses (26). However, these dosimetric advantages have only translated into reduced urinary, but not rectal, toxicity in reports with clinical outcomes (24, 26). No studies have compared QOL endpoints with MRI vs. CT-based planning. Conclusion Increased dose for prostate SBRT from 35 to 40 Gy, delivered in five once-weekly fractions, was not associated with a clinically significant difference in late QOL outcomes. Long-term effects on urinary and bowel QOL remain mild to moderate, with most patients reporting the same QOL scores in late follow-up compared to baseline. Author Contributions HQ, HM, and AL participated in the study conception and design, data analysis, and interpretation of results. PC, GP, and LZ were involved in the data analysis and interpretation. AM, LD, and AD participated in the collection and assembly of data. All authors were involved in writing and final approval of the 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. This study was funded by a joint Abbott-Canadian Association of Radiation Oncology Uro-Oncologic Radiation Award. It was presented in part at the 2012 American Society for Radiation Oncology (ASTRO) Annual Meeting. Supplementary Material The Supplementary Material for this article can be found online at http://journal.frontiersin.org/article/10.3389/fonc.2016.00185 Click here for additional data file. Click here for additional data file. Abbreviations ADT, androgen deprivation therapy; CTV, clinical target volume; Dmax, maximum dose; EORTC, European Organization for Research and Treatment of Cancer; EPIC, Expanded Prostate Cancer Index Composite; GI, gastrointestinal; GU, genitourinary; HR, hazard ratio; IMRT, intensity-modulated radiotherapy; IPSS, International Prostate Symptom Score; IQR, interquartile range; MCIC, minimum clinically important change; OAR, organs at risk; ProCaRS, prostate cancer risk stratification; PSA = prostate-specific antigen; PTV, planning target volume; QOL, quality of life; RTOG, Radiation Therapy Oncology Group; SBRT, stereotactic body radiation therapy. ==== Refs References 1 Miralbell R Roberts SA Zubizarreta E Hendry JH Dose-fractionation sensitivity of prostate cancer deduced from radiotherapy outcomes of 5,969 patients in seven international institutional datasets: alpha/beta = 1.4 (0.9-2.2) Gy . Int J Radiat Oncol Biol Phys (2012 ) 82 (1 ):e17 –24 .10.1016/j.ijrobp.2010.10.075 21324610 2 Yu JB Cramer LD Herrin J Soulos PR Potosky AL Gross CP Stereotactic body radiation therapy versus intensity-modulated radiation therapy for prostate cancer: comparison of toxicity . J Clin Oncol (2014 ) 32 (12 ):1195 –201 .10.1200/JCO.2013.53.8652 24616315 3 Loblaw A Cheung P D’Alimonte L Deabreu A Mamedov A Zhang L Prostate stereotactic ablative body radiotherapy using a standard linear accelerator: toxicity, biochemical, and pathological outcomes . Radiother Oncol (2013 ) 107 (2 ):153 –8 .10.1016/j.radonc.2013.03.022 23647750 4 Madsen BL Hsi RA Pham HT Fowler JF Esagui L Corman J . Stereotactic hypofractionated accurate radiotherapy of the prostate (SHARP), 33.5 Gy in five fractions for localized disease: first clinical trial results . Int J Radiat Oncol Biol Phys (2007 ) 67 (4 ):1099 –105 .10.1016/j.ijrobp.2006.10.050 17336216 5 Boike TP Lotan Y Cho LC Brindle J DeRose P Xie XJ Phase I dose-escalation study of stereotactic body radiation therapy for low- and intermediate-risk prostate cancer . J Clin Oncol (2011 ) 29 (15 ):2020 –6 .10.1200/JCO.2010.31.4377 21464418 6 Gravis G Marino P Joly F Oudard S Priou F Esterni B Patients’ self-assessment versus investigators’ evaluation in a phase III trial in non-castrate metastatic prostate cancer (GETUG-AFU 15) . Eur J Cancer (2014 ) 50 (5 ):953 –62 .10.1016/j.ejca.2013.11.034 24424105 7 Musunuru HB Quon H Davidson M Cheung P Zhang L D’Alimonte L Dose-escalation of five-fraction SABR in prostate cancer: toxicity comparison of two prospective trials . Radiother Oncol (2016 ) 118 (1 ):112 –7 .10.1016/j.radonc.2015.12.020 26796591 8 Tang CI Loblaw DA Cheung P Holden L Morton G Basran PS Phase I/II study of a five-fraction hypofractionated accelerated radiotherapy treatment for low-risk localised prostate cancer: early results of pHART3 . Clin Oncol (R Coll Radiol) (2008 ) 20 (10 ):729 –37 .10.1016/j.clon.2008.08.006 18838256 9 Wei JT Dunn RL Litwin MS Sandler HM Sanda MG . Development and validation of the expanded prostate cancer index composite (EPIC) for comprehensive assessment of health-related quality of life in men with prostate cancer . Urology (2000 ) 56 (6 ):899 –905 .10.1016/S0090-4295(00)00858-X 11113727 10 Norman GR Sloan JA Wyrwich KW . Interpretation of changes in health-related quality of life: the remarkable universality of half a standard deviation . Med Care (2003 ) 41 (5 ):582 –92 .10.1097/00005650-200305000-00007 12719681 11 Liang K-Y Zeger SL Longitudinal data analysis using generalized linear models . Biometrika (1986 ) 73 (1 ):13 –22 .10.1093/biomet/73.1.13 12 Rodrigues G Lukka H Warde P Brundage M Souhami L Crook J The prostate cancer risk stratification (ProCaRS) project: recursive partitioning risk stratification analysis . Radiother Oncol (2013 ) 109 (2 ):204 –10 .10.1016/j.radonc.2013.07.020 24231239 13 Talcott JA Rossi C Shipley WU Clark JA Slater JD Niemierko A Patient-reported long-term outcomes after conventional and high-dose combined proton and photon radiation for early prostate cancer . JAMA (2010 ) 303 (11 ):1046 –53 .10.1001/jama.2010.287 20233822 14 Beckendorf V Guerif S Le Prisé E Cosset JM Bougnoux A Chauvet B 70 Gy versus 80 Gy in localized prostate cancer: 5-year results of GETUG 06 randomized trial . Int J Radiat Oncol Biol Phys (2011 ) 80 (4 ):1056 –63 .10.1016/j.ijrobp.2010.03.049 21147514 15 Little DJ Kuban DA Levy LB Zagars GK Pollack A . Quality-of-life questionnaire results 2 and 3 years after radiotherapy for prostate cancer in a randomized dose-escalation study . Urology (2003 ) 62 (4 ):707 –13 .10.1016/S0090-4295(03)00504-1 14550448 16 Syndikus I Morgan RC Sydes MR Graham JD Dearnaley DP MRC RT01 collaborators . Late gastrointestinal toxicity after dose-escalated conformal radiotherapy for early prostate cancer: results from the UK Medical Research Council RT01 trial (ISRCTN47772397) . Int J Radiat Oncol Biol Phys (2010 ) 77 (3 ):773 –83 .10.1016/j.ijrobp.2009.05.052 19836155 17 King CR Collins S Fuller D Wang PC Kupelian P Steinberg M Health-related quality of life after stereotactic body radiation therapy for localized prostate cancer: results from a multi-institutional consortium of prospective trials . Int J Radiat Oncol Biol Phys (2013 ) 87 (5 ):939 –45 .10.1016/j.ijrobp.2013.08.019 24119836 18 Viani GA Stefano EJ Afonso SL . Higher-than-conventional radiation doses in localized prostate cancer treatment: a meta-analysis of randomized, controlled trials . Int J Radiat Oncol Biol Phys (2009 ) 74 (5 ):1405 –18 .10.1016/j.ijrobp.2008.10.091 19616743 19 Quon HC Loblaw DA Cheung P Tang CI Pang G Morton G Dose escalation of 5-fraction radiation therapy for prostate cancer: quality of life comparison of 2 prospective trials . International Journal of Radiation Oncology Biology Physics (2012 ) 84 (3 ):S148 –9 .10.1016/j.ijrobp.2012.07.383 20 Tucker SL Thames HD Michalski JM Bosch WR Mohan R Winter K Estimation of alpha/beta for late rectal toxicity based on RTOG 94-06 . Int J Radiat Oncol Biol Phys (2011 ) 81 (2 ):600 –5 .10.1016/j.ijrobp.2010.11.080 21377288 21 Korfage IJ Hak T de Koning HJ Essink-Bot ML Patients’ perceptions of the side-effects of prostate cancer treatment – a qualitative interview study . Soc Sci Med (2006 ) 63 (4 ):911 –9 .10.1016/j.socscimed.2006.01.027 16798130 22 Chen RC Clark JA Talcott JA . Individualizing quality-of-life outcomes reporting: how localized prostate cancer treatments affect patients with different levels of baseline urinary, bowel, and sexual function . J Clin Oncol (2009 ) 27 (24 ):3916 –22 .10.1200/JCO.2008.18.6486 19620493 23 Quon HC Ong A Cheung P Chu W Chung HT Vesprini D PATRIOT trial: randomized phase II study of prostate stereotactic body radiotherapy comparing 11 versus 29 days overall treatment time . J Clin Oncol (2015 ) 33 (7_suppl ):abstr6 . 24 Sander L Langkilde NC Holmberg M Carl J . MRI target delineation may reduce long-term toxicity after prostate radiotherapy . Acta Oncol (2014 ) 53 (6 ):809 –14 .10.3109/0284186X.2013.865077 24358954 25 Rasch C Barillot I Remeijer P Touw A van Herk M Lebesque JV . Definition of the prostate in CT and MRI: a multi-observer study . Int J Radiat Oncol Biol Phys (1999 ) 43 (1 ):57 –66 .10.1016/S0360-3016(98)00351-4 9989514 26 Ali AN Rossi PJ Godette KD Martin D Liauw S Vijayakumar S Impact of magnetic resonance imaging on computed tomography-based treatment planning and acute toxicity for prostate cancer patients treated with intensity modulated radiation therapy . Pract Radiat Oncol (2013 ) 3 (1 ):e1 –9 .10.1016/j.prro.2012.04.005 24674267

PMC005xxxxxx/PMC5003003.txt

==== Front Biores Open AccessBiores Open AccessbioresBioResearch Open Access2164-78442164-7860Mary Ann Liebert, Inc. 140 Huguenot Street, 3rd FloorNew Rochelle, NY 10801USA 10.1089/biores.2016.002310.1089/biores.2016.0023Original Research ArticleAssociation of Visual Acuity and Cognitive Impairment in Older Individuals: Fujiwara-kyo Eye Study Mine Masashi 1Miyata Kimie 1Morikawa Masayuki 2Nishi Tomo 1Okamoto Nozomi 3Kawasaki Ryo 4Yamashita Hidetoshi 5Kurumatani Norio 3Ogata Nahoko 1,*1 Department of Ophthalmology, Nara Medical University, Kashihara, Nara, Japan.2 Mie Prefectural Mental Care Center, Tsu, Mie, Japan.3 Department of Community Health and Epidemiology, Nara Medical University, Kashihara, Nara, Japan.4 Department of Public Health, Graduate School of Medical Science, Yamagata University, Yamagata, Japan.5 Department of Ophthalmology and Visual Science, Yamagata University Faculty of Medicine, Yamagata, Japan.* Address correspondence to: Nahoko Ogata, MD, PhD, Department of Ophthalmology, Nara Medical University, 840 Shijo-cyo, Kashihara, Nara, 634-8522, Japan, E-mail: ogata@naramed-u.ac.jp01 8 2016 2016 01 8 2016 5 1 228 234 © Masashi Mine et al. 2016; Published by Mary Ann Liebert, Inc.2016This Open Access article is distributed under the terms of the Creative Commons License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.Abstract Both visual impairment and cognitive impairment are essential factors that determine the quality of life in the aged population. The aim of this study was to determine if a correlation existed between visual acuity and cognitive impairment in an elderly Japanese population. The Fujiwara-kyo Eye Study was a cross-sectional study of individuals aged ≥68 years who lived in Nara Prefecture of Japan. Participants underwent ophthalmological examinations and cognitive function test. A mild visual impairment was defined as having a best corrected visual acuity (BCVA) >0.2 logarithm of the minimum angle of resolution (logMAR) units in the better eye. Cognitive impairment was defined as having a Mini-Mental State Examination (MMSE) score of ≤23 points. A total to 2818 individuals completed the examinations. The mean age of the participants was 76.3 ± 4.8 years (mean ± standard deviation). The mean BCVA of the better eye was −0.02 ± 0.13 logMAR units and 6.6% subjects were classified as being mildly visually impaired. The mean MMSE score was 27.3 ± 2.3 and 5.7% subjects were classified as being cognitively impaired. The proportion of subjects with cognitive or moderate visual impairment increased with age, and there was a significant correlation between the visual acuity and MMSE score (r = −0.10, p < 0.0001). Subjects with mild visual impairments had 2.4 times higher odds of having cognitive impairment than those without visual impairment (odds ratio 2.4, 95% confidence interval, 1.5–3.8, p < 0.001) after adjusting for age, sex, and length of education. We conclude that it may be important to maintain good visual acuity to reduce the risk of having cognitive impairment. Keywords:  agingneuroscienceregeneration ==== Body Introduction The proportion of older individuals in the population has been rapidly increasing globally. In developed countries, the proportion of the population ≥65 years was 16% in 2010, and it has been estimated that the proportion will increase to 27% in 2060.1 Japan is considered to be the most aged country in the world and the Japanese Cabinet Office has reported that the nationwide proportion of individuals ≥65 years old was 23%, and that of ≥75 years old was 11% in 2010.2 In 2060, these numbers are estimated to increase to 40% and 27%, respectively. In such an aged society, it is important that older individuals maintain their health and have a good quality of life (QOL) even at a later stage of life when various age-related functional and physical disorders develop. Both visual and cognitive impairments are essential factors that limit the activities of daily living and the QOL. According to the Japanese Ministry of Health, Labour and Welfare, the prevalence of dementia in persons aged ≥65 years was estimated to be 15% in 2010.3 Iwano et al. reported that the prevalence of vision impairments, defined as a having corrected Snellen visual acuity of <10/20 in the better eye, was 5.6% in subjects aged 70 to 79 years in 2004.4 They also reported that the risk of developing visual impairments increased by 3.9-fold for every 10-year increase in age. Earlier studies5–8 and population-based epidemiological studies involving older individuals9–14 have shown that there is an association between visual impairment and cognitive function. Such findings have been reported by the Berlin Aging Study,9 Australian Longitudinal Study of Aging,10,11 Maastricht Aging Study,12 Leiden 85+ Study,13 and Blue Mountains Eye Study.14 However, there are still contradictions in linking visual impairment and cognitive function.15,16 To date, there has been no study reporting the link between visual impairment and cognitive function in Japan.17–19 Because Japan has the longest life span in the world,20 it is important to determine whether there is a significant correlation between visual impairment and cognitive function in the elderly Japanese. Thus, the purpose of this study is to determine whether there is an association between visual acuity and cognitive function in an elderly Japanese population. Materials and Methods Participants The Fujiwara-kyo Study was a cohort study conducted in Nara and undertaken to investigate the functional capacity and the QOL of a community of elderly population.21,22 Study participants were ≥65 years, living in their own home, and residents of Nara Prefecture. The participants were volunteers who responded to our recruitment announcements. The Fujiwara-kyo Study was initially performed in 2007 with plans to do follow-up examinations every 5 years. The ophthalmological survey (Fujiwara-kyo Eye Study) was begun at the second stage in February to November 2012. Therefore, most of the individuals were ≥70 years because they were recruited at 2007 when they were 65 years or older, but new subjects aged ≥65 years (61 individuals) were also recruited during this second stage. We used data from the examinations done in 2012 as a cross-sectional study. All of the subjects underwent a basic interview to obtain their sociodemographic data, their general medical condition, and treatment histories. They also underwent systemic examinations, including physical anthropometric assessments, physical fitness assessments, and blood tests. These surveys were conducted in accordance with the tenets of the Declaration of Helsinki, and the protocol was approved by the Ethics Review Board of Nara Medical University. A signed informed consent form was obtained from all participants. Ophthalmological examinations The uncorrected and corrected visual acuities of both eyes were measured with a Landolt ring chart at 5 m with correction of any refractive errors determined by an automatic refractometer and keratometer (ARK-700A; Nidek, Aichi, Japan). The visual acuity was measured according to the standard of the International Organization for Standardization.23 The decimal best-corrected visual acuity (BCVA) of the better eye was converted to the logarithm of the minimum angle of resolution (logMAR) units for statistical analyses. The degree of visual impairment was classified as recommended by the International Council of Ophthalmology report24 and the Japanese driver's license requirement on vision. Thus, subjects with BCVA of >0.2 logMAR units in the better eye were classified as being mildly visually impaired. Cognitive function test The Mini-Mental State Examination (MMSE) was used for the cognitive function test. This test was developed by Folstein et al. in 1975 and is commonly used for dementia screening.25 It consists of five downstream items of orientation, memory, attentiveness for calculations, speech function, and design capacity. This test was performed by verbal questioning of 5- to 10-min duration by skilled clinical psychologists. The maximum score for the MMSE is 30 points, and individuals with a score of ≤23 points were classified as having cognitive impairments. We also analyzed the MMSE excluding the following vision-related five items: “naming two objects,” “following a 3-step command,” “reading and following instruction,” “writing a sentence,” and “visual reconstruction” and the maximum score for this was 22 points. This was conducted because Reischies and Geiselmann5 and Busse et al.8 reported that these five items of the MMSE were dependent on the visual acuity. Statistical analyses The differences in the sociodemographic and health characteristics of individuals with and without mild visual impairments were analyzed by the chi-square and unpaired t-tests. The correlations between the BCVA of the better eye and the MMSE score were determined by Pearson's correlation coefficient. A scatter plot was made with the visual acuity in logMAR units plotted on the abscissa and the MMSE scores plotted on the ordinate. The best fit regression line was calculated for these data. The differences of mean MMSE scores between individuals with or without mild visual impairment were analyzed by analysis of covariance (ANCOVA) with adjustments for age, sex, length of education, and history of stroke. Multiple logistic regression analyses were used to determine the association between visual and cognitive impairment with adjustments for age, sex, length of education, and history of stroke. The odds ratios (OR) and 95% confidence intervals (CI) for having cognitive impairment in terms of visual acuity were calculated. Statistical analyses were performed with the SPSS (version 22.0; SPSS, Inc., Chicago, IL). Two-tailed p-values were used in all of the analyses. A p-value <0.05 was taken to be significant. Results There were 2873 individuals who participated in the Fujiwara-kyo Eye Study. Of these, 2818 individuals completed all of the examinations (98.1%) and were used in the statistical analyses. Fifty-five participants were excluded because the vision or cognitive function tests were not performed or were incomplete. The age of the 2818 participants ranged from 68 to 100 years with a mean of 76.3 ± 4.8 years (±standard deviation [SD]) (Table 1). There were 1486 (52.7%) men and 1332 women. There were 1166 subjects whose age was ≤74 years, 965 subjects aged 75 to 79 years, 494 subjects aged 80 to 84 years, 166 subjects aged 85 to 90 years, and 27 subjects aged ≥90 years. There were 2172 (77.1%) subjects whose level of education was ≤12 years, and there were 646 (22.9%) subjects whose level of education was ≥13 years (Table 1). Table 1. Distribution of Characteristics of the Participants (n = 2818) Mean age ± SD (range), years 76.3 ± 4.8 (68–100) Men/women, n (%) 1486 (52.7)/1332 (47.3) Mean BCVA in logMAR units in the better eye ± SD (range) −0.02 ± 0.13 (−0.30–1.39) Mild visual impairment (%) 187 (6.6) Mean MMSE score ± SD (range) 27.3 ± 2.3 (14–30) Cognitive impairment (%) 160 (5.7) Mean MMSE score excluding 5 vision-related itemsa ± SD (range) 19.4 ± 2.2 (8–22) Length of education with less than 12 years (%) 2172 (77.1) MMSE score ranged from 0 to 30. MMSE score excluding five items ranged from 0 to 22. Cognitive impairment defined as MMSE score ≤23. Mild visual impairment defined as logMAR units >0.2 in the better eye. a Vision-related five items: “naming two objects,” “following a 3-step command,” “reading and following instruction,” “writing a sentence,” and “visual reconstruction.” BCVA, best corrected visual acuity; logMAR, logarithm of the minimum angle of resolution; MMSE, Mini-Metal State Examination; SD, standard deviation. The BCVA of the better eye ranged from −0.30 to 1.39 logMAR units with a mean of −0.02 ± 0.13 logMAR units. Of the 2818 participants, 187 (6.6%) were classified as being mildly visually impaired, that is, having BCVA >0.2 logMAR units in the better eye (Table 1). The MMSE scores ranged from 14 to 30 points with a mean of 27.3 ± 2.3. Of the 2818 participants, 160 individuals (5.7%) were classified as being cognitively impaired; that is, MMSE score ≤23 points. The MMSE scores excluding five vision-related items with a maximum score of 22 points ranged from 8 to 22 points with a mean of 19.4 ± 2.2 (Table 1). The number of subjects with mild visual impairment was 45 (3.9%) of 1166 subjects in the age ≤74 years, 68 (7.0%) of 965 subjects in the age 75 to 79 years, 44 (8.9%) of 494 subjects in the age 80 to 84 years, 22 (13.3%) of 166 subjects in the age 85 to 90 years, and 8 (29.6%) of 27 subjects in the age ≥90-year groups. The number of subjects with cognitive impairment was 36 (3.1%) of 1166 subjects in the age ≤74 years, 55 (5.7%) of 965 subjects in the age 75 to 79 years, 44 (8.9%) of 494 subjects in the age 80 to 84 years, 18 (10.8%) of 166 subjects in the age 85 to 90 years, and 7 (28.9%) of 27 subjects in the age ≥90-year groups. The proportion of subjects with cognitive or mild visual impairment increased with increasing age. The mean age of the 160 subjects with cognitive impairment was 79.2 ± 5.4 years, and that of the 2658 subjects without cognitive impairment was 76.1 ± 4.7 years (Table 2). The mean age of the 187 subjects (6.6%) with mild visual impairment was 78.7 ± 5.7 years, and that of 2631 subjects (93.4%) without mild visual impairment was 76.1 ± 4.7 years. The age of subjects in whom the cognitive function or visual acuity was classified as being impaired was significantly older than those without the impairments (p < 0.001) (Table 2). Table 2. Sociodemographic and Health Characteristics of Participants with Cognitive Impairment and Mild Visual Impairment   Cognitive impairment Mild visual impairment Factors Absent (n = 2658) Present (n = 160) p Absent (n = 2631) Present (n = 187) p Mean age ± SD, years 76.1 ± 4.7 79.2 ± 5.4 <0.001 76.1 ± 4.7 78.7 ± 5.7 <0.001 Men, % 52.4 58.8 0.116 53.7 38.5 <0.001 Length of education, less than 12 years, % 76.5 86.3 0.004 76.6 84.0 0.021 History of disease, % Stroke 6.0 5.0 0.447 5.9 9.6 0.037 Cerebral infarction 4.7 5.0 0.863 4.5 7.5 0.065 Angina 8.3 8.1 0.931 8.4 6.4 0.332 Myocardial infarction 2.7 3.1 0.779 2.8 2.1 0.586 Diabetes mellitus 14.8 16.3 0.605 14.8 16.0 0.632 Hypertension 46.0 39.5 0.118 45.2 50.8 0.143 Hypercholesterolemia 26.1 19.4 0.059 26.1 20.9 0.116 Current smoking 5.6 9.0 0.075 5.8 5.5 0.845 p-Values are for chi-square tests or unpaired t-tests performed. Cognitive impairment defined as MMSE score ≤23. Mild visual impairment defined as logMAR units >0.2 in the better eye. Cognitive impairments were significantly associated with age and length of education. Mild visual impairments were significantly associated with age, sex, length of education, and a history of stroke (Table 2). The correlation between the BCVA of the better eye and the MMSE score was significant (r = −0.10, p < 0.0001, Pearson's correlation coefficient, Fig. 1). FIG. 1. Relationship between visual acuity and MMSE score. The MMSE score is significantly correlated with the BCVA of the better eye (r = −0.10, p < 0.0001). BCVA, best corrected visual acuity; MMSE, Mini-Mental State Examination. The mean MMSE score in subjects with mild visual impairments was significantly lower than in those without mild visual impairments after adjusting for age, sex, length of education, and history of stroke (26.8 vs. 27.3, p < 0.05, Table 3). After excluding the five vision-related items in the MMSE, the significant association between the MMSE score and the presence of mild visual impairments remained after adjusting for age, sex, length of education, and history of stroke (19.1 vs.19.4, p < 0.05, Table 3). Table 3. Adjusted Mean MMSE Score (Standard Error) by the Presence or Absence of Mild Vision Impairment     Mean MMSE score (standard error)         Mild visual impairment     MMSE items Range of MMSE scores Absent Present F statistic p All items 0–30 27.3 (0.05) 26.8 (0.17) 6.359 0.012 Excluding vision-related 5 items 0–22 19.4 (0.04) 19.1 (0.16) 4.287 0.038 ANCOVA adjusted for age, sex, length of education, and history of stroke. Mild visual impairment defined as logMAR units >0.2 in the better eye. The subjects were classified into four groups by the BCVA, and the prevalence of cognitive impairment was significantly higher in the group with poorer visual acuity (Table 4). The prevalence of cognitive impairment in subjects with BCVA of 0 to 0.1 logMAR units in the better eye was twofold higher than in subjects with BCVA of ≤0 logMAR units (8.0% vs. 4.5%). In addition, compared to subjects with BCVA of ≤0 logMAR units, the OR for cognitive impairment increased to 1.8 (1.2–2.7, 95% CI, p = 0.007) in subjects with BCVA of 0 to 0.1 logMAR units and to 3.3 (2.1–5.4, 95% CI, p = 0.005) in subjects with BCVA of >0.2 logMAR units after adjusting for age, sex, length of education, and history of stroke (Table 4). Table 4. Associations Between Visual Acuity and Cognitive Impairment     Odds ratio (95%CI) BCVA in the better eye, logMAR Cognitive impairment (%) Unadjusted p Adjusteda p ≤0 96/2135 (4.5) 1.0 (reference)   1.0 (reference)   0–0.1 31/386 (8.0) 1.9 (1.2–2.8) 0.004 1.8 (1.2–2.7) 0.007 0.1–0.2 8/110 (7.3) 1.9 (0.9–3.8) 0.083 1.7 (0.8–3.5) 0.187 >0.2 25/187 (13.4) 3.3 (2.1–5.2) <0.001 3.3 (2.1–5.4) 0.005 a Multivariate regression model, adjusting for age, sex, length of education, and history of stroke. 95% CI, 95% confidence interval. The prevalence of cognitive impairment was higher in subjects with mild visual impairment than in those without mild visual impairment (13.3% vs. 5.1%, p < 0.001) (Table 5). When mild visual impairments (>0.2 logMAR) were present, the OR for the presence of cognitive impairment was 2.9 (1.8–4.5, 95% CI, p < 0.001). Age and length of education also significantly affected the presence of cognitive impairments (OR = 1.7 per 5 years of age, 95% CI, 1.4–1.9, p < 0.001; OR = 1.9 between education ≤12 years, 95% CI, 1.2–3.0, p < 0.001, respectively). After adjusting for age, sex, length of education, and history of stroke, mild visual impairment was significantly associated with a higher odds ratio of having cognitive impairment (OR = 2.4, 95% CI, 1.5–3.8, p < 0.001, Table 5). Table 5. Factors Associated with Cognitive Impairment     Odds ratio (95% CI) Factors Cognitive impairment (%) Unadjusted p Adjusteda p Age (each 5-year increase)   1.7 (1.4–1.9) <0.001 1.6 (1.4–1.9) <0.001 Sex  Men 94/1486 (6.3) 1.0 (reference)   1.0 (reference)    Women 66/1332 (5.0) 0.8 (0.6–1.1) 0.117 0.7 (0.5–1.0) 0.039 Length of education ≤12 years  No 22/646 (3.4) 1.0 (reference)   1.0 (reference)    Yes 138/2172 (6.4) 1.9 (1.2–3.0) 0.005 2.0 (1.2–3.1) 0.005 History of stroke  No 152/2646 (5.7) 1.0 (reference)   1.0 (reference)    Yes 8/172 (4.7) 0.8 (0.4–1.5) 0.448 1.0 (0.5–1.8) 0.956 Mild visual impairment  No 135/2631 (5.1) 1.0 (reference)   1.0 (reference)    Yes 25/187 (13.3) 2.9 (1.8–4.5) <0.001 2.4 (1.5–3.8) <0.001 a Multivariate regression model, adjusting for age, sex, length of education, and history of stroke. Discussion Our results confirmed that the prevalence of both cognitive impairment and visual impairment increases with increasing age. The subjects with visual impairment defined as BCVA of >0.2 logMAR units in the better eye had significantly lower MMSE scores than those without visual impairment. In addition, a significant correlation was found between the MMSE score and the BCVA of the better eye. A lower BCVA was associated with higher prevalence of cognitive impairment, and even a slight visual reduction (0 to 0.1 logMAR units) was associated with 1.8 times higher odds of having a cognitive impairment. Among the subjects with mild visual impairment (>0.2 logMAR units), the odds of having cognitive impairment were 2.4 times higher than those without mild visual impairment after age, sex, length of education, and history of stroke were adjusted. There have been earlier studies reporting an association between visual impairment and cognitive impairment. The Blue Mountain Eye Study reported a correlation between visual impairment and cognitive impairment.14 In addition, the Maastricht Aging Study,12 the Leiden 85+ Study,13 and the Aging, Demographics, and Memory study26 also found the correlation between visual impairment and cognitive impairment. Our results are comparable to these earlier studies. Our new finding is that the association was consistent in an older population whose age was ≥68 years with a mean of 76.3 years with a relatively larger sample number in an epidemiological study of 2818 subjects. This study is the first to report such association in elderly subjects in Japan. The association between visual impairment and cognitive impairment has not been observed in some studies.15,16 We suggest that this inconsistency may be because of the differences in the method of how the visual acuity was measured.12–14,26 For example, the vision was obtained by the self-reporting of visual acuity and no vision test was performed in one study.26 In another study, the vision was measured without refractive correction or only with their daily used spectacles.12 In our study, the visual acuities were measured with the corrective lenses based on the refraction determined by skilled orthoptists. Thus, the visual acuities in our study are likely to be more accurate, which makes the conclusions more reliable. To evaluate the cognitive functions, it has been pointed out that the visual acuity-dependent factors were included in the cognitive function test.5,8 Thus, Diaz et al.15 and Killen et al.16 used the MMSE excluding vision-related items and reported that there was no significant correlation between visual and cognitive impairments. Therefore, we performed additional analyses for the MMSE excluding the five vision-related items. We found that the significant association between visual impairment and cognitive impairment still remained. The exact mechanisms for the coexistence of vision and cognitive impairments have not been fully determined. A long-term loss of sensory stimulation caused by visual impairment could contribute to an atrophy of the central nervous system leading to cognitive hypofunction.9 Thus, it was hypothesized that the cognitive hypofunction would improve with an improvement of visual function. This hypothesis was tested by evaluating the cognitive function before and after cataract surgery.27–29 Prospective studies have showed that there were improvements in the MMSE scores after cataract surgery.27–29 We recently found in another cross-sectional study that the prevalence of cognitive impairment was significantly lower in subjects with history of cataract surgery than in subjects without history of cataract surgery, independent of visual acuity.30 However, Jefferis et al. reported that cognitive improvement was not significant considering a learning effect and the degree of vision recovery.31 In addition, a randomized study reported that there was no significant improvement in the cognitive function after cataract surgery.32 Although the progression of dementia can be slowed down, no improvement was achieved.33,34 Previous studies have reported visual impairment-related cognitive hypofunction based on longitudinal studies.11,26 However, Hong et al. recently reported that there was no significant association between visual impairment and cognitive function in a 10-year follow-up study.35 In any case, it may be difficult to demonstrate whether there is a significant relationship between visual acuity and cognitive functions. The association between the visual acuity and cognitive function is still unclear. The Fujiwara-kyo Study will have follow-up surveys every 5 years until 2027. With longitudinal observations and longer follow-up times, we should be able to clarify the temporality of the relationship between visual impairment and cognitive impairment. Our study was a cross-sectional study and some limitations are inherent. The participants were volunteers, which could potentially cause selection bias. The underlying diseases of visual impairment were unidentified and general medical conditions were obtained only by self-reporting. In addition, we used a very minimal definition of visual impairment (Snellen = 20/32) without consideration of visual fields or contrast sensitivity. They may represent a healthier population with better visual acuity and cognitive function than the general population. Actually, only 5.7% of subjects in our study were classified as being cognitive impaired, while the prevalence of cognitively impaired in individuals aged ≥65 years has been estimated to be 15% in Japan.3 Similarly, participants of this study may have had better visual acuity (mean of −0.02 ± 0.13 logMAR) than that of the general population. External validity of this study needs to be considered with careful interpretation. Further studies are needed to determine the temporal associations and to determine if an intervention to maintain or improve the visual acuity can delay or prevent the development of cognitive impairment. In conclusion, we found a significant association between visual and cognitive impairment. We also found that even mild visual impairment was significantly associated with cognitive impairment. Thus, maintaining good visual acuity might decrease the risk of having cognitive impairment. Acknowledgments We thank the participants and supporting organizations in the Fujiwara-kyo Study. We also thank Dr. Duco Hamasaki for advice and suggestions. This study was supported by Japan Society for the Promotion of Science (JSPS) KAKENHI grant Number 24592637. JSPS had no role in the design or conduct of this research. Author Disclosure Statement No competing financial interests exist. Abbreviations Used ANCOVAanalysis of covariance BCVAbest corrected visual acuity CIconfidence intervals logMARlogarithm of the minimum angle of resolution MMSEMini-Mental State Examination ORodds ratios QOLquality of life ==== Refs References 1 United Nations, Department of Economic and Social Affairs, Population Division . World Population Prospects: The 2010 Revision . United Nations , 2011 Available at: www.un.org/en/development/desa/population/publications/trends/population-prospects_2010_revision.shtml (accessed 6 30 , 2016 ) 2 Cabinet Office, Government of Japan . Annual Report on the Aging Society 2012 . Cabinet Office, Government of Japan ; 2013 Available at: www8.cao.go.jp/kourei/english/annualreport/2012/2012pdf_e.html (accessed 6 30 , 2016 ) 3 Asada T Prevalence of dementia in Japan: past, present and future . Rinsho Shinkeigaku . 2012 ;52 :962 –964 23196483 4 Iwano M , Nomura H , Ando F , et al. Visual acuity in a community-dwelling Japanese population and factors associated with visual impairment . Jpn J Ophtalmol . 2004 ;48 :37 –43 5 Reischies FM , Geiselmann B Age related cognitive decline and vision impairment affecting the detection of dementia syndrome in old age . Br J Psychiatry . 1997 ;171 :449 –451 9463604 6 Anstey KJ , Lord SR , William P Strength in the lower limbs, visual contrast sensitivity, and simple reaction time predict cognition in older women . Psychol Aging . 1997 ;12 :137 –144 9100274 7 Clemons TE , Rankin MW , McBee WL ; Age-Related Eye Disease Study Reseach Group. Cognitive impairment in the age-related eye disiease study: AREDS report no.16 . Arch Opthalmol . 2006 ;124 :537 –543 8 Busse A , Sonntag A , Bischkopf J , et al. Adaptation of dementia screening for vision-impaired older persons: administration of the Mini-Mental State Examination(MMSE) . J Clin Epidemiol . 2002 ;55 :909 –915 12393079 9 Lindenberger U , Baltes PB Sensory functioning and intelligence in old age: a strong connection . Psychol Aging . 1994 ;9 :339 –355 7999320 10 Anstey KJ , Luszcz MA , Sanchez L A reevaluation of the common factor theory of shared variance among age, sensory function, and cognitive function in older adults . J Gerontol B Psychol Sci Soc Sci . 2001 ;56 :3 –11 11 Anstey KJ , Luszcz MA , Sanchez L Two-year decline in vision but not hearing is associated with memory decline in very old adults in a population-based sample . Gerontology . 2001 ;47 :289 –293 11490149 12 Valentijn SA , van Boxtel MP , van Hooren SA , et al. Change in sensory functioning predicts change in cognitive functioning: results from a 6-year follow-up in the Maastricht aging study . J Am Geriatr Soc . 2005 ;53 :374 –380 15743277 13 Gussekloo J , de Craen AJ , Oduber C , et al. Sensory impairment and cognitive functioning in oldest-old subjects: the Leiden 85+ Study . Am J Geriatr Psychiatry . 2005 ;13 :781 –786 16166407 14 Tay T , Wang JJ , Kifley A , et al. Sensory and cognitive association in older persons: findings from an older Australian population . Gerontology . 2006 ;52 :386 –394 16921251 15 Diaz M , Norell M , Belkin J , et al. Cognitive profile of elders in an ophthalmic ambulatory setting . Br J Opthalmol . 2011 ;95 :24 –27 16 Killen A , Firbank MJ , Collerton D , et al. The assessment of cognition in visually impaired older adults . Age Aging . 2013 ;42 :98 –102 17 Oshima Y , Ishibashi T , Murata T , et al. Prevalence of age related maculopathy in a representative Japanese population: the Hisayama study . Br J Ophthalmol . 2001 ;85 :1153 –1157 11567955 18 Kawasaki R , Wang JJ , Rochtchina E , et al. Cardiovascular risk factors and retinal microvascular signs in an adult Japanese population: the Funagata Study . Ophthalmology . 2006 ;113 :1378 –1384 16877076 19 Nakata I , Yamashiro K , Nakanishi H ; Nagahama Cohort Research Group. Prevalence and characteristics of age-related macular degeneration in the Japanese population: the Nagahama study . Am J Ophthalmol . 2013 ;156 :1002 –1009 23938127 20 World Health Organization . World Health Statistics 2014 . World Health Organization Document Production Services , 2014 Available at: www.who.int/mediacentre/news/releases/2014/world-health-statistics-2014/en (accessed 6 30 , 2016 ) 21 Okamoto N , Morikawa M , Tomioka K , et al. Association between tooth loss and the development of mild memory impairment in the elderly: the Fujiwara-kyo study . J Alzheimers Dis . 2015 ;44 :777 –786 25362033 22 Morikawa M , Okamoto N , Kiuchi K , et al. Association between depressive symptoms and metabolic syndrome in Japanese community-dwelling older people: a cross-sectional analysis from the baseline results of the Fujiwara-kyo prospective cohort study . Int J Geriatr Psychiatry . 2013 ;28 :1251 –1259 23526542 23 International Organization for Standardization . Ophthalmic Optics—Visual Acuity Testing—Standard Optotype and Its Presentation . ISO 8596 :2009 , pp. 1 –5 24 International Council of Ophthalmology . Visual standards. Aspects and ranges of vision loss with emphasis on population surveys . In: The 29th International Congress of Ophthalmology : Sydney, Australia , 2002 Available at: www.icoph.org/downloads/visualstandardsreport.pdf (accessed 6 30 , 2016 ) 25 Folstein MF , Folstein SE , McHugh PR ‘Mini-Mental State’ a practical method for grading the cognitive state of patients for the clinician . J Psychiatr Res . 1975 ;12 :189 –198 1202204 26 Rogers MA , Langa KM Untreated poor vision: a contributing factor to late-life dementia . Am J Epidemiol . 2010 ;171 :728 –735 20150357 27 Gray CS , Karimova G , Hildreth AJ , et al. Recovery of visual and functional disability following cataract surgery in older people: Sunderland Cataract Study . J Cataract Refract Surg . 2006 ;32 :60 –66 16516780 28 Ishii K , Kabata T , Oshika T The impact of cataract surgery on cognitive impairment and depressive mental status in elderly patients . Am J Ophthalmol . 2008 ;146 :404 –409 18602079 29 Hall TA , McGwin G Jr , Owsley C Effect of cataract surgery on cognitive function in older adults . J Am Geriatr Soc . 2005 ;53 :2140 –2144 16398899 30 Miyata K , Obayashi K , Saeki K , et al. Higher cognitive function in elderly individuals with previous cataract surgery: cross-sectional association independent of visual acuity in the HEIJO-KYO cohort . Rejuvenation Res . 2016 ;19 :239 –243 26414122 31 Jefferis JM , Mosimann UP , Clarke MP Cataract and cognitive impairment: a review of the literature . Br J Ophthalmol . 2011 ;95 :17 –23 20807709 32 Anstey KJ , Lord SR , Hennessy M , et al. The effect of cataract surgery on neuropsychological test performance: a randomized controlled trial . J Int Neuropsychol Soc . 2006 ;12 :632 –639 16961944 33 World Health Organization . The ICD-10 Classification of Mental and Behavioural Disorders . World Health Organization ; 1992 34 Japanese Society of Neurology . Treatment Guidelines of Dementing Disorder 2010 . Japanese Society of Neurology , 2010 Available at: www.neurology-jp.org/guidelinem/nintisyo.html (accessed 6 30 , 2016 ) 35 Hong T , Mitchell P , Burlutsky G , et al. Visual impairment, hearing loss and cognitive function in an older population: longitudinal findings from the Blue Mountains Eye Study . PLoS One . 2016 ;11 :e0147646 26808979 Cite this article as: Mine M, Miyata K, Morikawa M, Nishi T, Okamoto N, Kawasaki R, Yamashita H, Kurumatani N, Ogata N (2016) Association of visual acuity and cognitive impairment in older individuals: Fujiwara-kyo eye study, BioResearch Open Access 5:1, 228–234, DOI: 10.1089/biores.2016.0023.

PMC005xxxxxx/PMC5003004.txt

==== Front Assay Drug Dev TechnolAssay Drug Dev TechnoladtAssay and Drug Development Technologies1540-658X1557-8127Mary Ann Liebert, Inc. 140 Huguenot Street, 3rd FloorNew Rochelle, NY 10801USA 2749473610.1089/adt.2016.72910.1089/adt.2016.729Original ArticlesPhenotypic Characterization of Toxic Compound Effects on Liver Spheroids Derived from iPSC Using Confocal Imaging and Three-Dimensional Image Analysis Sirenko Oksana 1Hancock Michael K. 2Hesley Jayne 1Hong Dihui 1Cohen Avrum 1Gentry Jason 1Carlson Coby B. 2Mann David A. 21 Molecular Devices, LLC, Sunnyvale, California.2 Cellular Dynamics International—A Fujifilm Company, Madison, Wisconsin.Address correspondence to:Oksana Sirenko, PhDMolecular Devices, LLC1311 Orleans DriveSunnyvale, CA 94089E-mail:oksana.sirenko@moldev.comDavid A. Mann, PhDCellular Dynamics International—A Fujifilm Company525 Science DriveMadison, WI 53711E-mail:dmann@cellulardynamics.com01 9 2016 01 9 2016 01 9 2016 14 7 381 394 © Oksana Sirenko et al., 2016; Published by Mary Ann Liebert, Inc.2016This Open Access article is distributed under the terms of the Creative Commons License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.Abstract Cell models are becoming more complex to better mimic the in vivo environment and provide greater predictivity for compound efficacy and toxicity. There is an increasing interest in exploring the use of three-dimensional (3D) spheroids for modeling developmental and tissue biology with the goal of accelerating translational research in these areas. Accordingly, the development of high-throughput quantitative assays using 3D cultures is an active area of investigation. In this study, we have developed and optimized methods for the formation of 3D liver spheroids derived from human iPS cells and used those for toxicity assessment. We used confocal imaging and 3D image analysis to characterize cellular information from a 3D matrix to enable a multi-parametric comparison of different spheroid phenotypes. The assay enables characterization of compound toxicities by spheroid size (volume) and shape, cell number and spatial distribution, nuclear characterization, number and distribution of cells expressing viability, apoptosis, mitochondrial potential, and viability marker intensities. In addition, changes in the content of live, dead, and apoptotic cells as a consequence of compound exposure were characterized. We tested 48 compounds and compared induced pluripotent stem cell (iPSC)-derived hepatocytes and HepG2 cells in both two-dimensional (2D) and 3D cultures. We observed significant differences in the pharmacological effects of compounds across the two cell types and between the different culture conditions. Our results indicate that a phenotypic assay using 3D model systems formed with human iPSC-derived hepatocytes is suitable for high-throughput screening and can be used for hepatotoxicity assessment in vitro. ==== Body Introduction Liver injury caused by toxicity of pharmaceutical drugs and environmental chemicals is a subject of great concern. Therefore, development of new complex systems that would allow effective testing for potential liver toxicity is an area of active investigation.1–3 There is a growing interest in using three-dimensional (3D) models for studying complex biology and tissue engineering. Recently, 3D models have been employed for developmental biology research and toxicity screening.4–9 Numerous studies have reported more liver-specific functional activity in hepatocyte 3D cultures than in conventional two-dimensional (2D) monolayers,5,10–12 since mechanical stress mediated by adhesion to extracellular matrix or other cells alters signal transduction and gene transcription in a variety of cell types. In the human liver hepatocellular carcinoma cell line (HepG2), cells respond to differing physical environments of 2D and 3D culture with altered actin cytoskeleton structure and cell shape. Further, global gene expression analysis has shown that distinct genetic programs are initiated depending on the physical structure of the cells: metabolic and synthetic functional genes, including cytochrome P450 and albumin, are upregulated in 3D spheroid structures.1,5,7,10,13–22 The majority of published liver toxicity studies have used transformed hepatocyte cell lines (HepG2, HepaRG) or primary hepatocytes for toxicity screening,10,12,16,18,19,23 whereas induced pluripotent stem cell (iPSC)-derived hepatocytes present a valuable model that can closely resemble the phenotypes and functionality of primary hepatocytes11,13,22,24–29 while minimizing variability and other limitations of primary cells. Human iPSC-derived hepatocytes show great promise with respect to having a primary tissue-like phenotype, consistent and unlimited availability, and the potential to establish genotype-specific cells from different individuals. There has been significant progress in the development of 3D cell models and techniques during the past several years. Development strategies have included biodegradable scaffolds, organ-on-a chip structures, and self-assembled organoids.24,29–31 Recently, spheroid formation in low-attachment round-bottom plates has become popular, as the method offers a simple workflow and is compatible with high-content imaging.7,17,23,32–35 Common methods of analysis include disruption of spheroids and analysis of cell lysates or suspensions for ATP or other metabolites with microplate readers,11 whereas high-content imaging methods have been proven productive for the characterization of phenotypic effects of chemical compounds on morphology and viability.33,36,37 High-content imaging can be used with numerous fluorophores in combination, including stains for viability, DNA binding, apoptosis markers, or mitochondria markers.19,38 This method can be extended to more complex multicellular models that express a plurality of fluorescent markers. The use of higher magnification, as well as confocal imaging and 3D analysis provides single-cell resolution and characterization of cell content and morphology in 3D volume. Higher magnification confocal imaging and 3D analysis also allow derivation of multi-parametric outputs for characterizing complex phenotypes of spheroids treated with compounds.33 The goal of this study was to develop and characterize confocal high-content imaging in combination with 3D image analysis methods that would be suitable for the high-throughput compound screening using liver spheroids made from iPSC-derived hepatocytes. Sample handing steps for cell culture, treatment, and staining have been reduced to minimize spheroid disturbances and increase assay reproducibility. We optimized and compared imaging and image analysis methods and described measurements for multi-parametric characterization of different spheroid phenotypes and determination of IC50 values. Furthermore, we characterized the assay using 50 benchmark cytotoxic and known hepatotoxic compounds and compared IC50 values for 3D and 2D cultures. Finally, we compared this model with spheroids formed with HepG2, and we discovered significant differences in toxicity assessment between those systems. The method described here can enhance development of relevant cell-based assays for efficient assessment of the hepatotoxicity of chemicals and drug candidates in high-throughput quantitative screening. Materials and Methods Cell Models Human iPSC-derived hepatocytes, iCell Hepatocytes 2.0 (Cellular Dynamics International), and HepG2 (ATCC) were used in the study. Cryopreserved cells were thawed and maintained according to provided protocols. To prepare spheroid cultures of human iPSC-derived hepatocytes, iCell Hepatocytes 2.0 were first thawed and plated at a high density (∼300,000 cells/cm2) onto collagen I-coated 24-well plates to allow the cells to recover from cryopreservation and to establish a confluent 2D culture. After 7 days in culture, cells were gently detached using StemPro Accutase (ThermoFisher Scientific), pelleted by centrifugation, and resuspended in William's E Medium containing Hepatocyte Maintenance Supplement (ThermoFisher Scientific). Immediately before plating the cells, the cell suspension was further supplemented with a 10% final volume/volume of ready-to-use hESC-qualified Geltrex reduced growth factor basement membrane matrix (ThermoFisher Scientific). The combined cells plus matrix suspension was plated directly into 96-well GravityTRAP ultra-low attachment spheroid plates (InSphero) at ∼1,000 cells/well, which were immediately centrifuged (300 g, 2 min) to settle the cells and to remove any bubbles before placing the plate in a humidified incubator (37°C, 5% CO2). Spheroid formation was observed within 24–48 h. HepG2 cells were cultured in minimum essential medium (Corning) supplemented with 10% fetal bovine serum and plated into 384-well ultra-low attachment (ULA) Corning U-shaped black clear-bottom plates (Corning 3830), at densities of 1,000 cells/well in the appropriate media. Chemicals and Treatments For compound screening, test agents were prepared as 10–100 mM stock solutions in tissue culture-grade dimethyl sulfoxide (DMSO; Sigma-Aldrich); the final concentration of DMSO in media was 0.1%–0.3%. Compounds (Sigma-Aldrich) were tested in triplicate or duplicate, in a six-point dilution series. For spheroid assays, iCell Hepatocyte and HepG2 spheroids were plated 48 h before initiation of the experiment. Cells were then exposed to the indicated concentrations of compounds for 72 h. For mitochondria toxicity or caspase activation assay, spheroids were cultured for 48 h and then treated with compounds for 24 h. For conventional 2D assays, cells were plated on collagen-coated 384-well plates (BD Biocoat) for 48 h and then treated with compounds for 72 h (Supplementary Fig. S1). Multiparametric Live Cell Toxicity Assay After incubation with test compounds, spheroids were stained with a mixture of three dyes: 2 μM calcein AM, 3 μM of EthD-1, and 10 μM Hoechst 33342 (Life Technologies). Dyes were prepared in sterile phosphate-buffered saline (Corning). In separate experiments, CellEvent Caspase 3/7 reagent (Life Technologies) was used to evaluate compound ability to trigger apoptosis signaling. Then, 7.5 μM CellEvent was added with 10 μM Hoechst nuclear dye diluted in Hank's Balanced Salt Solution (Corning). For mitochondria toxicity assay, MitoTracker Orange from Life Technologies (200 nM) was combined with Hoechst (10 μM). Dye solutions were added directly to the media without aspiration. Spheroids were incubated with dye for 2 h before imaging. Dye solution was not washed out, and care was taken during pipetting to avoid spheroid loss, disintegration, or displacement. Image Acquisition Images were acquired using an ImageXpress Micro® Confocal High-Content Imaging System (Molecular Devices), with a 20× Plan Fluor or 10× Plan Fluor objective. Typical image acquisition settings are given in Supplementary Table S1 (Supplementary Data are available online at www.liebertpub.com/adt). A stack of 11–17 planes separated by 5–10 μm was acquired, starting at the well bottom and covering approximately the lower half of each spheroid. Typically, a Z-stack of images covered 100–120 μm for iCell Hepatocyte or HepG2 spheroids. All individual images were saved and used for 3D analysis, as well as for 2D projection (Maximum Projection [MaxPro]) images. 3D Image Analysis Images were analyzed using the 3D analysis module of MetaXpress® High-Content Image and Analysis Software (Molecular Devices). The analysis method is described in the Supplementary Data. Find Spherical Objects function was used to define spheroids. Then, Count Nuclei, Cell Scoring, and Live-Dead application modules were used for cell count and viability assessment. A customized analysis for additional multi-parametric outputs was done using a protocol created in the MetaXpress custom module editor (CME). The custom module analysis first identified the spheroid using Hoechst staining. Spheroids were identified using the Find Spherical Objects function. Spheroid object sizes were characterized and compared by volume, diameter, and fluorescence intensities. Live cells were then identified by the presence of calcein AM signal or EthD-1 exclusion, and dead cells were identified by the presence of EthD-1 signal. Then, objects were connected in 3D volume by using Connect by Best Match function. Measurements from individual cells included nuclear or cellular number, nuclear or cell volume, diameter, average intensities for calcein AM, EthD-1, or Hoechst; counts of all nuclei; and evaluation of average nuclear size, intensity, and distances between nuclei. Calcein AM- or EthD-1-positive or -negative cells were counted, and their volumes and intensities values were measured. In some cases, images were smoothed with a Gaussian filter or sharpened with Top Hat function. EC50 values were determined using a four-parameter curve fit from SoftMax® Pro 6 software (Molecular Devices). Results Development and Optimization of the Live Cell High-Content Assay with 3D Spheroid Cultures The goal of this study was to develop and evaluate fast, accurate, and reproducible high-content imaging methods to investigate the effects of toxic compounds on the morphology and viability of 3D liver spheroids using techniques for confocal imaging and 3D image analysis. The workflow of the assay is presented in Figure 1A. Human iPSC-derived hepatocytes were cultured for 7 days in a 2D format before detaching the cells, supplementing the cell suspension with extracellular matrix, and plating the mixture into ULA plates to initiate 3D spheroid formation. Fig. 1. (A) Schematic diagram of 3D liver spheroid formation and high-content imaging assay workflow. The spheroid images shown here were from seeding with ∼1,000 cells/well. (B) Images of spheroids (calcein AM staining) shown for different numbers of cells plated per well (125–8,000 cells). (C, E) Dependence of spheroid diameter and volume on the number of cells plated (n = 6). (D) Nuclei counts of stained spheroids relative to cell plating numbers. (F) Average fluorescence intensities from spheroids stained with Hoechst, calcein AM, and EthD-1 for indicated times (n = 3). Optical filters used were DAPI (Hoechst, solid gray), FITC (calcein AM, solid black), and TexasRed (EthD-1, dashed line). 3D, three-dimensional. Both conical-shaped 96-well GravityTRAP ULA plates (InSphero) and U-shaped ULA plates (Corning) were tested for spheroid formation, compound addition, fluorescent staining, and confocal imaging. Both plates eliminate spheroid transfer steps and center the spheroids in a small focus area, facilitating capture of an entire spheroid in one 10× image. Cells aggregated at the bottom of the wells, forming spheroids within 24–48 h. The results reported in this study for iPSC-derived hepatocyte spheroids were all obtained using the GravityTRAP plates. We studied the reproducibility of spheroid formation from iPSC-derived hepatocytes and the dependence of the spheroid size on the number of plated cells. Cells were plated at different densities (60–8,000 cells/well), incubated for 48 h, and imaged. A single spheroid per well was typically formed at each cell plating density from ∼125–8,000 cells/well, with diameter and volume sizes increasing proportionally to the number of cells plated (Fig. 1B, C, E). We found that a plating density of 1,000 cells/well resulted in a consistent spheroid size and shape, with object sizes suitable for image acquisition and analysis. At this density, the average spheroid diameter after 2 days was consistent as measured by transmitted light or fluorescent imaging with a value of 220 ± 24 μm (n = 6), yielding a coefficient of variation (CV) of 10.9%. Smaller spheroids exhibited greater size variability (CV = 20% for 500 cells plated, n = 6), whereas variability in size decreased for larger spheroids (CV = 4% for 2,000 cells plated, n = 6). Since iPSC-derived hepatocyte spheroids do not proliferate, the diameter of untreated spheroids remained consistent in culture through the experiment. We had previously described staining protocols for spheroids made from transformed cell lines.39 Here, we used a previously developed staining protocol and concentrations with a one-step dye mixture addition that eliminates the need for fixing cells or repeated wash steps.22,39 Images of representative spheroids stained with a mixture of three dyes are shown in Figure 2A, B. Calcein AM was used to measure cell metabolic activity, viability, and a variety of morphological parameters. Hoechst was utilized to measure total cell count and nuclear shape. EthD-1 selectively penetrates cells with damaged outer membranes and was used to mark dead or necrotic cells. Fig. 2. (A) Untreated and staurosporine-treated spheroids (1 μM) were stained with a combination of three dyes: Hoechst (10 μM), EthD-1 (3 μM), and Calcein AM (1 μM). Numbers represent the appropriate objects positive for each marker counted in 3D space. (B) Composite images of all three channels shown for the control and select compound-treated spheroids. The spheroids were treated with staurosporine (1 μM), ketoconazole (30 μM), or haloperidol (30 μM) for 72 h and images were generated using maximum projection from a Z-stack of 11 images taken at 10 μm intervals. Counted numbers of calcein AM-positive (live) cells are shown for the different treatments. (C) Apoptosis is apparent in the composite images of spheroids stained with CellEvent Caspase 3/7 reagent to detect caspase 3/7 (green), EthD-1 (red) for dead cells, and Hoechst (blue) for all nuclei. Spheroids treated for 24 h with methyl mercury (1 μM) exhibited a higher number of apoptotic cells (green) than the control (number of apoptotic cells is indicated). (D) Composite images for the control and Antimycin A (3 μM) treated spheroids stained with MitoTracker Orange (red) and Hoechst (blue). Antimycin A disrupts mitochondrial membrane integrity. Number of cells with intact mitochondria is indicated. Spheroids of varying sizes were tested using this protocol to confirm its suitability for 3D analysis (Fig. 1B–E). The raw nuclei counts relative to the cell plating densities were less than expected due to imaging limitations described next. For spheroids containing ∼1,000 cells/well (∼220 μm in diameter), the counted number of nuclei was ∼500 (Fig. 1D). An additional experiment (Fig. 1F) demonstrated the time dependency of dye penetration into spheroids, with staining times of 90–120 min required to achieve maximal signal intensity. Spheroid imaging could be accomplished for up to 6 h post-staining if plates were kept in the incubator (data not shown). Additional staining protocols were evaluated for apoptosis (Caspase 3/7, Fig. 2C) and mitochondria integrity (MitoTracker Orange, Fig. 2D) dyes. Imaging 3D Spheroids A challenge in performing spheroid assays is the ability to image and analyze cells in 3D. Common issues previously described include poor light penetration into or out of the spheroid center, light scattering by cells, and high background from out-of-plane fluorescence.40 These issues tend to be less important for smaller spheroids (∼220 μm) used in the present study. Confocal microscopy provides efficient background rejection and sharper images. In this study, images were taken using an automated confocal microscope with a large field-of-view camera. Imaging protocols were additionally optimized for 3D image analysis. The microscopy system effectively captures spheroids in a single image using a 10× objective. At 10× magnification, we captured either 11 images with 10 μm distance or 17 images with 5 μm distance, which resulted in comparable cell counts. Images acquired with a 20× objective allowed better visualization of details, but they did not provide a significant difference in cell count. Because of the limitation of sufficient light penetration into dense objects, the approximate depth of imaging into the hepatocyte spheroids prepared and stained as described earlier was ∼110 microns. Although longer staining times up to about 2 h improved dye penetration and average signal intensities (Fig. 1F), no further increase in light penetration or depth of view was obtained. Therefore, for spheroids containing ∼1,000 cells/well (∼220 μm diameter), visualization and enumeration of objects (cells, as defined) within the object yields approximately half of the actual number of cells present in the intact spheroid (∼500 counted). This can be illustrated by Figure 1D, showing dependence of the number of cells counted in spheroids from the number of cells plated initially. In contrast, for bigger spheroids, total cell number is undercounted due to the limitations of the light penetration into samples. In the case of disintegrated, deformed, or “loosened up” spheroids, the same depth of penetration might produce an artificially higher cell count based on analysis of more than a half of the compromised spheroid. The MaxPro algorithm, which combines the highest intensity pixels along lines orthogonal to the image planes into a single image, was used for cell count comparisons. As expected, nuclei counted in 2D MaxPro represented only a fraction of nuclei counted by 3D analysis (data not shown). 3D Analysis of Spheroid Images In the present study, 3D analysis of images was accomplished using recently added features of MetaXpress software. The software converts a stack of 2D images obtained by confocal acquisition into 3D space with appropriate detection and segmentation of objects. Spherical object analysis allowed for measuring several parameters that are important for phenotypic characterization of drug-induced hepatotoxicity, including spheroid diameter size, volume, and live/dead cell fluorescence marker intensities (Fig. 3). For example, spheroids treated with methyl mercury were readily distinguished from the vehicle control by significant morphological changes in spheroid size and volume (i.e., toxicity-induced loss of spheroid integrity leading to gross enlargement of the 3D object) as well as by decreased viability staining (FITC channel intensity). Fig. 3. (A) Analysis functions segment objects in 3D volume using predetermined size and intensity thresholds. (B) Analysis of the spheroid as a single object. Phenotypic measurements can then be calculated, including: spheroid diameter, spheroid volume, and average intensities for nuclear (DAPI channel) and viability staining (FITC channel). Here, values were determined for the control (light gray) and methyl mercury treated (1 μM, dark gray) spheroid wells. The values were normalized to the corresponding control values that were set at 100%. Error bars represent standard deviations (n = 3). Values significantly different from controls (*P < 0.05, T-test) are marked with asterisks. The 3D analysis allows for defining individual nuclei and cells as well as for appropriate characterization of the individual cells and subcellular objects. Figure 4 illustrates how more detailed segmentation was performed, for example, to identify and count nuclei within a spheroid and score the results for live/dead populations based on the staining patterns obtained for each marker dye. Briefly, individual Z-planes were first segmented and analyzed as 2D images, for example, for nuclei count, live/dead, or cell scoring features; then, the objects were “Connected by Best Match” function with user-defined maximum displacement of each object relative to another one (e.g., 5–10 microns for the maximum displacement of the nuclei and 20–30 microns for cytoplasm). Nuclei or individual cells were, thus, segmented and scored in the context of a layered 3D volume, ideally without missing objects or counting the same objects twice. Fig. 4. (A) Confocal images of a spheroid stained with Hoechst acquired using a 10 × Plan Fluor objective as a Z-stack of 11 images 10 μm apart, with three selected planes shown. Results from image analysis object masks are shown to the right of each image, with individual nuclei segmentation given a random pseudo-color. (B) Confocal images of the same spheroids showing the overlays of Hoechst (blue), calcein AM (green), and EthD-1 (red) staining. The resulting object masks are shown to the right of each image, with masks for the entire spheroid light blue, viable cell nuclei indicated in yellow, dead cell nuclei in dark blue, and cytoplasm of viable cells shown in pink. (C) Examples of 3D masks for nuclei and (D) calcein AM-positive cytoplasm. These steps allow defining and counting the number of total cells, number of calcein AM-positive or -negative cells, and number of ethidium homodimer-positive and -negative cells. In addition, the intensities of individual cells can be defined in different colors, as well as cell volumes, diameters, distances between objects, or spatial locations of objects in 3D matrix. Significantly, the larger spheroid mask can be used to define the numbers of smaller objects within the spheroid, or externally. Phenotypic Analysis of Compound Treatment Effects Significant changes in spheroid phenotypes and cell content were observed after compound treatment. Many spheroids lost their spherical shape, appeared disintegrated, “loose,” “flattened,” or “irregular,” had cells detached from the main body, or exhibited condensed nuclei due to cell death. These phenotypic changes in the spheroids after compound treatment occurred over a range of concentrations (Fig. 5). Fig. 5. Representative images of spheroids seeded at 1,000 cells/well and treated with different compounds for 72 h (concentrations indicated). Composite images of Hoechst (blue), calcein AM (green), and EthD-1 (red) are presented. Quantitation is presented in Figure 6. For example, for amiodarone-treated spheroids, there was a notable increase in spheroid volume and diameter, as well as a decrease in the intensity of the live-cell calcein AM signal. Quantitative analysis of the images included derivation of parameters to assess morphological features of spheroids, cell content, and complexity (Fig. 6A). The spheroid volume, diameter, and fluorescence intensities were measured, along with the number of calcein AM-positive “live” cells, the number of EthD-1-positive “dead” cells, as well as the average volumes and intensities of cells exhibiting these markers (Fig. 6B). Fig. 6. (A) Image analysis masks for a single plane shown for amiodarone-treated (5 and 100 μM) and control spheroids. Spheroid masks are indicated in blue, live cell nuclei are marked in yellow, live cell cytoplasm is shown in pink, and dead cell nuclei are marked dark blue. (B) Image analysis readouts were derived as a result of a cell scoring analysis performed in 3D space. Measurements: number of live cells and dead cells per spheroid, sum of the volumes of live cells (determined by calcein AM-positive staining of cytoplasm), integrated intensity for calcein AM signal, volume of spheroid, diameter of spheroid, and average intensity of the Hoechst nuclear stain. Data shown correspond to the following compounds (72 h exposure): control (0.1% dimethyl sulfoxide), mitomycin C (5 μM), rotenone (60 μM), haloperidol (45 μM), pimozide (60 μM), amiodarone (45 μM), staurosporine (1 μM), and methyl mercury (20 μM). All values are presented as normalized relative to the control, with the exception of the number of dead cells that was normalized to the number of total counted cells in the control. Error bars represent the standard deviations of the replicates (n = 3). Significant differences from control (P < 0.05) are indicated on the graph with asterisks. Representative images for the corresponding treatments can be found in Figure 5. (C) The dose–dependent effects and the four-parameter curve fits for the selected compounds for the 3D spheroid toxicity assay using the number of live cells (calcein AM-positive cells) per spheroid. The number of live cells was determined by the Cell Scoring analysis of images using the 3D analysis features (n = 3). IC50 values are included in Table 1. Total cell number did not tend to decrease in response to compound effects. Instead, a concentration-dependent decrease in the number of viable cells was observed concomitant with an increase in the number of dead cells. Average fluorescent intensity for calcein AM was dramatically reduced for the entire spheroid as well as for individual segmented cells (cytoplasm). In addition, the average volume of cytoplasm stained with live cell dye calcein AM was decreased (live cell volume), as well as the sum of the volumes of calcein AM-positive cells. For certain compound treatments such as methyl mercury and staurosporine, there was an increase in Hoechst fluorescence intensity, likely indicating nuclear condensation and apoptosis. An increase in the distance between the nuclei was also observed, and this could be attributed to a loss of cell-to-cell junctions. When using calcein AM, the number of metabolically active live cells, the sum of the volumes of the live cells, or their integrated fluorescent intensities provided the greatest sensitivity for assessment of cytotoxicity, as indicated by the largest fold changes between the control and treated samples (Fig. 6B). MetaXpress CME was employed for multi-parametric image analysis. Spheroid diameter, spheroid volume, or count of dead cells changed significantly for some compounds, but typically did not follow four-parametric concentration-dependent responses. In contrast, the live cell number, the sum of their volumes, and the calcein AM integrated fluorescence intensities resulted in the largest assay windows, statistically significant differences between treated and control samples, and successfully fit into four-parameter dose–response curve models. Z′-factors, indicators of data quality, for the different readouts are presented in Supplementary Table S2. For an investigation into the cytotoxicity mechanisms, we also evaluated the apoptotic phenotype and mitochondria integrity. Activation of apoptosis using Caspase 3/7 dye was measured 24 h after compound treatment. Typical activated caspase 3/7 staining patterns for control cells, and those treated with methyl mercury, are shown in Figure 2C. Treatment with compounds causing apoptosis resulted in an increase of caspase 3/7 staining intensity and the number of caspase 3/7 positive (apoptotic) cells. Mitochondrial potential was evaluated with MitoTracker Orange dye. Treatment of spheroids with compounds affecting mitochondria integrity resulted in a decrease of the intensity of MitoTracker Orange staining and a decrease of cells with intact mitochondria (Fig. 2D). Effects of Known Hepatotoxic Compounds Spheroid assay performance was characterized using a representative set of 48 compounds that included known cytotoxic and hepatotoxic agents, plus six chemicals considered “safe” or nontoxic. Table 1 presents IC50 values for different classes of hepatotoxic compounds: known liver toxins (aflatoxin, rotenone, methyl mercury), anti-cancer drugs (e.g., doxorubicin, mitomycin), anti-depressants, anti-psychotics (e.g., paroxetine, fluoxetine, thioridazine, haloperidol), anti-malarial agents (quinacrine, chloroquine), and several other types of drugs. Of the 42 selected known hepatotoxins in the set, 36 elicited detectable toxicity effects in the spheroid hepatotoxicity assay (86% sensitivity), whereas all compounds considered “safe” had no effect on the number of live cells or spheroid phenotype (100% predictivity). Table 1. Comparison of Compound Responses Between 3D and 2D Cultures of iPSC-Derived Hepatocytes   iCell hepatocytes   Compounds IC50 (μM)a 3D assay IC50 (μM) 2D assay Compound description Toxic substances  TAB 1.69 ± 0.082 1.22 ± 1.02 Toxin  Methyl mercury 2.82 ± 0.49 5.05 ± 1.31 Toxin  Fuzariotoxin 0.481b 3.75 Toxin  Aflatoxin B1 0.595 ± 0.145 1.7 Toxin  Aflatoxin G1 7.78 ± 2.13 9.45 ± 2.01 Toxin  Rotenone 21.4 ± 6.9 n.d.c Pesticide Anti-proliferative agents  Mitoxantrone 0.533 ± 0.011d 3.7 ± 0.32 Topoisomerase II inhibitor  Daunorubicin 0.671 ± 0.046 0.5 DNA-intercalator, anti-cancer  Idarubicin 1.11 0.94 DNA-intercalator, anti-cancer  Doxorubicin·HCl 5.69 ± 4.04 2.17 ± 0.08 DNA-intercalator, anti-cancer  Staurosporine 0.91 ± 0.12 1.71 ± 0.87 Kinase inhibitor, apoptosis inducer  Mitomycin C 2.28 ± 0.25 2.36 ± 0.65 DNA-intercalator, anti-cancer Other drugs  Quinacrine 1.029 ± 0.212 2.52 Anti-malaria  Coralgil 1.24 6.71 Inducer of steatosis  Puromycin 3.82 7.2 ± 11.2 Protein synthesis inhibitor  Amlodipine 4.12 30.2 Ca blocker  rac Perhexiline Maleate 6.36 ± 0.15 9.89 Mitochondria inhibitor  Tamoxifen 12.4 10.23 Estrogen receptor antagonist  Miconazole nitrate 13.71 ± 0.789 103.1 ± 132 Anti-fungal  Fluphenazine di·HCl 13.0 ± 0.824 29.5 Anti-psychotic  Paroxetine 14.73 ± 3.92 30.4 Antidepressant  Thioridazine 15.83 ± 4.47 31.4 Anti-psychotic  Chloroquine 18.60 ± 0.247 38.5 Anti-malaria  Ketoconazole 20.4 ± 10.7 61.0 ± 46.2 Anti-fungal  Amiodarone·HCl 22.6 ± 9.68 15.3 ± 18.3 Anti-arrhythmia, autophagy inducer  Paclitaxel 26.2 ± 6.39 No tox (100)e Microtubule inhibitor  Fluoxetine 28.9 10.17 ± 0.321 Antidepressant  Imatinib 36.8 n.d. Kinase inhibitor  Pimozide 35.1 ± 1.39 13.5 ± 15.9 Anti-psychotic  Apigenin 36.9 ± 10.2 n.d. Autophagy inducer  Haloperidol·HCl 52.2 ± 21.6 No tox (100) Anti-psychotic  Isradipine 168 ± 137 n.d. Ca blocker  Colchicine >100f >100 Microtubule inhibitor  Tolcapone >100 >100 Parkinson disease drug  Doxocycline >100 >100 Antibiotic  Retinoic acid ∼500 480 Retinoic acid receptor agonist  Imipramine No tox (100) >100 Antidepressant (TCA)  Carboplatin No tox (500) No tox (500) DNA repair, anti-cancer  Cytarabine No tox (1,000) No tox (1,000) DNA intercalator, anti-cancer  Etoposide No tox (1,000) No tox (1,000) Topoisomerase inhibitor, anti-cancer  Phenylbutazone No tox (100) No tox (100) Nonsteroidal anti-inflammatory  Acetaminophen No tox (1,000) No tox (1,000) Nonsteroidal anti-inflammatory Negative controls  Penicillin V potassium salt No tox (1,000) No tox (1,000) Antibiotic  Streptomycin sulfate No tox (1,000) No tox (1,000) Antibiotic  Aspirin No tox (1,000) No tox (1,000) Anti-inflammatory  Ampicillin No tox (1,000) No tox (1,000) Antibiotic  Sucrose No tox (1,000) No tox (1,000) Carbohydrate  Sorbitol No tox (1,000) No tox (1,000) Carbohydrate a IC50 values (in μM) measured for tested compounds using number of live cells per spheroid as the readout. Error limits are standard error of the parameter estimate defined from the curve fit. b The undefined standard errors for some parameters indicate that although the curve fits have converged, the uncertainty in the parameter estimates could not be determined. c n.d., not determined. d Difference between 3D and 2D is significant (P < 0.05). Indicated in bold. e “No tox” means that no apparent effects were observed at the highest concentration tested (value in μM in parentheses). f “>100” means that toxic effects (decrease of calcein AM-positive cells) were observed at the highest concentration tested (100 μM), but IC50 values were not determined. 2D, two-dimensional; 3D, three-dimensional; iPSC, induced pluripotent stem cell; TAB, tetraoctylammonium bromide. We also compared the IC50 values obtained from 3D cultures versus those obtained from 2D cultures of iPSC-derived hepatocytes (Table 1), which had also been treated with compounds for 72 h. With some notable exceptions (e.g., mitoxantrone, P < 0.05), the majority of the IC50 values were not statistically significantly different (P > 0.05, or not determined) between the 3D and 2D cultures. It is noteworthy that the standard errors were not calculated for many of the compound treatments, thus preventing a complete evaluation of differences in IC50 between the 3D and 2D cultures. However, 21 of the 33 compounds with detectable effects in both 3D and 2D formats yielded slightly lower IC50 values for the 3D cultures, suggesting a possible trend toward stronger toxicity effects in 3D culture. We also compared effects for 23 selected compounds on spheroids made from transformed liver cells. Differences observed were related to the continuous proliferation of HepG2 cells in 3D culture. HepG2 spheroids were larger in size (300–350 μm diameter per 1,000 plated cells) in comparison with spheroids composed of iPSC-derived hepatocytes, since HepG2 cells appear to continuously grow in the culture during compound incubation. Also, there was a marked decrease in the size of HepG2 spheroids as well as a decrease in total cell number, in contrast to the nondividing iCell hepatocyte spheroids that did not typically decrease in size or cell number in response to compound treatment. A comparison of IC50 values for 23 compounds between the two cell types is given in Table 2. There were significant differences (P < 0.05) between the data with respect to several anti-proliferative compounds (e.g., staurosporine, paclitaxel, and mitomycin). The IC50s determined for HepG2 spheroid cultures were in the nanomolar range, whereas the iPSC-derived hepatocyte spheroids either displayed markedly right-shifted IC50s or were unaffected by these types of agents. With occasional exceptions (e.g., methyl mercury, pimozide, tetraoctylammonium bromide), the majority of the hepatotoxic compounds acting via mechanisms that do not inhibit replication (e.g., haloperidol and ketoconazole) showed similar effects in both cell models. Table 2. Comparison of Compound Responses Between 3D Cultures of HepG2- and iPSC-Derived Hepatocytes   HepG2, IC50 (μM)a iPSC-derived hepatocytes, IC50 (μM) Anti-proliferation agents  Idarubicin 0.00513 ± 0.00523 1.11  Staurosporine 0.00582 ± 0.00098b 0.91 ± 0.12  Doxorubicin·HCl 0.008 ± 0.0005 5.69 ± 4.04  Mitomycin C 0.012 ± 0.0006 2.28 ± 0.25  Paclitaxel 0.01 ± 0.0012 26.2 ± 6.39  Cytarabine 0.307 ± 0.097 No tox (1,000)c  Etoposide 4.18 ± 2.99 No tox (1,000)  Carboplatin 15.82 ± 2.61 No tox (1,000) Toxic substances  TAB 0.005 ± 0.0047 1.69 ± 0.082  Methyl mercury 9.19 ± 2.98 2.82 ± 0.49  Rotenone 8.32 ± 4.37 21.4 ± 6.94 Other drugs  Pimozide 4.73 ± 3.18 35.1 ± 1.39  Tamoxifen 18.7 ± 29.4 12.4 ± 13.2  Ketoconazole 19.2 ± 12.5 20.4 ± 10.7  Apigenin 21.9 ± 13.4 36.9 ± 10.18  Isradipine 22.17 168 ± 137  Haloperidol·HCl 25.08 ± 5.50 52.2 ± 21.6  Chloroquine 33.28 18.60 ± 0.247  Amiodarone·HCl 42.4 ± 8.39 22.6 ± 9.68  Imatinib 178 ± 40.54 36.8 Negative controls  Aspirin No tox (1,000) No tox (1,000)  Ampicillin No tox (1,000) No tox (1,000)  Sucrose No tox (1,000) No tox (1,000) a IC50 values (in μM) were determined for tested compounds using number of live cells per spheroid as the readout. Error limits are standard error of the parameter estimate defined from the curve fit. b Differences between HepG2- and iPSC-derived hepatocytes is significant (P < 0.05) for compounds indicated in bold. c “No tox” means that no apparent effects were observed at the highest concentration tested (value in μM in parentheses). Discussion Human 3D tissue models derived from iPSC-differentiated cells can provide useful systems for analyzing the pathogenesis of diseases or drug-induced toxic effects.11,41,42 Although in vivo toxicity assessment remains the “gold standard,” 3D models can accelerate the process by enabling rapid compound prioritization for in vivo testing and establishing mechanisms of action. Although interest in 3D cell models is increasing rapidly, the currently limited characterization of 3D assays impedes widespread adoption. Relative to established 2D methods, 3D methods are less developed with respect to imaging and analysis, and compatibility with screening systems. There are still important limitations to 3D systems, including imaging and analysis, sensitivity, and compatibility with screening systems. The aim of the present study was to develop and characterize confocal imaging and 3D analysis methods that would be compatible with automated instruments and show feasibility of using 3D spheroid cultures for toxicity screening. In contrast to previous studies, we have focused here on optimizing higher resolution imaging and 3D analysis. We have detailed the analysis protocol and characterized the cell population analysis method that allows multi-parametric quantification of different phenotypes. We have shown how 3D image analysis can deliver a number of informative phenotypic readouts that enable screening for deleterious effects of test compounds on cell morphology and viability from a simple assay workflow. Using a set of representative hepatotoxic compounds, we demonstrated how different readouts can be used in combination to characterize complex drug-induced effects on aspects such as spheroid morphology (e.g., changes in size or volume), integrity (e.g., changes in the distance between cells), viability (e.g., live/dead cell counts), and other parameters (e.g., average fluorescence intensity) to assess toxicity effects (Figs. 2–4). IC50 values from different readouts showed that counting the number of calcein AM-positive (metabolically active) cells in spheroids provided the greatest sensitivity to compound effects and yielded the best curve fits for ranking compound toxicities. Other parameters measured would provide supporting information about specific drug-induced phenotypes, for example, spheroid size, integrity, or content of dead cells. It should also be noted that due to better light penetration into deformed spheroids, it is possible that they might yield higher live cell counts, resulting in an underestimation of certain IC50 values. However, deformed spheroids are typically expected to contain few live cells that would minimize the effect of overcounting live cells on IC50 values. A comparison of compound toxicity effects between iPSC-derived hepatocytes and HepG2 spheroid cultures demonstrated significant differences in assay response. Not surprisingly, the cancer-derived HepG2 cell line was >100× more sensitive to compounds with anti-proliferative effects. Thus, the rank order for toxicity was also very different, with HepG2 cells ranking the anti-proliferative compounds as more toxic. In addition, the observed phenotypic changes were also very different. A decrease in spheroid size and total cell number for HepG2 spheroids was not observed for iPSC-derived hepatocytes. These observations suggest that HepG2 spheroids represent a very different model that reflects special properties of cancer cells that are not present in primary cells. In addition, the iPSC-derived hepatocyte 3D spheroid assay results were compared with results obtained using a typical 2D culture format. It has been previously shown that 3D cultures can often be more or less sensitive than 2D cultures to toxicity effects, and that these relative sensitivities correlate with compound mechanisms of action.1,9,11,32 Nevertheless, the studies here found that the IC50 values (Table 1) obtained for many of the compounds in this small test set were comparable between the 3D spheroids and the 2D culture, and the rank order of compound potency was similar. However, since the assays here were conducted over relatively brief exposure times, it is possible that more functional differences between the 2D and 3D formats could be revealed if the treatments were extended over longer periods. It is also possible that testing a larger set of compounds, including compounds shown to elicit marginal toxicity effects in 2D, could identify additional differences between the two models. Similarly, drugs influenced by CYP450 activity or other metabolic processing events may respond differently in 3D systems owing to enhanced functions. Some comparisons of long-term drug-induced toxicities in 3D versus 2D cultures have recently been noted using primary cells (InSphero23), which demonstrated the importance of extended exposure times to better discriminate compound effects between 2D and 3D culture formats. Going forward, we anticipate successful applications of the method described here to high-content assays utilizing other cell types and additional markers for readouts such as hypoxia, cytoskeleton, and kinase activation. The approach may also be extensible to more complex 3D systems, such as cultures containing multiple cell types (e.g., Kuppfer cells, fibroblasts, endothelial cells) where 3D analysis would allow characterization of different cell populations and their roles in toxicity and liver injury. In summary, the 3D spheroid cell model combined with high-content 3D assays presented here shows promise as a sensitive and reproducible screening tool for assessing hepatotoxicity. Although the predictivity of the assay in comparison with animal and clinical data still needs to be established, further development of the methods and models will increase the utility of this in vitro tool for screening. Supplementary Material Supplemental data Acknowledgments The authors would like to thank Evan Cromwell for his helpful discussions, Trisha Mitlo for technical help, and Gaoder Austin for help with data presentation. Disclosure Statement O.S., J.H., D.H., A.C., and J.G. are employed by Molecular Devices LLC, which sells the ImageXpress Micro Confocal system and MetaXpress software; M.K.H., C.B.C., and D.A.M. are employed by Cellular Dynamics Int., which provides iPSC-derived cells. Abbreviations Used 3Dthree-dimensional 2Dtwo-dimensional ATPadenosine triphosphate CMEcustom module editor CVcoefficient of variation DAPI4′,6-diamidino-2-phenylindole DMSOdimethyl sulfoxide EthD-1ethidium homodimer-1 FITCfluorescein isothiocyanate iPSCinduced pluripotent stem cell MaxPromaximum projection SDstandard deviation TABtetraoctylammonium bromide TRITCtetramethyl rhodamine isothiocyanate ULAultra-low attachment ==== Refs References 1 Hartung T : Toxicology for the twenty-first century . Nature 2009 ;460 :208 –212 19587762 2 Atienzar FA , Novik EI , Gerets HH , et al. : Predictivity of dog co-culture model, primary human hepatocytes and HepG2 cells for the detection of hepatotoxic drugs in humans . Toxicol Appl Pharmacol 2014 ;275 :44 –61 24333257 3 Gerets HH , Tilmant K , Gerin B , et al. : Characterization of primary human hepatocytes, HepG2 cells, and HepaRG cells at the mRNA level and CYP activity in response to inducers and their predictivity for the detection of human hepatotoxins . Cell Biol Toxicol 2012 ;28 :69 –87 22258563 4 Ramsden D , Tweedie DJ , Chan TS , Taub ME , Li Y : Bridging in vitro and in vivo metabolism and transport of faldaprevir in human using a novel cocultured human hepatocyte system, HepatoPac . Drug Metab Dispos 2014 ;42 :394 –406 24366904 5 Chang TT , Hughes-Fulford M : Monolayer and spheroid culture of human liver hepatocellular carcinoma cell line cells demonstrate distinct global gene expression patterns and functional phenotypes . Tissue Eng Part A 2009 ;15 :559 –567 18724832 6 Wang W , Itaka K , Ohba S , et al. : 3D spheroid culture system on micropatterned substrates for improved differentiation efficiency of multipotent mesenchymal stem cells . Biomaterials 2009 ;30 :2705 –2715 19215979 7 Wang J , Chen F , Liu L , et al. : Engineering EMT using 3D micro-scaffold to promote hepatic functions for drug hepatotoxicity evaluation . Biomaterials 2016 ;91 :11 –22 26994875 8 Drewitz M , Helbling M , Fried N , et al. : Towards automated production and drug sensitivity testing using scaffold-free spherical tumor microtissues . Biotechnol J 2011 ;6 :1488 –1496 22102438 9 Tung YC , Hsiao AY , Allen SG , Torisawa YS , Ho M , Takayama S : High-throughput 3D spheroid culture and drug testing using a 384 hanging drop array . Analyst 2011 ;136 :473 –478 20967331 10 Al-Qubaisi M , Rozita R , Yeap SK , Omar AR , Ali AM , Alitheen NB : Selective cytotoxicity of goniothalamin against hepatoblastoma HepG2 cells . Molecules 2011 ;16 :2944 –2959 21471934 11 Kunz-Schughart LA , Freyer JP , Hofstaedter F , Ebner R : The use of 3-D cultures for high-throughput screening: the multicellular spheroid model . J Biomol Screen 2004 ;9 :273 –285 15191644 12 Dunn JC , Yarmush ML , Koebe HG , Tompkins RG : Hepatocyte function and extracellular matrix geometry: long-term culture in a sandwich configuration . FASEB J 1989 ;3 :174 –177 2914628 13 Anson BD , Kolaja KL , Kamp TJ : Opportunities for use of human iPS cells in predictive toxicology . Clin Pharmacol Ther 2011 ;89 :754 –758 21430658 14 Elkayam T , Amitay-Shaprut S , Dvir-Ginzberg M , Harel T , Cohen S : Enhancing the drug metabolism activities of C3A—a human hepatocyte cell line—by tissue engineering within alginate scaffolds . Tissue Eng 2006 ;12 :1357 –1368 16771648 15 Gomez-Lechon MJ , Castell JV , Donato MT : The use of hepatocytes to investigate drug toxicity . Methods Mol Biol 2010 ;640 :389 –415 20645064 16 Groneberg DA , Grosse-Siestrup C , Fischer A : In vitro models to study hepatotoxicity . Toxicol Pathol 2002 ;30 :394 –399 12051557 17 Li L , Zhou Q , Voss TC , Quick KL , LaBarbera DV : High-throughput imaging: focusing in on drug discovery in 3D . Methods 2016 ;96 :97 –102 26608110 18 Medine CN , Lucendo-Villarin B , Storck C , et al. : Developing high-fidelity hepatotoxicity models from pluripotent stem cells . Stem Cells Transl Med 2013 ;2 :505 –509 23757504 19 O'Brien PJ , Irwin W , Diaz D , et al. : High concordance of drug-induced human hepatotoxicity with in vitro cytotoxicity measured in a novel cell-based model using high content screening . Arch Toxicol 2006 ;80 :580 –604 16598496 20 Presley AD , Fuller KM , Arriaga EA : MitoTracker Green labeling of mitochondrial proteins and their subsequent analysis by capillary electrophoresis with laser-induced fluorescence detection . J Chromatogr B Analyt Technol Biomed Life Sci 2003 ;793 :141 –150 21 Scotter EL , Narayan P , Glass M , Dragunow M : High throughput quantification of mutant huntingtin aggregates . J Neurosci Methods 2008 ;171 :174 –179 18367250 22 Sirenko O , Hesley J , Rusyn I , Cromwell EF : High-content assays for hepatotoxicity using induced pluripotent stem cell-derived cells . Assay Drug Dev Technol 2014 ;12 :43 –54 24229356 23 Messner S , Agarkova I , Moritz W , Kelm JM : Multi-cell type human liver microtissues for hepatotoxicity testing . Arch Toxicol 2013 ;87 :209 –213 23143619 24 Berger DR , Ware BR , Davidson MD , Allsup SR , Khetani SR : Enhancing the functional maturity of induced pluripotent stem cell-derived human hepatocytes by controlled presentation of cell-cell interactions in vitro . Hepatology 2015 ;61 :1370 –1381 25421237 25 Mann DA , Einhorn S. , Block K : Human iPSC-derived hepatocytes: functional model tissue for in vitro predictive metabolism, toxicity, and disease modeling . Gen Eng Biotech News 2013 ;33 :28 –29 26 Mann DA. Human induced pluripotent stem cell-derived hepatocytes for toxicology testing . Expert Opin Drug Metab Toxicol 2015 ;11 :1 –5 25385341 27 Lu J , Einhorn S , Venkatarangan L , et al. : Morphological and functional characterization and assessment of iPSC-derived hepatocytes for in vitro toxicity testing . Toxicol Sci 2015 ;147 :39 –54 26092927 28 Kim JH , Jang YJ , An SY , et al. : Enhanced metabolizing activity of human ES cell-derived hepatocytes using a 3D culture system with repeated exposures to xenobiotics . Toxicol Sci 2015 ;147 :190 –206 26089346 29 Takebe T , Zhang RR , Koike H , et al. : Generation of a vascularized and functional human liver from an iPSC-derived organ bud transplant . Nat Protoc 2014 ;9 :396 –409 24457331 30 Domansky K , Inman W , Serdy J , Dash A , Lim MH , Griffith LG : Perfused multiwell plate for 3D liver tissue engineering . Lab Chip 2010 ;10 :51 –58 20024050 31 Takebe T , Sekine K , Enomura M , et al. : Vascularized and functional human liver from an iPSC-derived organ bud transplant . Nature 2013 ;499 :481 –484 23823721 32 Vinci M , Gowan S , Boxall F , et al. : Advances in establishment and analysis of three-dimensional tumor spheroid-based functional assays for target validation and drug evaluation . BMC Biol 2012 ;10 :29 22439642 33 Krausz E , de Hoogt R , Gustin E , et al. : Translation of a tumor microenvironment mimicking 3D tumor growth co-culture assay platform to high-content screening . J Biomol Screen 2013 ;18 :54 –66 22923784 34 Zhang M , Luo G , Zhou Y , Wang S , Zhong Z : Phenotypic screens targeting neurodegenerative diseases . J Biomol Screen 2014 ;19 :1 –16 23958650 35 Kim JY , Fluri DA , Marchan R , et al. : 3D spherical microtissues and microfluidic technology for multi-tissue experiments and analysis . J Biotechnol 2015 ;205 :24 –35 25592049 36 Horman SR , To J , Orth AP : An HTS-compatible 3D colony formation assay to identify tumor-specific chemotherapeutics . J Biomol Screen 2013 ;18 :1298 –1308 23918920 37 Wenzel C , Riefke B , Grundemann S , et al. : 3D high-content screening for the identification of compounds that target cells in dormant tumor spheroid regions . Exp Cell Res 2014 ;323 :131 –143 24480576 38 Waite CL , Roth CM : PAMAM-RGD conjugates enhance siRNA delivery through a multicellular spheroid model of malignant glioma . Bioconjug Chem 2009 ;20 :1908 –1916 19775120 39 Sirenko O , Mitlo T , Hesley J , Luke S , Owens W , Cromwell EF : High-content assays for characterizing the viability and morphology of 3D cancer spheroid cultures . Assay Drug Dev Technol 2015 ;13 :402 –414 26317884 40 Hama H , Kurokawa H , Kawano H , et al. : Scale: a chemical approach for fluorescence imaging and reconstruction of transparent mouse brain . Nat Neurosci 2011 ;14 :1481 –1488 21878933 41 Khademhosseini A , Langer R , Borenstein J , Vacanti JP : Microscale technologies for tissue engineering and biology . Proc Natl Acad Sci U S A 2006 ;103 :2480 –2487 16477028 42 Langer R , Tirrell DA : Designing materials for biology and medicine . Nature 2004 ;428 :487 –492 15057821

PMC005xxxxxx/PMC5003005.txt

==== Front Biores Open AccessBiores Open AccessbioresBioResearch Open Access2164-78442164-7860Mary Ann Liebert, Inc. 140 Huguenot Street, 3rd FloorNew Rochelle, NY 10801USA 10.1089/biores.2016.002710.1089/biores.2016.0027Original Research ArticleHepatic Differentiation of Human Induced Pluripotent Stem Cells in a Perfused Three-Dimensional Multicompartment Bioreactor Freyer Nora 1Knöspel Fanny 1Strahl Nadja 1Amini Leila 1Schrade Petra 2Bachmann Sebastian 2Damm Georg 3,4Seehofer Daniel 3,4Jacobs Frank 5Monshouwer Mario 5Zeilinger Katrin 1,*1 Bioreactor Group, Berlin Brandenburg Center for Regenerative Therapies (BCRT), Charité—Universitätsmedizin Berlin, Berlin, Germany.2 Charité Centrum Grundlagenmedizin, Institut für Vegetative Anatomie, Charité—Universitätsmedizin Berlin, Berlin, Germany.3 Department of General-, Visceral- and Transplantation Surgery, Charité—Universitätsmedizin Berlin, Berlin, Germany.4 Department of Hepatobiliary Surgery and Visceral Transplantation, University of Leipzig, Leipzig, Germany.5 Janssen Research and Development, Beerse, Belgium.Part of this work was previously presented at the following meeting: Freyer N, Strahl N, Knöspel F, Urbaniak T, Zeilinger K. Hepatic differentiation of hiPSC in a 3D multicompartment bioreactor, 30th Annual Meeting of the German Association of the Study of the Liver (GASL), 2014, 24 and 25th January, Tübingen, Germany, abstract published in Gastroenterol 52, p. 3.12. * Address correspondence to: Dr. med. vet. Katrin Zeilinger, Bioreactor Group, Berlin Brandenburg Center for Regenerative Therapies (BCRT), Charité—Universitätsmedizin Berlin, Campus Virchow-Klinikum, Augustenburger Platz 1, Berlin 13353, Germany, E-mail: katrin.zeilinger@charite.de01 8 2016 2016 01 8 2016 5 1 235 248 © Nora Freyer et al. 2016; Published by Mary Ann Liebert, Inc.2016This Open Access article is distributed under the terms of the Creative Commons License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.Abstract The hepatic differentiation of human induced pluripotent stem cells (hiPSC) holds great potential for application in regenerative medicine, pharmacological drug screening, and toxicity testing. However, full maturation of hiPSC into functional hepatocytes has not yet been achieved. In this study, we investigated the potential of a dynamic three-dimensional (3D) hollow fiber membrane bioreactor technology to improve the hepatic differentiation of hiPSC in comparison to static two-dimensional (2D) cultures. A total of 100 × 106 hiPSC were seeded into each 3D bioreactor (n = 3). Differentiation into definitive endoderm (DE) was induced by adding activin A, Wnt3a, and sodium butyrate to the culture medium. For further maturation, hepatocyte growth factor and oncostatin M were added. The same differentiation protocol was applied to hiPSC maintained in 2D cultures. Secretion of alpha-fetoprotein (AFP), a marker for DE, was significantly (p < 0.05) higher in 2D cultures, while secretion of albumin, a typical characteristic for mature hepatocytes, was higher after hepatic differentiation of hiPSC in 3D bioreactors. Functional analysis of multiple cytochrome P450 (CYP) isoenzymes showed activity of CYP1A2, CYP2B6, and CYP3A4 in both groups, although at a lower level compared to primary human hepatocytes (PHH). CYP2B6 activities were significantly (p < 0.05) higher in 3D bioreactors compared with 2D cultures, which is in line with results from gene expression. Immunofluorescence staining showed that the majority of cells was positive for albumin, cytokeratin 18 (CK18), and hepatocyte nuclear factor 4-alpha (HNF4A) at the end of the differentiation process. In addition, cytokeratin 19 (CK19) staining revealed the formation of bile duct-like structures in 3D bioreactors similar to native liver tissue. The results indicate a better maturation of hiPSC in the 3D bioreactor system compared to 2D cultures and emphasize the potential of dynamic 3D culture systems in stem cell differentiation approaches for improved formation of differentiated tissue structures. Keywords:  stem cellstissue engineering ==== Body Introduction During drug development, only one out of nine compounds gets approved by the regulatory authorities, usually due to a lack of efficacy or toxic side effects.1 Thus, models for assessment of drug toxicity, especially hepatotoxicity, in the early phase of drug development are needed. Animal models, although indispensable in preclinical studies, are not sufficiently predictive for humans due to interspecies differences.2 Primary human hepatocytes (PHH) have been widely accepted as the gold standard for predictive in vitro studies on hepatic drug toxicity.3 However, PHH display a huge variation in cell function and enzyme activities because of interdonor variances,4 and the high demand of freshly isolated PHH is difficult to address due to the scarce availability of human liver tissue. Human induced pluripotent stem cells (hiPSC) represent a promising cell source for the generation of human hepatocytes for studies on hepatic drug toxicity. Due to the unlimited self-renewing capacity of hiPSC, they provide the option for cell production in large amounts and at a constant quality. In addition, variances due to genetic polymorphism can be investigated by using different hiPSC lines representative of individual patient groups.5 Several protocols have been established to generate stem cell-derived hepatocytes from human pluripotent stem cells.6–9 These procedures mimic the embryonic development of the liver by adding different growth factors necessary for each developmental stage. The resulting hepatocyte-like cells (HLC) were successfully applied for in vitro studies on human drug exposure,10,11 hepatitis B and C infection,12,13 or malaria pathogenesis14 among others, and they have been shown to repopulate the livers of chimeric mice and rescue the disease phenotype in these animals.15 However, the HLC obtained with existing protocols still show an immature phenotype with reduced hepatic functionality when compared to PHH.16,17 To overcome these drawbacks, improved culture models are demanded, which address the needs of the cells in their natural environment. Several studies have shown that three-dimensional (3D) culture of PHH in natural or synthetic scaffolds supports cell–cell contacts, cell polarization, and preservation of liver functions such as cytochrome P450 (CYP) activities, albumin production, and glycogen synthesis.18–20 To improve oxygenation and medium exchange in hepatocyte cultures, various perfused 3D culture systems have been developed.21–23 In the 3D multicompartment bioreactor used in this study, the cells are maintained in a perfused 3D environment allowing for physiological signal exchange and autocrine or paracrine stimulation, close to the natural situation in the organ. We have previously shown that this 3D bioreactor system supports stable culture of PHH under serum-free conditions24,25 and is suitable for differentiation of human embryonic stem cells (hESC).26,27 Thus, we hypothesize that the usage of the 3D bioreactor system could improve the hepatic maturation and liver-specific functionality of hiPSC-derived hepatocytes compared with conventional two-dimensional (2D) cultures. The functionality of the cells upon differentiation in 2D cultures or 3D bioreactors was evaluated by measurement of typical hepatocyte products (albumin, urea) and CYP activities. Cultures were further characterized by means of immunohistochemical investigations, transmission electron microscopy (TEM), and analysis of liver-specific mRNA expression. Data from hiPSC-derived differentiated cells were compared to those from freshly isolated or 2D cultured PHH. Materials and Methods Bioreactor technology The 3D multicompartment bioreactor consists of three independent, but interwoven hollow fiber capillary systems that serve for counter-current medium perfusion (two medium compartments). Cells are supplied with oxygen by direct membrane oxygenation through integrated gas capillaries (gas compartment), which are perfused with an air/CO2 mixture. Cells are cultured in the extracapillary space (cell compartment). The analytical scale bioreactors used in this study have a cell compartment volume of 2 mL. A detailed description of the technology is provided elsewhere.28 Bioreactors are operated in a perfusion device with two modular pump units, one for medium recirculation and one for medium feed. The bioreactor incubation chamber is heated by two heating units located inside the chamber, each consisting of a heating cartridge and a fan. A platinum measuring resistor monitors the temperature inside the chamber and software is used to set and maintain the desired temperature. Gas flow rates and gas compositions are regulated using electronically operated gas valves for air, CO2, and the resulting gas mixture (Vögtlin Instruments). Bioreactors, tubing systems, and perfusion devices were manufactured by Stem Cell Systems. Hepatic differentiation of hiPSC in 3D bioreactors or 2D cultures The hiPSC line DF6-9-9T29 (WiCell Research Institute) was cultured under feeder-free conditions on Nunclon™ six-well cell culture plates (ThermoScientific Nunc™) coated with 8.68 μg/cm2 Matrigel (growth factor reduced). Cells were expanded with the mTeSR™1 medium (StemCell Technologies) with 0.05 mg/mL gentamicin (Merck). Afterward, a total of 100 × 106 hiPSC were seeded into a precoated bioreactor (8.68 μg/cm2; Matrigel) and cultured over a total of 20 days. Bioreactors were maintained at 37°C, the medium recirculation rate was 10 mL/min, and the feed rate was 1 mL/h. Based on daily measurements of the pH, glucose, and lactate values, CO2 and medium perfusion rates were adjusted, if necessary, to maintain a stable pH between 7.2 and 7.4 and sufficient glucose levels (>25 mg/dL). After a proliferation phase of 3 days with mTeSR™1, differentiation of the cells in 3D bioreactors was induced based on the protocols described by Hay et al.6,30,31 for 2D cultures. In the first step, differentiation into definitive endoderm (DE) was induced by perfusion with the Roswell Park Memorial Institute (RPMI) 1640 culture medium (Merck) supplemented with 100 ng/mL activin A (Peprotech), 50 ng/mL Wnt3a (R&D Systems), 1 μM sodium butyrate (Sigma-Aldrich), and 2% (v/v) B27 supplements without insulin (Life Technologies) for 3 days. Subsequently, bioreactors were perfused over 13 days with a hepatocyte culture medium consisting of basal medium and single quots (Lonza) and 10 ng/mL hepatocyte growth factor (HGF; Peprotech) to induce differentiation of DE-cells to hepatoblasts. For further maturation to HLC, 10 ng/mL oncostatin M (Peprotech) was added during the last 4 days of differentiation. 2D cultures were performed in parallel with 3D bioreactor cultures for control, applying the same differentiation protocol with daily medium exchange. Metabolic parameters The metabolic activity of the cells was assessed by daily measurement of glucose and lactate concentrations with a blood gas analyzer (ABL 700; Radiometer). Potential cell damage was detected by analyzing the release of lactate dehydrogenase (LDH) using an automated clinical chemistry analyzer (Cobas® 8000; Roche Diagnostics) as well as the production of urea and the albumin precursor protein alpha-fetoprotein (AFP). Albumin secretion, as a marker for mature hepatocytes, was detected using an ELISA Quantitation kit and tetramethylbenzidine (TMB) substrate (both from Bethyl Laboratories) according to the manufacturer's instructions. Culture of PHH The PHH were isolated from macroscopically healthy tissue that remained from resected human liver of patients with primary or secondary liver tumors or benign local liver diseases. Informed consent of the patients for the use of tissue for research purposes was obtained according to the ethical guidelines of the Charité—Universitätsmedizin Berlin. Part of the tissue sample was fixed in formaldehyde for immunofluorescence staining. Cell isolation was performed according to Pfeiffer et al.32 Hepatocytes were seeded at a density of 2.0 × 105 cells/cm2 in six-well plates (BD Sciences) coated with rat tail collagen. Cells were cultivated using Heparmed Vito 143 supplemented with 10% fetal calf serum (PAA), 0.8 mg/mL insulin, 5 mg/L transferrin, 0.003 mg/L glucagon, 100 U/mL penicillin, and 100 μg/mL streptomycin (all Merck). Measurement of cytochrome P450 (CYP) isoenzyme activities Activities of the pharmacologically relevant CYP isoenzymes CYP1A2, CYP2B6, CYP2C9, and CYP3A4 were measured in (1) undifferentiated hiPSC, (2) HLC differentiated in 2D cultures or (3) 3D bioreactors, and (4) PHH cultures (24 h after seeding) serving as control. The cells were incubated with a cocktail containing phenacetin (Sigma) as a substrate for CYP1A2, bupropion (Toronto Research Chemicals) as a substrate for CYP2B6, diclofenac as a substrate for CYP2C9, and midazolam as a substrate for CYP3A4/5 (both from Sigma-Aldrich). Samples were taken at 1, 2, 4, and 6 h subsequent to substrate application. An overview of the used substrates, their final concentrations, and the corresponding CYP isoenzymes is provided in Table 1. Table 1. CYP Isoenzymes Tested and Their Corresponding Substrates with Resulting Products and Applied Concentrations Enzyme Substrate Product Final concentration [μM] Transition reactions for analysis of probe products CYP1A2 Phenacetin Acetaminophen 100 152 → 110 CYP2B6 Bupropion 6-Hydroxybupropion 500 256 → 238 CYP2C9 Diclofenac 4′-Hydroxydiclofenac 25 312 → 266 CYP3A4/5 Midazolam 1′-Hydroxymidazolam 10 342 → 324 Formed metabolites (acetaminophen, 6-OH-bupropion, 4-OH-diclofenac, 1-OH-midazolam) were quantified by liquid chromatography tandem-mass spectrometry. Deuterated 1-OH-midazolam was added as an internal standard. Separation was carried out using an Acquity UPLC C18 column and eluted fractions were directly passed through a Xevo TQ-S tandem mass spectrometer (both from Waters Corp.). Acquired data were processed with Thermo Xcaliber 20 software (Thermo Scientific). Gene expression analysis RNA was isolated from undifferentiated hiPSC, from HLC differentiated in 2D cultures or 3D bioreactors, and from freshly isolated PHH serving as reference cells. RNA isolation and subsequent cDNA synthesis were performed as described elsewhere.33 Each cDNA template was mixed with polymerase chain reaction (PCR) Master mix (Applied Biosystems) and human-specific primers and probes (TaqMan Gene Expression Assay system; Life Technologies; Table 2). Quantitative real-time PCR (qRT-PCR) was performed using a Real time cycler (Mastercycler ep Realplex 2; Eppendorf). The expression of specific genes was normalized to that of the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and fold changes of expression levels were calculated with the ΔΔCt method.34 Table 2. Applied Biosystems TaqMan Gene Expression Assays® Gene symbol Gene name Assay ID POU5F1 POU domain, class 5, transcription factor 1 HS00999632_g1 NANOG Nanog homeobox HS02387400_g1 SOX7 SRY-box 7 HS00846731_s1 SOX17 SRY-box 17 HS00751752_s1 AFP Alpha fetoprotein HS00173490_m1 ALB Albumin HS00910225_m1 CYP1A2 Cytochrome P450 family 1 subfamily A member 2 HS00167927_m1 CYP2B6 Cytochrome P450 family 2 subfamily B member 6 HS03044634_m1 CYP2C9 Cytochrome P450 family 3 subfamily A member 4 HS00426397_m1 CYP3A4 Cytochrome P450 family 3 subfamily A member 4 HS00604506_m1 GATA2 GATA binding protein 2 HS00231119_m1 NEFL Neurofilament, light polypeptide HS00196245_m1 GAPDH Glyceraldehyde-3-phosphate dehydrogenase HS03929097_g1 Immunofluorescence studies Immunofluorescence staining was performed with antibodies listed in Table 3 as described elsewhere33 with following changes for 3D and tissue sections. The hollow fiber bed was excised en-bloc, fixed with 4% formaldehyde solution (Herbeta Arzneimittel), dehydrated, paraffinized, and cut into slides of 2.5 μm thickness. Subsequently, the slides were deparaffinized, rehydrated, and subjected to antigen retrieval in a citrate buffer (pH 6.0) in a pressure cooker for 15 min. The same procedure was applied to native human liver tissue serving as a positive control. Nuclei in native human liver tissue were stained with bisBenzimide H 33342 trihydrochloride (Sigma). Table 3. Antibodies Used for Immunofluorescence Staining   Protein symbol Species Manufacturer Cat.-No. Final conc. [μg/mL] Primary antibody Alpha fetoprotein AFP mouse Santa Cruz Sc-8399 2 Albumin ALB mouse Sigma A6684 67 Cytokeratin 18 CK18 mouse Santa Cruz Sc-6259 2 Cytokeratin 19 CK19 rabbit Santa Cruz Sc-25724 2 Cytochrome P450 1A2 CYP1A2 mouse Santa Cruz Sc-53614 2 Cytochrome P450 2B6 CYP2B6 rabbit Santa Cruz Sc-67224 2 Hepatocyte nuclear factor 4 alpha HNF4A rabbit Santa Cruz Sc-8987 2 Marker of proliferation Ki-67 MKI67 mouse BD Biosciences 556003 10 Multidrug resistance-associated protein 2 MRP2 rabbit Sigma M8316 9 POU domain, class 5, transcription factor 1 OCT3 rabbit Santa Cruz Sc-9081 2 Stage-specific embryonic antigen 4 SSEA4 mouse R&D Systems MAB1435 10 Tight junction protein 1 TJP1 mouse LSBio LS B2228 30 Secondary antibody Alexa Fluor® 488 anti-mouse   goat Life Technologies A-11029 2 Alexa Fluor 594 anti-rabbit   goat Life Technologies A-11037 2 Final concentrations are given in μg/mL. Ki-67 positive cells were quantified in undifferentiated hiPSC, 2D cultured DE cells, and HLC from 2D cultures or 3D bioreactors with the open source image processing program ImageJ using at least 10 randomly chosen visual fields for each group. Transmission electron microscopy For TEM, part of the 3D bioreactor cell compartment was fixed with 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (both from Serva) over night, followed by postfixing in 1% OsO4 (Science Services) with 0.8% potassium ferrocyanide (Merck) in 0.1 M sodium cacodylate buffer for 1.5 h. Subsequently, samples were progressively dehydrated in ethanol and then embedded in Epon (Serva). Ultrathin sections were stained with uranyl acetate and Reynold's lead citrate (Merck) and microphotographs were taken using an electron microscope EM 906 (Carl Zeiss). Statistical evaluation Experiments were performed in triplicates, unless stated otherwise, and results are presented as mean ± standard error of the mean. The area under curve (AUC) was calculated for time courses of biochemical parameters and differences between culture systems were detected with a subsequent two-tailed Student's t-test. CYP activities of undifferentiated hiPSC were compared with those of HLC from 2D cultures or 3D bioreactors by one-way analysis of variance (ANOVA) followed by Bonferroni's multiple comparison test. Differences in gene expression between both culture systems were detected by applying an unpaired, two-tailed Student's t-test. PHH were compared with all other groups by one-way ANOVA followed by Dunnett's multiple comparison test. Differences were judged as significant if the p-value was <0.05. Results Metabolic activity and integrity of HLC in 2D cultures or 3D bioreactors In both culture systems, the metabolic activity of the cells during hepatic differentiation, assessed by glucose consumption and lactate production, showed an increase over time (Fig. 1A, B). Average glucose consumption rates were significantly higher (p ≤ 0.05) in 2D cultures than in 3D bioreactors (Fig. 1A). The time course of lactate production reflected that of glucose consumption (Fig. 1B). LDH release, indicating cell injury, was significantly higher (p ≤ 0.01) in 2D cultures than in 3D bioreactors during the first 3 days of differentiation, but decreased from day 3 on to basal levels. In contrast, LDH release in 3D bioreactors remained on a basal level over the whole culture period (Fig. 1C). FIG. 1. Metabolic activity of hiPSC during hepatic differentiation in 2D cultures (dotted line) or in 3D bioreactor cultures (black line). (A) Glucose consumption, (B) lactate production, (C) release of LDH, (D) secretion of AFP, (E) albumin production, and (F) urea secretion. Values were normalized to 1 × 106 inoculated cells. AUC was calculated and differences were detected with the unpaired, two-tailed Student's t-test (3D bioreactors: n = 3, 2D cultures: n = 4, mean ± SEM). AFP, alpha-fetoprotein; AUC, area under curve; 2D, two dimensional; 3D, three dimensional; hiPSC, human induced pluripotent stem cells; LDH, lactate dehydrogenase; SEM, standard error of the mean. Secretion of stage-specific proteins and metabolites The synthesis of liver-specific proteins and the activity of the urea cycle were evaluated during hepatic differentiation. The endoderm-specific marker AFP showed an increase during the hepatoblast differentiation stage, starting on day 7 in 2D cultures and on day 9 in 3D bioreactors, but the increase was significantly higher (p ≤ 0.05) in 2D cultures (Fig. 1D). Albumin also showed an increase during the hepatoblast differentiation phase beginning on day 12 of differentiation in 2D cultures and 3D bioreactors. However, maximum values achieved were thrice higher in 3D bioreactors than in 2D cultures (Fig. 1E), although the difference was not significant. Values of urea secretion showed a constant increase in 3D bioreactors until day 17 to the threefold of secretion rates detected on day 1 of culture. A fluctuating time course was observed in 2D cultures with peaks at the beginning of culture and between day 9 and 13 (Fig. 1F). Values approximated those of 3D bioreactors during the final days of differentiation. Functional analysis of different cytochrome P450 (CYP) isoenzymes To investigate the CYP activity, the ability of the HLC to metabolize various substrates into their isoenzyme-specific products was analyzed and compared with CYP activities of undifferentiated hiPSC and those of PHH cultures determined 24 h after seeding. CYP1A2 showed a higher activity in 3D bioreactors than in 2D cultures, while it was lower compared with undifferentiated hiPSC or PHH (Fig. 2A). A significant CYP2B6-dependent conversion of bupropion to 6-OH-bupropion was observed in 3D bioreactors compared to 2D cultures (p ≤ 0.05 Fig. 2B). In contrast, CYP3A4 showed higher activities in undifferentiated hiPSC and in 2D cultures than in 3D bioreactors, although the differences were not significant (Fig. 2C). The isoenzyme CYP2C9 activity was detected neither in differentiated nor in undifferentiated hiPSC (data not shown). All enzyme activities investigated were significantly higher in PHH than in the HLC generated from hiPSC in 2D or 3D cultures (Fig. 2A–C). FIG. 2. Activities of different cytochrome P450 (CYP) isoenzymes in undifferentiated hiPSC (white), in hiPSC after hepatic differentiation in 2D cultures (gray) or 3D bioreactors (black), or in PHH (dotted). CYP activities were determined by measuring the conversion rates of selected substrates into isoenzyme-specific products. (A) Formation of acetaminophen from phenacetin by CYP1A2, (B) formation of 6-OH-bupropion from bupropion by CYP2B6, and (C) formation of 1-OH-midazolam from midazolam by CYP3A4/5. Differences in metabolic activity between undifferentiated hiPSC, 2D cultures and 3D bioreactors, were calculated using one-way ANOVA with Bonferroni's multiple comparison test (solid line). In addition, differences between PHH to all other groups were calculated using one-way ANOVA with Dunnett's multiple comparison test (dotted line). (3D bioreactors: n = 3, 2D cultures and undifferentiated hiPSC: n = 4, PHH: n = 5; mean ± SEM), *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. ANOVA, analysis of variance; PHH, primary human hepatocytes. Gene expression of pluripotency factors, and endodermal and hepatic markers The expression of 12 stage-specific genes was comparatively analyzed by qRT-PCR in 2D or 3D cultures of HLC and in freshly isolated PHH (Fig. 3). Expression levels were plotted relative to those of undifferentiated hiPSC (d0) to investigate the hepatic maturation state of the cells. FIG. 3. Gene expression of pluripotency markers (A) Oct-3/4 and (B) Nanog, mesodermal marker (C) GATA-2, ectodermal marker (D) neurofilament, endodermal markers (E) AFP and (F) SOX 17, extra-embryonic marker (G) SOX 7 and hepatic markers (H) albumin, (I) CYP1A2, (J) CYP2B6, (K) CYP2C9 and (L) CYP3A4/5 in hiPSC after hepatic differentiation in 2D cultures or 3D bioreactors, and in PHH relative to undifferentiated hiPSC (d0). Samples for mRNA expression analysis were taken after hepatic differentiation of hiPSC in 2D cultures (black) or 3D bioreactors (gray). For mRNA expression analysis of PHH (dotted), freshly isolated cells were used. Fold changes relative to undifferentiated hiPSC were calculated with normalization to GAPDH expression by the ΔΔCt method. Differences in gene expression between 2D cultures and 3D bioreactors were calculated using the unpaired, two-tailed Student's t-test (solid line). In addition, differences between PHH and all other groups were calculated by means of one-way ANOVA with Dunnett's multiple comparison test (dotted line) (3D bioreactors: n = 3, 2D cultures and undifferentiated hiPSC: n = 4, PHH: n = 3; mean ± SEM), *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. The expression of the pluripotency markers POU5F1 and NANOG decreased in both culture systems compared to undifferentiated hiPSC, but the decrease was significantly higher in 2D cultures (p ≤ 0.05, Fig. 3A, B). PHH showed POU5F1 levels similar to HLC in 2D cultures, while NANOG expression was similar to those in 3D bioreactors (Fig. 3A, B). To exclude differentiation into nonendodermal cells, markers for mesoderm (GATA2) and endoderm (neurofilament, NEFL) were analyzed as well. The analysis revealed a >100-fold increase for GATA2 in 3D bioreactors, which is significantly higher than in 2D cultures showing an ∼20-fold increase (Fig. 3C). The expression of NEFL did not change significantly in comparison to undifferentiated hiPSC (Fig. 3D). PHH showed only scarce expression of NEFL and GATA2 (Fig. 3C, D). In contrast, most markers for endoderm and mature hepatocytes increased in both culture systems. AFP showed a >60,000-fold increased expression in both 2D cultures and 3D bioreactors, relative to undifferentiated hiPSC (Fig. 3E), and the expression of SOX17 (Fig. 3F), another endodermal marker, showed a 20-fold increase, while PHH showed no relevant expression of those markers. Since AFP and SOX17 are markers for both, definitive and extraembryonic endoderm, SOX7 (Fig. 3G) was analyzed as a specific marker for extraembryonic endoderm. The expression of SOX7 was increased in both culture systems and was similar to that detected in PHH. Among the markers investigated for mature hepatocytes, the highest increase was detected for albumin (ALB), which was increased in both groups by >2,000-fold, with 3D bioreactors showing twice the expression compared to 2D cultures (Fig. 3H). In contrast, the expression of CYP1A2 was decreased compared with undifferentiated hiPSC both in 2D cultures and 3D bioreactors (Fig. 3I), which is in line with the results from activity measurements (Fig. 2A). A decrease of mRNA expression was also detected for CYP2B6, but the expression was still significantly higher in 3D bioreactors (p ≤ 0.05, Fig. 3J) and is in accordance to the findings from activity measurements (Fig. 2B). Expression of CYP2C9 and CYP3A4/5 was increased by around 40- or five-fold, respectively, compared with undifferentiated cells showing no significant differences between 2D cultures and 3D bioreactors (Fig. 3K, L). All markers for mature hepatocytes showed a distinctly lower expression in HLC than in PHH, irrespective of the culture system. Immunohistochemical characterization of hiPSC-derived differentiated cells To determine the amount of proliferating cells before, during, and after the hepatic differentiation of hiPSC, an immunofluorescence staining with the proliferation marker Ki-67 was performed. In cultures of undifferentiated hiPSC, almost all cells (97.7% ± 1.6%) were positive for Ki-67 (Fig. 4A1). In 2D cultured DE cells, approximately two thirds of the cells (63.6% ± 12.6%) were proliferating (picture not shown). After hepatic differentiation, around one third of the cells were still proliferating in 2D cultures (32.6% ± 20.4%, Fig. 4B1) and 3D bioreactors (30.3% ± 18.7%, Fig. 4C1). In the native human liver, no Ki-67-positive cells could be detected (Fig. 4D1). FIG. 4. Immunofluorescence analysis of hiPSC before and after hepatic differentiation in 2D cultures or 3D bioreactors compared with native human liver tissue. Samples from cultures or liver tissue were stained with (A1–D1) Ki-67, (A2–D2) SSEA-4 and Oct-3/4, (A3–D3) albumin and CK19, (A4–D4) CK18 and CK19, (A5–D5) AFP and HNF4α, (A6–D6) TJP1 and MRP2, and with (A7–D7) CYP1A2 and CYP2B6. Nuclei were counterstained with DAPI (blue) or with bisBenzimide H 33342 trihydrochloride in native human liver tissue (blue). CK19, cytokeratin 19. The pluripotency markers SSEA4 and OCT3 were expressed in the majority of undifferentiated hiPSC (Fig. 4A2), but could not be detected in hiPSC-derived HLC in 2D cultures (Fig. 4B2) or 3D bioreactors (Fig. 4C2), or in native human liver tissue (Fig. 4D2). Regarding the expression of hepatocyte-specific markers, the majority of cells was positive for albumin in 2D cultures as well as 3D bioreactors (Fig. 4B3, C3), similar to native human liver tissue (Fig. 4D3). Cytokeratin 19 (CK19), characteristic for biliary cells, was only detected in a few sparsely distributed cells in 2D cultures (Fig. 4B4), whereas in 3D bioreactors, ring-shaped cell arrangements lined with cells positive for CK18 and CK19 were observed, indicating formation of bile duct-like tubular structures (Fig. 4C4) resembling those occurring in native human liver tissue (Fig. 4D4). Staining for hepatocyte nuclear factor 4-alpha (HNF4A) showed a few positive cells, which were negative for AFP in both 2D cultures and 3D bioreactors (Fig. 4B5, C5). This is in contrast to native human liver tissue where almost all cells showed immunoreactivity for HNF4A (Fig. 4D5). Staining of tight junction protein (TJP1) and the transporter multidrug resistance-associated protein 2 (MRP2) revealed no immunoreactivity in 2D cultures (Fig. 4B6), while in 3D bioreactors (Fig. 4C6), abundant cells were positive for TJP1 as indicated by thin borders between the cells. In the native liver tissue, most of the cells were double positive for MRP2 and TJP1 (Fig. 4D6). Immunofluorescence analysis of CYP isoenzymes showed that most of the cells in 2D cultures and 3D bioreactors were positive for CYP1A2, but negative for CYP2B6 (Fig. 4B7, C7), which is in line with the functional analysis showing higher activities for CYP1A2. In 3D bioreactors, the inner ring of tubular-like structures lacked CYP1A2 immunoreactivity indicating a zonation of the cells. In native human liver tissue, cells were positive for both CYP isoenzymes (Fig. 4D7). The undifferentiated hiPSC were negative for all liver-specific markers (Fig. 4A3–A7). Ultrastructural characteristics of hiPSC-derived differentiated cells Ultrastructural studies of hiPSC-derived cells after hepatic differentiation in 3D bioreactors showed a heterogeneous cell population with some similarities to native liver tissue. Cells displaying a high nucleus to cytoplasm ratio indicate the presence of still immature cells (Fig. 5A). In other areas, cells showed microvilli at their apical side with abundant cell–cell contacts consisting of tight junctions and desmosomes indicating cell polarization typical for hepatocytes (Fig. 5B, C). The majority of cells contained numerous mitochondria, which is also a typical feature of hepatocytes due to their high metabolic activity. Distinct interdigitations between neighboring cells and rough endoplasmic reticulum in the cytoplasm were observed in addition to well-developed Golgi apparatuses (Fig. 5D). FIG. 5. Ultrastructural characteristics of hiPSC after hepatic differentiation in 3D bioreactors. (A) Cells with a high nucleus:cytoplasm ratio indicating immature cells (N). (B) Cells with distinct microvilli (Mv) and abundant cell–cell contacts (arrows) between neighboring cells. (C) Tight junction (TJ) and desmosome (De) between two neighboring cells. (D) Interdigitations (Id) between two cells with mitochondria (M), rough endoplasmic reticulum (rER), and Golgi apparatus (G). Discussion Hepatocytes derived from hiPSC represent a promising cell source for pharmacological toxicity testing. Although numerous protocols for the hepatic differentiation of hiPSC exist and are continuously refined, the obtained cells still show a fetal phenotype,16,17 and the maturation state of the obtained cells needs to be further improved. In this study, we investigated the potency of a dynamic 3D bioreactor technology to promote the hepatic maturation of hiPSC-derived hepatocytes when compared to conventional 2D cultures. The differentiation outcome of hiPSC cultured in 2D cultures or 3D bioreactors was analyzed by means of metabolic, phenotypical, and functional parameters. To allow a direct comparison of the two culture systems, the results for secreted metabolites and proteins as well as for CYP activities were normalized to the initial cell number. Potential proliferation activities of HLC in 2D cultures or 3D bioreactors were quantified by means of Ki-67 staining. The protein Ki-67 is expressed during all active phases of the cell cycle (G1, S, G2, mitosis), but is absent in resting cells (G0).35 The quantitative analysis of the Ki-67 staining revealed that at the end of hepatic differentiation, there was still mitotic activity. However, since the amount of proliferating cells with ∼30% was comparable in 2D cultures and 3D bioreactors, it can be concluded that the observed differences in metabolite/protein secretion and CYP activities are not associated with differences in cell growth. Glucose consumption rates and lactate production rates as indicators for energy metabolism were doubled in 2D cultures compared to 3D bioreactors. This could be caused by the formation of 3D cell aggregates within the bioreactor leading to a gradient and a differential metabolic activity of cells growing either on the surface or in the center of the aggregates. It has been reported that concentration gradients influence cell differentiation and tissue formation.36,37 Thus, formation of gradients could also be a reason for the heterogeneous cell population observed in 3D bioreactors as indicated by immunohistochemical analyses and increased GATA2 mRNA expression. In the described hollow fiber bioreactor, the aggregate size is controlled by the distance between the capillaries. In other bioreactor technologies such as stirred tank vessels, the aggregate size is influenced by the impeller design38 and the stirring velocity,39 but stirring can cause shear stress and cell damage.40 Another bioreactor technology with minimized shear stress is the rotating-wall vessel, which consists of a horizontally rotating cylindrical culture vessel with a coaxial tubular oxygenator avoiding nutrient compartmentalization and barrier formation.41 Furthermore, a significantly lower LDH release was detected in 3D bioreactors compared with 2D cultures. One possible reason could be a protective effect of 3D cell aggregates against damage caused by the applied activin A, which can induce apoptosis.42 However, LDH is released during both necrosis and apoptosis43 and therefore does not allow distinguishing between necrosis and apoptosis. An apoptosis assay would be needed to clarify whether the observed LDH release is due to apoptosis processes. AFP expression and secretion were detected in HLC in both 2D cultures and 3D bioreactors, with significantly higher secretion rates in 2D cultures. In contrast, albumin secretion showed a distinctly higher increase in 3D cultures than in 2D cultures. In vivo AFP is only expressed during liver embryogenesis, in fetal liver cells, and hepatocellular carcinomas,44 while albumin is produced by differentiated hepatocytes in the adult liver. Thus, the findings of AFP and ALB expression indicate a higher maturation grade of HLC in 3D bioreactors than in 2D cultures. Expression of AFP in HLC in association with expression of markers for mature hepatocytes such as ALB and various CYP isoenzymes has been described in several studies.7,45,46 Gieseck et al.47 also observed lower AFP secretion rates and higher albumin secretion rates in 3D clump cultures compared to 2D cultures, which emphasizes the value of 3D culture systems to promote hepatic maturation of hiPSC. Since albumin secretion was still increasing at the end of hepatic differentiation, a further maturation could be achieved by prolongation of the differentiation period. This finding is supported by the observed increase of urea secretion in 3D bioreactors. Urea is the major end product of protein nitrogen metabolism and is synthesized by the urea cycle in the liver from ammonia. Hence, urea secretion is an important functional marker for mature hepatocytes. To further validate results from urea secretion, the expression of enzymes from the urea cycle such as carbamoyl phosphate synthetase I or transcarbamylase could be analyzed. In addition to AFP and SOX17 as markers for both, definitive and extraembryonic endoderm, SOX7 was analyzed as a specific marker for extraembryonic endoderm. The gene expression of SOX7 was similar in HLC in 2D cultures, in 3D bioreactors, and in PHH, indicating that the hiPSC-derived HLC did not contain significant amounts of extraembryonic cells, which confirms the hepatic origin of AFP and also of SOX17 in this study. According to the Human Protein Atlas, the SOX7 expression is strongest in the placenta, but a weak expression is also found in the human liver.48 The ability to metabolize drugs through specific CYP isoenzymes is crucial for potential application of hiPSC-derived HLC in pharmacological studies. Functional activities of CYP1A2 and CYP3A4 were detected in both 2D cultures and 3D bioreactors, while the CYP2B6 activity was observed only in 3D bioreactors. However, all CYP isoenzyme activities were significantly lower in HLC than in PHH. Data from other studies also showed a lower expression and activity of CYPs in differentiated hiPSC compared to PHH.5,49 However, since different experimental conditions were used in those studies, a direct comparison of CYP activities from different publications is difficult. Measurements of mRNA expression levels revealed that the pluripotency markers POU5F1 and NANOG were stronger downregulated in 2D cultures than in 3D bioreactors. This could be explained by the formation of an activin A gradient within the 3D cell aggregates. Activin A recapitulates the nodal signaling pathway in vitro, which stimulates DE derivation from pluripotent stem cells.50 In lower concentrations, activin A is used to maintain pluripotency of hiPSC51 and hESC.52 This could imply that DE formation is decreased in the center of the cell aggregates and other factors such as dimethylsulfoxide (DMSO) might be required to block the effect of activin A on the pluripotency.53 On the other hand, the formation of gradients of activin A and other factors might also promote 3D tissue formation from HLC by reflecting the physiological situation in the liver lobule, which is characterized by an oxygen gradient between the periportal and the perivenous area. This is supported by immunohistochemical and TEM pictures showing tissue-like organization of the cells within 3D bioreactors, including bile duct-like structures characterized by CK18/CK19 double staining. Miki et al.26 also observed the formation of bile duct-like structures after hepatic differentiation of hESC in the 3D bioreactor. Bile ducts are lined with cholangiocytes, which, like the hepatocytes, originate from the bipotent hepatoblasts.54 Current protocols for direct differentiation of pluripotent stem cells to cholangiocytes and their in vitro generation from hepatoblasts are to some extent similar to protocols used for hepatocyte differentiation since they use factors such as epidermal growth factor, insulin, and hydrocortisone.55,56 De Assuncao et al.55 observed that cholangiocyte markers already became apparent during the hepatic progenitor phase, which would explain the appearance of cells positive for CK19 in this study. In addition, abundant tight junctions were detected by immunohistochemical staining and ultrastructural analysis using TEM. The presence of tight junctions is a prerequisite for the formation of bile canaliculi as a characteristic feature of differentiated hepatocytes. In contrast to tight junction protein, the transporter protein MRP2, which is expressed in the canalicular (apical) membrane area of the hepatocyte and contributes to biliary transport, was negative both in 2D cultures and 3D bioreactor cultures, indicating incomplete cell polarization. The detailed comparison of hiPSC-derived HLC with PHH showed that further improvement of HLC maturation is needed. This could be achieved by combining other promising differentiation strategies with the 3D bioreactor culture: for example, differentiation into the mesodermal direction during DE differentiation, detected by GATA2 expression in the 3D bioreactor, could be excluded by adding specific inhibitors57 and cocultivation with primitive endothelial cells or mesenchymal stem cells could recapitulate the in vivo organogenesis even more closely in the 3D culture.15 In addition, Kim et al.46 could show that repeated stimulation of HLC with xenobiotics could improve their metabolizing activity. A further strategy is based on the usage of an hiPSC line, which is derived from primary hepatoblasts as already shown for mouse iPSC.58 This could also improve the hepatic differentiation since recent studies suggest that iPSC retain a residual donor cell memory, which may impact their capacity to differentiate into the cell type of origin.59 Conclusion In conclusion, the hepatic maturation of hiPSC-derived HLC was improved in 3D bioreactors compared with 2D cultures in terms of albumin secretion, CYP2B6 activity, and formation of tissue-like structures with cell–cell contacts. However, some aspects of hepatic functionality were still insufficient, for example, AFP was still expressed, and the activity of CYP2C9 and expression of the biliary transporter MRP2 could not be detected. The 3D bioreactor could be used to investigate approaches to improve the maturation of hiPSC-derived hepatocytes and generate fully differentiated hepatocytes from hiPSC in a physiological-like 3D environment in the future. Acknowledgments The research leading to these results has received support from the Innovative Medicines Initiative Joint Undertaking under grant agreement no. 115439, resources of which are composed of financial contribution from the European Union's Seventh Framework Programme (FP7/2007-2013) and EFPIA companies' in kind contribution. This article reflects only the author's views and neither the IMI JU, EFPIA, nor the European Commission is liable for any use that may be made of the information contained therein. Author Disclosure Statement No competing financial interests exist. Abbreviations Used ANOVAanalysis of variance AUCarea under curve DEdefinitive endoderm hESChuman embryonic stem cells HGFhepatocyte growth factor hiPSChuman induced pluripotent stem cells HLChepatocyte-like cells LDHlactate dehydrogenase PHHprimary human hepatocytes TEMtransmission electron microscopy ==== Refs References 1 Kola I , Landis J Can the pharmaceutical industry reduce attrition rates? Nat Rev Drug Discov . 2004 ;3 :711 –715 15286737 2 Lewis DF , Ioannides C , Parke DV Cytochromes P450 and species differences in xenobiotic metabolism and activation of carcinogen . Environ Health Perspect . 1998 ;106 :633 –641 9755138 3 Giri S , Bader A Improved preclinical safety assessment using micro-BAL devices: the potential impact on human discovery and drug attrition . Drug Discov Today . 2011 ;16 :382 –397 21354326 4 Ingelman-Sundberg M Pharmacogenetics of cytochrome P450 and its applications in drug therapy: the past, present and future . Trends Pharmacol Sci . 2004 ;25 :193 –200 15063083 5 Takayama K , Morisaki Y , Kuno S , et al. Prediction of interindividual differences in hepatic functions and drug sensitivity by using human iPS-derived hepatocytes . Proc Natl Acad Sci U S A . 2014 ;111 :16772 –16777 25385620 6 Hay DC , Zhao D , Ross A , et al. Direct differentiation of human embryonic stem cells to hepatocyte-like cells exhibiting functional activities . Cloning Stem Cells . 2007 ;9 :51 –62 17386014 7 Sullivan GJ , Hay DC , Park IH , et al. Generation of functional human hepatic endoderm from human induced pluripotent stem cells . Hepatology . 2010 ;51 :329 –335 19877180 8 Hannan NR , Segeritz CP , Touboul T , et al. Production of hepatocyte-like cells from human pluripotent stem cells . Nat Protoc . 2013 ;8 :430 –437 23424751 9 Szkolnicka D , Farnworth SL , Lucendo-Villarin B , et al. Deriving functional hepatocytes from pluripotent stem cells . Curr Protoc Stem Cell Biol . 2014 ;30 :1G.5.1 –12 10 Medine CN , Lucendo-Villarin B , Storck C , et al. Developing high-fidelity hepatotoxicity models from pluripotent stem cells . Stem Cells Transl Med . 2013 ;2 :505 –509 23757504 11 Sjogren AK , Liljevald M , Glinghammar B , et al. Critical differences in toxicity mechanisms in induced pluripotent stem cell-derived hepatocytes, hepatic cell lines and primary hepatocytes . Arch Toxicol . 2014 ;88 :1427 –1437 24912781 12 Schwartz RE , Trehan K , Andrus L , et al. Modeling hepatitis C virus infection using human induced pluripotent stem cells . Proc Natl Acad Sci U S A . 2012 ;109 :2544 –2548 22308485 13 Shlomai A , Schwartz RE , Ramanan V , et al. Modeling host interactions with hepatitis B virus using primary and induced pluripotent stem cell-derived hepatocellular systems . Proc Natl Acad Sci U S A . 2014 ;111 :12193 –12198 25092305 14 Ng S , Schwartz RE , March S , et al. Human iPSC-derived hepatocyte-like cells support Plasmodiumliver-stage infection in vitro . Stem Cell Reports . 2015 ;4 :348 –359 25660406 15 Takebe T , Sekine K , Enomura M , et al. Vascularized and functional human liver from an iPSC-derived organ bud transplant . Nature . 2013 ;499 :481 –484 23823721 16 Baxter M , Withey S , Harrison S , et al. Phenotypic and functional analyses show stem cell-derived hepatocyte-like cells better mimic fetal rather than adult hepatocytes . J Hepatol . 2015 ;62 :581 –589 25457200 17 Godoy P , Schmidt-Heck W , Natarajan K , et al. Gene networks and transcription factor motifs defining the differentiation of stem cells into hepatocyte-like cells . J Hepatol . 2015 ;63 :934 –942 26022688 18 Kern A , Bader A , Pichlmayr R , et al. Drug metabolism in hepatocyte sandwich cultures of rats and humans . Biochem pharmacol . 1997 ;54 :761 –772 9353130 19 Tuschl G , Hrach J , Walter Y , et al. Serum-free collagen sandwich cultures of adult rat hepatocytes maintain liver-like properties long term: a valuable model for in vitro toxicity and drug-drug interaction studies . Chem Biol Interact . 2009 ;181 :124 –137 19482013 20 Schyschka L , Sánchez JJM , Wang Z , et al. Hepatic 3D cultures but not 2D cultures preserve specific transporter activity for acetaminophen-induced hepatotoxicity . Arch Toxicol . 2013 ;87 :1581 –1593 23728527 21 Powers MJ , Janigian DM , Wack KE , et al. Functional behavior of primary rat liver cells in a three-dimensional perfused microarray bioreactor . Tissue Eng . 2002 ;8 :499 –513 12167234 22 Mazzei D , Guazzardi S A low shear stress modular bioreactor for connected cell culture under high flow rates . Bioctechnol Bioeng . 2010 ;106 :127 –137 23 Tilles AW , Baskaran H , Roy P , et al. Effects of oxygenation and flow on the viability and function of rat hepatocytes cocultured in a microchannel flat-plate bioreactor . Biotechnol Bioeng . 2015 ;73 :379 –389 11320508 24 Hoffmann SA , Müller-Vieira U , Biemel K , et al. Analysis of drug metabolism activities in a miniaturized liver cell bioreactor for use in pharmacological studies . Biotechnol Bioeng . 2012 ;109 :3172 –3181 22688505 25 Lübberstedt M , Müller-Vieira U , Biemel KM , et al. Serum-free culture of primary human hepatocytes in a miniaturized hollow-fibre membrane bioreactor for pharmacological in vitro studies . J Tissue Eng Regen Med . 2015 ;9 :1017 –1026 23165723 26 Miki T , Ring A , Gerlach J Hepatic differentiation of human embryonic stem cells is promoted by three-dimensional dynamic perfusion culture conditions . Tissue Eng Part C Methods . 2011 ;17 :557 –568 21210720 27 Stachelscheid H , Wulf-Goldenberg A , Eckert K , et al. Teratoma formation of human embryonic stem cells in three-dimensional perfusion culture bioreactors . J Tissue Eng Regen Med . 2013 ;7 :729 –741 22438087 28 Zeilinger K , Schreiter T , Darnell M , et al. Scaling down of a clinical 3D perfusion multi-compartment hollow fiber liver bioreactor developed for extracorporeal liver support to an analytical scale device useful for hepatic pharmacological in vitro studies . Tissue Eng Part C Methods . 2011 ;17 :549 –556 21210724 29 Yu J , Hu K , Smuga-Otto K , et al. Human induced pluripotent stem cells free of vector and transgene sequences . Science . 2009 ;324 :797 –801 19325077 30 Hay DC , Zhao D , Fletcher J , et al. Efficient differentiation of hepatocytes from human embryonic stem cells exhibiting markers recapitulating liver development in vivo . Stem Cells . 2008 ;26 :894 –902 18238852 31 Hay DC , Fletcher J , Payne C , et al. Highly efficient differentiation of hESCs to functional hepatic endoderm requires Activin A and Wnt3a signaling . Proc Natl Acad Sci U S A . 2008 ;105 :12301 –12306 18719101 32 Pfeiffer E , Kegel V , Zeilinger K , et al. Featured article: isolation, characterization, and cultivation of human hepatocytes and non-parenchymal liver cells . Exp Biol Med (Maywood) . 2015 ;240 :645 –656 25394621 33 Knöspel F , Freyer N , Stecklum M , et al. Periodic harvesting of embryonic stem cells from a hollow-fiber membrane based four-compartment bioreactor . Biotechnol Prog . 2016 ;32 :141 –151 26486457 34 Livak KJ , Schmittgen TD Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method . Methods . 2001 ;25 :402 –408 11846609 35 Scholzen T , Gerdes J The Ki-67 protein: from the known and the unknown . J Cell Physiol . 2000 ;182 :311 –322 10653597 36 Singh M , Berkland C , Detamore MS Strategies and applications for incorporating physical and chemical signal gradients in tissue engineering . Tissue Eng Part B Rev . 2008 ;14 :341 –366 18803499 37 Uzel SG , Amadi OC , Pearl TM , et al. Simultaneous or sequential orthogonal gradient formation in a 3D cell culture microfluidic platform . Small . 2016 ;12 :612 –622 26619365 38 Olmer R , Lange A , Selzer S , et al. Suspension culture of human pluripotent stem cells in controlled, stirred bioreactors . Tissue Eng Part C Methods . 2012 ;18 :772 –784 22519745 39 Storm MP , Orchard CB , Bone HK , et al. Three-dimensional culture systems for the expansion of pluripotent embryonic stem cells . Biotechnol Bioeng . 2010 ;107 :683 –695 20589846 40 Kinney MA , Sargent CY , McDevitt TC The multiparametric effects of hydrodynamic environments on stem cell culture . Tissue Eng Part B Rev . 2011 ;17 :249 –262 21491967 41 Hammond TG , Hammond JM Optimized suspension culture: the rotating-wall vessel . Am J Physiol Renal Physiol 2001 ;281 :F12 –F25 11399642 42 Chen YG , Wang Q , Lin SL , et al. Activin signaling and its role in regulation of cell proliferation, apoptosis, and carcinogenesis . Exp Biol Med (Maywood) . 2006 ;231 :534 –544 16636301 43 Denecker G , Vercammen D , Steemans M , et al. Death receptor-induced apoptotic and necrotic cell death: differential role of caspases and mitochondria . Cell Death Differ . 2001 ;8 :829 –840 11526436 44 Terentiev AA , Moldogazieva NT Alpha-fetoprotein: a renaissance . Tumour Biol . 2013 ;34 :2075 –2091 23765762 45 Cameron K , Tan R , Schmidt-Heck W , et al. Recombinant laminins drive the differentiation and self-organization of hESC-derived hepatocytes . Stem Cell Reports . 2015 ;5 :1250 –1262 26626180 46 Kim JH , Jang YJ , An SY , et al. Enhanced metabolizing activity of human ES cell-derived hepatocytes using a 3D culture system with repeated exposures to xenobiotics . Toxicol Sci . 2015 ;147 :190 –206 26089346 47 Gieseck RL 3rd , Hannan NR , Bort R , et al. Maturation of induced pluripotent stem cell derived hepatocytes by 3D-culture . PLoS One . 2014 ;9 :e86372 24466060 48 Uhlén M , Fagerberg L , Hallström BM , et al. Tissue-based map of the human proteome . Science . 2015 ;347 :1260419 25613900 49 Ek M , Söderdahl T , Küppers-Munther B , et al. Expression of drug metabolizing enzymes in hepatocyte-like cells derived from human embryonic stem cells . Biochem Pharmacol . 2007 ;74 ;496 –503 17568565 50 D'Amour KA , Agulnick AD , Eliazer S , et al. Efficient differentiation of human embryonic stem cells to definitive endoderm . Nat Biotechnol . 2005 ;23 :1534 –1541 16258519 51 Tomizawa M , Shinozaki F , Sugiyama T , et al. Activin A is essential for feeder-free culture of human induced pluripotent stem cells . J Cell Biochem . 2013 ;114 :584 –588 22991093 52 Beattie GM , Lopez AD , Bucay N , et al. Activin A maintains pluripotency of human embryonic stem cells in the absence of feeder layers . Stem Cells . 2005 ;23 :489 –495 15790770 53 Czysz K , Minger S , Thomas N DMSO efficiently down regulates pluripotency genes in human embryonic stem cells during definitive endoderm derivation and increases the proficiency of hepatic differentiation . PLoS One . 2015 ;10 :e0117689 25659159 54 Zhang Y , Bai XF , Huang CX Hepatic stem cells: existence and origin . World J Gastroenterol . 2003 ;9 :201 –204 12532431 55 De Assuncao TM , Sun Y , Jalan-Sakrikar N , et al. Development and characterization of human-induced pluripotent stem cell-derived cholangiocytes . Lab Invest . 2015 ;95 :684 –696 25867762 56 Dianat N , Dubois-Pot-Schneider H , Steichen C , et al. Generation of functional cholangiocyte-like cells from human pluripotent stem cells and HepaRG cells . Hepatology . 2014 ;60 :700 –714 24715669 57 Loh KM , Ang LT , Zhang J , et al. Efficient endoderm induction from human pluripotent stem cells by logically directing signals controlling lineage bifurcations . Cell Stem Cell . 2014 ;14 :237 –252 24412311 58 Lee SB , Seo D , Choi D , et al. Contribution of hepatic lineage stage-specific donor memory to the differential potential of induced mouse pluripotent stem cells . Stem Cells . 2012 ;30 :997 –1007 22378611 59 Vaskova EA , Stekleneva AE , Medvedev SP , et al. “Epigenetic memory” phenomenon in induced pluripotent stem cells . Acta Naturae . 2013 ;5 :15 –21 24455179 Cite this article as: Freyer N, Knöspel F, Strahl N, Amini L, Schrade P, Bachmann S, Damm G, Seehofer D, Jacobs F, Monshouwer M, Zeilinger K (2016) Hepatic differentiation of human induced pluripotent stem cells in a perfused three-dimensional multicompartment bioreactor, BioResearch Open Access 5:1, 235–248, DOI: 10.1089/biores.2016.0027.

PMC005xxxxxx/PMC5003007.txt

==== Front J Endourol Case RepJ Endourol Case RepcrenJournal of Endourology Case Reports2379-9889Mary Ann Liebert, Inc. 140 Huguenot Street, 3rd FloorNew Rochelle, NY 10801USA 2757944510.1089/cren.2016.008610.1089/cren.2016.0086Case ReportIsiris: A Novel Method of Removing Foreign Bodies from the Lower Urinary Tract to Avoid Unnecessary Hospitalization and Anesthesia Smith Peter Mark MBChBHarbias Aman MBChB BSc (Hons) MRCSRobinson Richard MBChB MD FRCS (Urol)Palmer Anne RGN BScGrey Benjamin Robin MBChB MD FRCS (Urol)Department of Urology, Central Manchester Hospitals Foundation Trust, Manchester, United Kingdom.Address correspondence to:Peter Mark Smith, MBChBDepartment of UrologyCentral Manchester Hospitals Foundation TrustOxford RoadManchester M13 9WLUnited Kingdom E-mail:peter.smith@doctors.org.uk01 8 2016 2016 01 8 2016 2 1 144 147 © Peter Mark Smith et al. 2016; Published by Mary Ann Liebert, Inc.2016This Open Access article is distributed under the terms of the Creative Commons License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.Abstract Background: Polyembolokoilamania refers to the practice of inserting foreign bodies (FBs) into natural orifices. A FB within the urethra is a relatively rare phenomenon with 646 cases recorded last year in the United Kingdom. Management of these patients presents technical challenges and complexities because of underlying psychiatric disorders that are often associated. This case illustrates a novel way of removing FBs from the genitourinary tract, requiring less resources, preventing hospital admission, and attempts to break the cycle of behavior, leading to recurrent attendance with polyembolokoilamania. Case Presentation: A 38-year-old Caucasian male prisoner, with psychiatric history presented to the emergency department (ED) with a history of inserting FBs into his urethra on 12 different occasions over a 6-week period. Of these 12 attendances, 3 resulted in admission and 2 required emergency intervention in theater under general anesthesia. After the third attendance in 5 days, it was decided to use Isiris™, a single-use flexible cystoscopy device with a built-in ureteral stent grasper, to remove the FBs and check the integrity of the urethra. The procedure was performed within the ED, without the need for admission to a ward bed or general anesthesia. Furthermore, only two members of staff were required to remove all of the urethral FBs. Conclusion: Isiris, although marketed as a stent removal device, enabled us to remove all the patient's FBs in one procedure. Isiris is an easy to use device, similar to a flexible cystoscope, that a specialist nurse or resident would be familiar using. It allows efficient and safe removal of lower urinary tract FBs, even out of hours. It requires minimal staffing support and can be done in the ED. It has the potential to reduce associated sequela of urethral polyembolokoilamania, saving resources while preserving the availability of the emergency theater. Keywords:  Isirisforeign bodypolyembolokoilamaniagenitourinary ==== Body Introduction Polyembolokoilamania refers to the practice of inserting foreign bodies (FBs) into natural orifices. FBs in the urethra are a relatively rare phenomenon. Of the 5.6 million emergency hospital admissions recorded in England in a 1-year period between April 2014 and March 2015, 646 were because of FBs in the genitourinary (GU) tract with 89 resulting procedures recorded (6 open, 83 endoscopic) for their removal.1 A large number of patients who insert FBs into their urethra may do so on multiple occasions, thus the management of these patients poses challenges of both a psychiatric and technical nature. When theater time and resources are required to remove FBs, the costs to the healthcare provider organization markedly increase. This case illustrates a novel method of removing FBs from the lower GU tract, which has the potential to save resources, prevent hospital admissions, and to break the cycle of behavior that leads to recurrent polyembolokoilamania. Case Report This case describes the management of a 38-year-old Caucasian male prisoner, with a previous psychiatric history of borderline personality disorder and deliberate self-harm, who presented to the emergency department (ED) with a history of inserting multiple FBs into his urethra on 12 different occasions in the space of a 6-week period. Of these 12 attendances to the ED, 3 resulted in admission and 2 required transfer to the emergency operating theater for their removal under general anesthesia. Representative plain radiographs demonstrating the FBs are shown in Figure 1. FIG. 1. A selection of pelvic radiographs demonstrating FBs inserted into the patient's urethra, including parts of pens (B), paperclips (A), glass (B), and dismantled suprapubic catheter (C, D). FBs, foreign bodies. On one occasion, the patient developed acute urinary retention necessitating emergency insertion of a suprapubic catheter (SPC). Subsequently, the patient deliberately dismantled the SPC and reinserted it into the urethra. Thereafter, attendances to the ED became more frequent. After consultation with the prison and psychiatry services, it was decided that, whenever possible, hospital admission and urethral catheterization should be avoided. Further management plan at that stage was for a future elective admission for removal of the retained SPC tip and associated FBs from the bladder under general anesthesia. In the subsequent week, the patient presented to the ED three times after insertion of further FBs. On each attendance, the FBs were easily palpable in the urethra and their removal was effective using forceps. However, on each occasion despite there being no other palpable FBs in the urethra, the patient was adamant that he was unable to void, until a catheter was inserted. On his next attendance, the third in 5 days, it was decided to use Isiris™, a single-use flexible cystoscopy device with built-in stent grasper (Fig. 2), to remove the FBs and check the integrity of the urethra. The procedure was performed quickly within the ED, thus avoiding financial penalties for delay in discharge from the ED to a place of residence or to a hospital bed. The procedure only required intraurethral lubricant gel and two members of staff and it was well tolerated by the patient, only requiring basic oral analgesia. All remaining FBs that had been inserted over the previous 6 weeks were removed in this minimally invasive manner (See Figures 3 and 4). The procedure allowed a confident endoscopic examination to reassure the patient that no significant urethral trauma had been sustained, that there were no residual FBs, and that ultimately no reason for him not to be able to void. The use of the Isiris system permitted a safe and cost-efficient intervention with discharge directly from the ED. FIG. 2. Showing the Isiris™ with monitor. FIG. 3. Still shots taken from the Isiris monitor while removing the FBs. A, C demonstrate how the stent grasping forceps are opened and B, D show how they are used to grasp and subsequently remove the FB. FIG. 4. Various objects removed from the patient's urethra upon use of Isiris. Pictured are various parts of a dismantled pen and the tip of a dismantled catheter. Discussion Because of the rare nature of these cases and the added psychiatric component, a holistic approach to the management of polyembolokoilamania is required. A multidisciplinary approach was adopted with consultation of all key stakeholders (the patient, urology, psychiatry, ED, and prisoner services), resulting in a safe and efficient intervention. This minimized both reinforcement of the patient's behavior and the impact upon the organization's capacity and resources. Such a response was required because of the spiraling situation, culminating in three attendances in 5 days with the patient becoming more demanding and manipulative so as to remain an inpatient in hospital. Although marketed as a stent removal device, Isiris enabled the removal of all the patient's FBs in one procedure in the ED. As Isiris is an all-in-one single-use system, there are multiple advantages of using it over rigid or flexible cystoscopy in such cases. First, the equipment can be kept in a 24-hour access storeroom and can be easily transported by a single person; it was taken from our storeroom to the patient and the procedure was performed in its entirety in the ED at the patient's bedside. This negated the need for admission of the patient and subsequent transfer to the emergency operating theater suite for conventional flexible or rigid cystoscopy. A single-use Isiris cystoscope plus monitor costs £258 ($333). Table 1 illustrates a comparative cost breakdown for FB removal from the GU tract in both the ED and the emergency operating theater for organizations such as the UK's National Health Service (NHS). As Isiris is an all-in-one system that can be operated by a single individual, the cost savings are significant compared with FB removal in the operating theater with the need for an anesthetist, theater nurse, surgeon, assistant, operating department practitioner, and recovery staff; approximated at £1140 ($1470). Furthermore, as Isiris is a single-use system, the cost of cleaning the theater, sterilizing equipment, and disinfecting cystoscopes is also avoided. Table 1. Illustrating the Costs Involved in Different Stages of Removal of a Foreign Body from the Genitourinary Tract   Admission ED using Isiris™ ED attendance £132 ($170) £132 ($170) 24 hours on a ward £303 ($391) — 1 hour emergency theater £1000 ($1291) — Equipment cost £95 ($122) £258 ($333) Total £1530 ($1975) £390 ($503) Costing's are based upon U.K. NHS practice.2,3 ED = emergency department; NHS = National Health Service. Cost is not the only advantage of being able to remove FBs in the ED. There are a number of patient factors that would make immediate endoscopic removal of FBs in the ED favorable, in particular the ability to avoid a general anesthetic. The procedure itself is well tolerated and markedly reduces time in hospital, allowing a quicker return to normal activities. Furthermore, endoscopic removal of FBs using Isiris can be performed by a wide spectrum of clinicians who are competent to perform flexible cystoscopies, including more junior medical staff. This further reduces waiting times for patients. Removing the FBs immediately upon presentation may also offer the potential benefit of avoiding the complications associated with urethral polyembolokoilamania such as strictures, retention, periurethral abscess, and fistulation. There are many reasons people insert objects into natural orifices and there is a vast amount of literature surrounding polyembolokoilamania and the psychologic nature of the disease. Malingering refers to the act of deliberately feigning physical or psychologic symptoms to achieve a tangible secondary gain. In this case, the patient had a substantial and lengthy history of nonsuicidal, self-injurious malingering behavior. The phenomenon of urethral polyembolokoilamania by prisoners in a maximum-security hospital to control prison and hospital staff has been reported before and is difficult to manage.4 By avoiding admission and dealing with the problem within the ED, there is the potential, along with the help of psychiatry/psychology, to begin to break these patterns of behavior and relieve some of the economic burden posed by malingering patients. Conclusions Although this scenario is admittedly rare, it provides a novel way of dealing with urethral polyembolokoilamania. Isiris is an easy-to-use device with which a urology specialist nurse or urology resident would be familiar. We have reported its utility for efficiently and safely removing urethral FBs, even out of hours. It requires minimal staffing support and can be done in the ED itself, thereby providing better and more expedient patient care. It has the potential to reduce the likelihood of complications associated with urethral polyembolokoilamania, while also saving resources in addition to preserving the availability of the operating theater for other surgical emergencies. Disclosure Statement B.R. Grey has been invited teaching faculty on ureteroscopy courses sponsored by Coloplast over the last three years, and spoke at a Coloplast symposium as part of the British Association of Urological Surgeons annual meeting in June 2016. No honorarium was received for these events, but expenses were paid (travel and/or accommodation). Recently (June 2016), Mr. Grey has been invited to be a key opinion leader for Coloplast regarding Isiris, though had not accepted the role at the time of acceptance and no payment is expected. Coloplast has offered financial support of travel and accommodation expenses only as part of both his overnight attendance at the British Association of Urological Surgeons (BAUS) annual Section of Endourology meeting in September 2016. P.M. Smith is part of the Coloplast “Young Lion” program, where they identify future urologists and provide financial support in their attendance to urological educational events. Coloplast has provided accommodation expenses only in attendance to the BAUS annual meeting in June 2016. For all other authors, no competing financial interests exist. Abbreviations Used FBforeign body EDemergency department SPCsuprapubic catheter NHSNational Health Service ==== Refs References 1 Hospital Episode Statistics, Admitted Patient Care—England . Health and Social Care Information Centre . Available at www.hscic.gov.uk/article/2021/Website-Search?q=Hospital+Episode+Statistics%2c+Admitted+Patient+Care%2c+England&topics=13205&sort=Relevance&size=10&page=1&area=both-top (Accessed 7 2 , 2016 ) 2 Department of Health . Reference Costs 2014–15 . Available at www.gov.uk/government/uploads/system/uploads/attachment_data/file/477919/2014-15_Reference_costs_publication.pdf (Accessed 7 3 , 2016 ) 3 ISD Scotland – Theatre Costs 2014–15 : Available at www.isdscotland.org/Health-Topics/Finance/Publications/2015-11 (Accessed 7 3 , 2016 ) 4 Rada RT , James W Urethral insertion of foreign bodies: A report of contagious self-mutilation in a maximum-security hospital . Arch Gen Psychiatry 1982 ;39 :423 –429 7065851 Cite this article as: Smith PM, Harbias A, Robinson R, Palmer A, Grey BR (2016) Isiris: a novel method of removing foreign bodies from the lower urinary tract to avoid unnecessary hospitalization and anesthesia, Journal of Endourology Case Reports 2:1, 144–147, DOI: 10.1089/cren.2016.0086.

PMC005xxxxxx/PMC5003008.txt

==== Front Front PsycholFront PsycholFront. Psychol.Frontiers in Psychology1664-1078Frontiers Media S.A. 10.3389/fpsyg.2016.01219PsychologyCorrectionCorrigendum: Gender Differences in the Physical and Psychological Manifestation of Childhood Trauma and/or Adversity in People with Psychosis Sweeney Shaun 1*Air Tracy 1Zannettino Lana 2Shah Sonal S. 3Galletly Cherrie 11Discipline of Psychiatry, University of AdelaideAdelaide, SA, Australia2School of Nursing and Midwifery, Flinders UniversityAdelaide, SA, Australia3Neuropsychiatric Epidemiology Research Unit, School of Psychiatry and Clinical Neurosciences Laboratory, University of Western AustraliaPerth, WA, AustraliaEdited by: Christina Andreou, Universitäre Psychiatrische Kliniken Basel, Switzerland Reviewed by: Michelle Dow Keawphalouk, Harvard University/Massachusetts Institute of Technology, USA *Correspondence: Shaun Sweeney shaun.sweeney@adelaide.edu.auThis article was submitted to Psychology for Clinical Settings, a section of the journal Frontiers in Psychology 29 8 2016 2016 7 121928 5 2016 02 8 2016 Copyright © 2016 Sweeney, Air, Zannettino, Shah and Galletly.2016Sweeney, Air, Zannettino, Shah and GalletlyThis is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.A corrigendum onGender Differences in the Physical and Psychological Manifestation of Childhood Trauma and/or Adversity in People with Psychosis by Sweeney, S., Air, T., Zannettino, L., and Galletly, C. (2015). Front. Psychol. 6:1768. doi: 10.3389/fpsyg.2015.01768childhood trauma and/or adversitypsychosisself-reported health statusgenderoutcomes ==== Body Due to an oversight by the authors, the co-author Sonal S. Shah was missed. This has been now added in the author list above and the Author Contributions statement has been revised below. In the section Materials and Methods, sub-section CHILDHOOD TRAUMA AND/OR ADVERSITY, the whole paragraph should be replaced with: Questions about the occurrence and nature of CTA were included in the SHIP interview. CTA was coded on the basis of responses to the question: “Were there any other very distressing or traumatic events in your childhood (not including parental separation or divorce, or loss of a close relative)?” Finally, Health (2012) and Bromfield and Holzer (2008) should be removed from the text and from the Reference List. The authors apologize for these errors. These corrections do not affect the data or the conclusions contained in the manuscript. The original article has been updated. Author contributions SS was primarily responsible for the conception of the research upon which this paper is based. CG, LZ, and TA assisted in the development of the ideas for the paper and in refining the final manuscript. SS produced the first draft of the manuscript and submitted the final manuscript and completed the literature search and wrote the main body of the manuscript. TA analyzed the data. SSS contributed to acquisition, analysis and interpretation of the national SHIP data; was part of initial discussions on the substance of this paper, and critically contributed to the draft of the revised version. All authors approved the final 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.

PMC005xxxxxx/PMC5003011.txt

==== Front Biores Open AccessBiores Open AccessbioresBioResearch Open Access2164-78442164-7860Mary Ann Liebert, Inc. 140 Huguenot Street, 3rd FloorNew Rochelle, NY 10801USA 10.1089/biores.2016.002610.1089/biores.2016.0026Original Research ArticleFunctional Effects of Delivering Human Mesenchymal Stem Cell-Seeded Biological Sutures to an Infarcted Heart Hansen Katrina J. 1Favreau John T. 1Guyette Jacques P. 1,2Tao Ze-Wei 1,3Coffin Spencer T. 1Cunha-Gavidia Anny 1D'Amore Brian 1Perreault Luke R. 1Fitzpatrick John P. 1DeMartino Angelica 1Gaudette Glenn R. 1,*1 Department of Biomedical Engineering, Worcester Polytechnic Institute, Worcester, Massachusetts.2 Center for Regenerative Medicine, Massachusetts General Hospital, Boston, Massachusetts.3 Department of Pharmaceutical Sciences, University of Arkansas for Medical Sciences, Little Rock, Arkansas.* Address correspondence to: Glenn R. Gaudette, PhD, Department of Biomedical Engineering, Worcester Polytechnic Institute, 60 Prescott Street, Worcester, MA 01605, E-mail: gaudette@wpi.edu01 8 2016 2016 01 8 2016 5 1 249 260 © Katrina J. Hansen et al. 2016; Published by Mary Ann Liebert, Inc.2016This Open Access article is distributed under the terms of the Creative Commons License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.Abstract Stem cell therapy has the potential to improve cardiac function after myocardial infarction (MI); however, existing methods to deliver cells to the myocardium, including intramyocardial injection, suffer from low engraftment rates. In this study, we used a rat model of acute MI to assess the effects of human mesenchymal stem cell (hMSC)-seeded fibrin biological sutures on cardiac function at 1 week after implant. Biological sutures were seeded with quantum dot (Qdot)-loaded hMSCs for 24 h before implantation. At 1 week postinfarct, the heart was imaged to assess mechanical function in the infarct region. Regional parameters assessed were regional stroke work (RSW) and systolic area of contraction (SAC) and global parameters derived from the pressure waveform. MI (n = 6) significantly decreased RSW (0.026 ± 0.011) and SAC (0.022 ± 0.015) when compared with sham operation (RSW: 0.141 ± 0.009; SAC: 0.166 ± 0.005, n = 6) (p < 0.05). The delivery of unseeded biological sutures to the infarcted hearts did not change regional mechanical function compared with the infarcted hearts (RSW: 0.032 ± 0.004, SAC: 0.037 ± 0.008, n = 6). The delivery of hMSC-seeded sutures exerted a trend toward increase of regional mechanical function compared with the infarcted heart (RSW: 0.057 ± 0.011; SAC: 0.051 ± 0.014, n = 6). Global function showed no significant differences between any group (p > 0.05); however, there was a trend toward improved function with the addition of either unseeded or seeded biological suture. Histology demonstrated that Qdot-loaded hMSCs remained present in the infarcted myocardium after 1 week. Analysis of serial sections of Masson's trichrome staining revealed that the greatest infarct size was in the infarct group (7.0% ± 2.2%), where unseeded (3.8% ± 0.6%) and hMSC-seeded (3.7% ± 0.8%) suture groups maintained similar infarct sizes. Furthermore, the remaining suture area was significantly decreased in the unseeded group compared with that in the hMSC-seeded group (p < 0.05). This study demonstrated that hMSC-seeded biological sutures are a method to deliver cells to the infarcted myocardium and have treatment potential. Keywords:  biomaterialscardiologystem cellstissue engineering ==== Body Introduction Cardiovascular diseases account for more than 30% of deaths worldwide with myocardial infarction (MI) contributing to about half of these deaths.1 Although many advances have been made in revascularization therapy, these therapies do not repair infarcted myocardium.2,3 Cell-based therapies have been proposed as a treatment for MI. Importantly for the clinical use of these cells, Hare et al. demonstrated no deleterious effects of human mesenchymal stem cells (hMSCs) when delivered intravenously to patients.4 Several clinical trials delivering hMSCs have demonstrated promising initial results with short-term improvements in ejection fraction; longer term studies have shown that functional benefits dissipate within 1 year.5 The ability of mesenchymal stem cells (MSCs) to differentiate into contractile myocytes that can couple with the native heart has been debated.6,7 However, it appears that this cell type is not able to differentiate into a significant quantity of cardiomyocytes that leads to improved, active contractile function in the myocardium.8 Recent studies have suggested that cardioprotective paracrine signaling is largely responsible for the immunomodulatory capacity of MSCs and their ability to improve function after delivery to infarcted tissue.9,10 One limitation of MSC therapy is the inefficient engraftment rate using current cell delivery methods.11 The most efficient delivery methods are intramyocardial injection, resulting in ∼10% cell retention,11 and transendocardial delivery, resulting in ∼19% cell retention.12 This small 10–20% retention rate would require a large starting cell population to replace the billion myocytes lost due to an MI.13 In addition, targeting cell delivery to a specific region is challenging. To overcome these issues, we developed fibrin-based biological sutures that can be seeded with hMSCs.14 Using this delivery method, previous studies have shown a 64% engraftment rate of hMSCs to cardiac tissue.15 Herein, we use this delivery method to deliver hMSCs to infarcted myocardium and determine whether these cells lead to improved regional mechanical function. Materials and Methods Fibrin microthread production Discrete fibrin microthreads were produced as described previously.14 In brief, fibrinogen (70 mg/mL; MP Biomedical) and thrombin (8 U/mL; Sigma) were coextruded in a 10 mM HEPES (Calbiochem)-buffered bath (pH 7.4) using a custom-built system. After extrusion, the microthreads were allowed to polymerize before being removed and air dried. To create biological sutures, 12 individual microthreads were hydrated with distilled water and twisted together. After drying, each bundle was cut to 4 cm lengths threaded through the eye of a surgical needle (size 26; Havel's, Inc.), hydrated, and twisted together to form a 2 cm long suture. Each suture was placed in a 4 cm long piece of silastic tubing (1.98 mm inner diameter, Dow Corning) with a 27-gauge 0.5 inch needle (Becton Dickinson and Co.) and secured with a slide clamp across the needles. Suture constructs were ethylene oxide sterilized and stored in a desiccator before cell seeding. Human MSC culture Passages 4–8 hMSCs (Lonza, Inc.) were grown in monolayer culture in mesenchymal stem cell growth medium (MSCGM; Lonza, Inc.) incubated at 37°C in 5% CO2 until cells reached 80–90% confluence. All implanted hMSCs were treated with MSCGM supplemented with 8.2 nM 655 ITK carboxyl quantum dots (Qdots; Invitrogen) to allow for cell tracking. Qdots were added to 70–80% confluent layers of hMSCs for 24 h, after which Qdot medium was replaced and cells were maintained until further use. Qdot loading was verified using fluorescent microscopy. Suture seeding Before cell seeding, sutures were hydrated with dPBS for 20 min; dPBS is expelled before cells were introduced. For cell seeding, 100 μL of a 2.5 × 105, 5 × 105, and 1 × 106 hMSC/mL solution was drawn into a 1 mL syringe (Becton Dickinson) attached to the 27-gauge needle and dispensed into the tubing containing the suture. The 27-gauge needle and syringe were removed and a slide clamp was added to seal both ends. The seeding construct was placed in a vented 50 mL conical tube and placed in a MACsmix rotator (Miltenyi Biotech) set to four rotations per minute and placed into a 37°C incubator with 5% CO2. To determine the role of incubation time, sutures were incubated for 1, 3, 6, 12, and 24 h. All implanted sutures were seeded with 100 μL of a 1 × 106 hMSC/mL solution for 24 h. hMSC attachment quantification After seeding, hMSC-seeded sutures not destined for implantation were either fixed in 4% paraformaldehyde for staining or prepped for a CyQuant DNA assay. For CyQuant DNA assays, sutures were placed in a 1.5 mL Eppendorf tube with 100 μL of phosphate-buffered saline (PBS) in a −80°C freezer for cell lysis. The CyQuant assay was run according to manufacturer's specifications. Fluorescent measurements were taken using a plate reader (Victor 3; Perkin Elmer) at an excitation of 480 nm and absorption of 520 nm. Values are reported as mean ± standard error of the mean (SEM) cell count per linear centimeter of suture. Infarct model Male nude rats (NIH nude; Taconic) were used for all procedures. Animals were anesthetized using 5% inhaled isoflurane in oxygen, intubated, and mechanically ventilated with anesthesia maintained with 3% isoflurane. A 2 cm incision along the fourth intercostal space and subsequent blunt dissection of the muscle was used to expose the heart. To induce an infarct, the left anterior descending (LAD) artery was ligated for 1 h using 5-0 Prolene (Ethicon) suture. After 1 h, the ligature was cut to restore blood flow and the chest was either closed or the suture was implanted and the chest was closed. The suture was implanted from the base of the heart toward the apex of the heart, and excess suture was cut and removed (Fig. 1). Sham operations underwent the same procedure already described without the ligation of the LAD and suture implantation. Animals were allowed to recover for 1 week. FIG. 1. Implanted hMSC-seeded suture. Male athymic rat heart was exposed and the LAD was ligated to induce infarct. One hour postligation, ligature was cut and blood flow was resumed. The hMSC-seeded fibrin suture was implanted from the base to the apex such that suture entered in the ischemic zone (i), passed through the infarcted myocardium (ii), and exited through the ischemic zone at the apex (iii) (A). Implantation of fibrin suture in rat heart (B). Excess suture was removed. hMSC, human mesenchymal stem cell; LAD, left anterior descending. Terminal surgery One week postoperation, animals were anesthetized, intubated, and placed on a respirator as already described. The heart was exposed through the original suture lines and retractors were placed into the intercostal space to expose the left ventricle of the heart. A 3.5 French size micro tip catheter (Millar) was inserted into the left ventricle through the apex and sutured in place. A high speed camera (Fastec HiSpec4; Fastec, Inc.) was used to record high speed videos (250 frames/sec, 8 bit depth, 1696 × 1710) of contracting ventricular muscle. Pressure data were synchronized to video frame acquisition through a data acquisition board (National Instruments) and a LabVIEW-based controller. Videos were acquired from multiple angles and video and pressure data were saved for offline data analysis. After video acquisition, a 0.3 mg/kg dose of Beuthanasia-D (VetOne) was injected into the left atrium to arrest the heart, and heart tissue was removed for histological processing and analysis. Assessment of global and regional mechanical function Global mechanical function was assessed using the collected pressure waveforms as described previously.16 Maximum developed pressure was determined by averaging the difference between the maximum and minimum pressures for each beat. Minimum rate of pressure decline and maximum rate of pressure development were determined by applying a five-point linear averaging filter to the pressure wave form and finding the maximum and minimum of the discrete derivative of the signal for each beat. Finally, the diastolic relaxation time constant was determined by applying a three-parameter least squared fitting method to pressure data from its negative inflection point to its minimum for each beat.17 All global function data are averaged over at least five beats per animal. High density mapping (HDM) was used to assess regional mechanical function as previously described.15,16,18 HDM is a speckle tracking algorithm based on Fourier domain phase correlation with the subpixel registration algorithm described by Foroosh et al.19 Using this method, a region of interest was defined on the rat heart over the thread and/or infarct region. The region was subsequently divided into small (32 × 32 pixel) windows. Individual windows were tracked from frame to frame across several cardiac cycles to yield a displacement field. Displacement fields were used to quantify systolic area of contraction (SAC) and regional stroke work (RSW), as described previously.18 Histology Sutures not prepared for CyQuant were fixed in 4% paraformaldehyde for 10 min and stained for F-actin. In brief, sutures were washed in PBS, blocked for 30 min in 1% albumin from bovine serum, and stained with 488 phalloidin (Invitrogen) for 30 min and counterstained with Hoechst 33342 (0.5 μg/mL; Invitrogen) for 5 min at room temperature. After terminal surgery, hearts were removed and cut into half bisecting the infarct and biological suture, and fixed in 4% paraformaldehyde overnight. After fixation, heart sections were moved to a 30% sucrose solution for 24 h. Each half was embedded using OCT and placed in a Leica CM 3050 cryostat (Leica Microsystems) and cut into 10 μm thick sections onto charged glass microscope slides (Globe Scientific, Inc.). Sections were stained with Masson's trichrome reagents according to manufacturer's instruction. Adjacent sections were used for immunofluorescence staining. In brief, sections were fixed in acetone at −20°C for 10 min and blocked with 5% goat serum for 45 min. The sections were incubated in mouse anti-α-actinin monoclonal antibody (1:100; Abcam) overnight at 4°C. Subsequently, sections were treated in goat antimouse secondary antibody (1:400; Alexa Fluor 488; Thermo Fisher Scientific) 1 h at room temperature and counterstained with Hoechst 33342 (0.5 μg/mL) for 5 min at room temperature. Images were acquired with a Leica Upright DMLB2 or a Leica TCP SP5 confocal laser scanning microscope. Infarct region and suture area quantification Microsoft Image Composite Editor was used to generate composite images of the full cross sections from sections stained with Masson's trichrome. Images were uploaded into a custom MATLAB analysis program (Mathworks) that evaluates color of individual pixels in the image and extrapolates the location of the lumen and outer boundary of the section. Image regions outside the boundary of the heart and within the lumen (white space) were removed, leaving the image region corresponding to intact heart tissue. As Masson's trichrome stains cardiac muscle tissue red and fibrosis blue, blue regions of the image were programmatically identified and used to determine infarct size. The program approximates and outputs total section area and total area of the infarct. Sectional and infarct area was calculated by the sum of squared pixel lengths and converted to microns to determine the percentage of the section that is infarcted. For suture area measurement, images with Masson's trichrome staining were uploaded to ImageJ (NIH). Manual tracings around the suture in each cross section were obtained to give the area of the suture. Every section containing a suture was used for the measurements to give an average area. Statistics Results are represented as mean ± SEM. Comparisons between seeding conditions were analyzed using a one-way analysis of variance, with a post hoc Tukey test for multiple comparisons. Comparisons between in vivo treatment conditions were analyzed using a Kruskal–Wallis test with a Dunn's test for multiple comparisons. An unpaired t test was used to determine significance between suture areas. All conditions were considered significant at a value of p < 0.05. Results Concentration and duration of cell seeding affect cell attachment to biological sutures Three concentrations of hMSCs (2.5 × 105, 5 × 105, and 1 × 106 hMSCs/mL solution) were seeded onto 2 cm long biological sutures. There were significantly more hMSCs attached to the sutures seeded with the higher concentrations (5 × 105 and 1 × 106 hMSC/mL) than a concentration of 2.5 × 105 (n = 10; p < 0.05) (Fig. 2A), a finding that was further supported by F-actin staining (Fig. 2B–D). FIG. 2. Effect of cell seeding concentration on hMSC attachment to biological suture. Biological sutures were seeded with three different concentrations of hMSCs: 25, 50, and 100 k cells per suture. Highest attachment was found for the higher concentrations (A) with no statistical difference found between the 50 and 100 k. F-actin staining for 25 k (B), 50 k (C), and 100 k (D) revealed similar cell attachment for the 50 and 100 k conditions. Some blue autofluorescence seen in (B) and (D) is due to the suture. Mean ± SEM, p < 0.05, n = 10 per group. SEM, standard error of the mean. To determine the effect of incubation time, five different times were examined (1, 3, 6, 12, and 24 h). Seeding for 1 h (1216 ± 192, n = 10) and for 3 h (7052 ± 1085, n = 10) significantly reduced cell attachment compared with seeding for 12 h (21,512 ± 3526, n = 10) and 24 h (17,887 ± 2746, n = 10) (p < 0.05) (Fig. 3A). There was no significant difference in the number of hMSCs attached at 6 (15,224 ± 2213, n = 10), 12, or 24 h. F-actin staining shows cell attachment beginning at 1 h (Fig. 3B), more cells attached at 6 and 12 h (Fig. 3C, D), with cells appearing to elongate along the length of the suture at 12 and 24 h. FIG. 3. Effect of seeding time on hMSC attachment to biological suture. hMSC attachment for five different time points (1, 3, 6, 12, and 24 h) was examined. Seeding times of 1 and 3 h resulted in significantly fewer cells attached compared with seeding times 12 and 24 h, where there was no significant differences in attachment at 6, 12, and 24 h (A). F-actin staining shows cells attached at 1 h (B), but with higher number of cells attached at 6 and 12 h (C, D). Mean ± SEM, p < 0.05, n = 10 per group. hMSC-seeded biological sutures do not improve global function after infarction After the model was successfully developed, five sham, two infarct, one unseeded, and three animals from the hMSC-seeded group died before successful completion of all data acquisition. These animals were not included in this study. No differences were noted in the maximum or minimum rate of pressure development or end diastolic pressures between groups (sham, MI, MI with unseeded sutures, and MI with hMSC-seeded sutures, n = 6 per group; Fig. 4A–C). There were no significant differences in the heart rates in all groups at the time of terminal surgery (Fig. 4D). Similarly, no significant differences were seen in diastolic relaxation time constant, although sham-operated hearts had the shortest relaxation time (8.6 ± 1.1 msec) and infarcted hearts had the longest relaxation time (14.1 ± 4.3 msec) (Fig. 4E). FIG. 4. Global function is not impacted by hMSC-seeded biological suture treatment after infarct. No significant changes were noted for any of the following measurements: minimum dP/dt (A), maximum dP/dt (B), end diastolic pressure (C), heart rate (D), or diastolic time relaxation constant (E). Mean ± SEM. n = 6 per group. hMSC-seeded biological sutures improve regional mechanical function after infarction To assess regional mechanical function, a speckle tracking algorithm, HDM, was used for determination of RSW and SAC. The average region in which function was determined was similar in all groups. Creation of an MI resulted in a significant decrease in RSW (0.026 ± 0.011) and SAC (0.022 ± 0.015) compared with sham-operated animals (RSW 0.141 ± 0.009; SAC 0.166 ± 0.005) (n = 6, p = 0.0027 RSW and SAC) (Fig. 5A, B). Delivery of unseeded sutures to the infarcted regions also resulted in statistically reduced RSW (0.037 ± 0.005) and SAC (0.039 ± 0.009) compared with sham-operated animals (n = 6, p = 0.012 RSW; p = 0.026 SAC). hMSC-seeded sutures had a lower RSW (0.057 ± 0.011) and SAC (0.051 ± 0.014) than sham-operated hearts, although statistical significance was not achieved (n = 6, p = 0.165 RSW; p = 0.086 SAC). There were no significant differences in regional mechanical function between infarct, unseeded, and seeded groups; however, infarcted hearts treated with cell-seeded sutures showed a trend toward increase of RSW when compared with infarcted hearts (p = 0.07). Representative pressure area work loops (Fig. 5C) indicate improved function in hMSC-seeded group over unseeded and infarcted groups. FIG. 5. hMSC-seeded biological sutures improve regional mechanical function after infarct. RSW and SAC were significantly decreased for infarct and unseeded suture groups compared with the sham group (A, B). Representative pressure area work loop (C) shows improved function for seeded group compared with infarct and unseeded groups. Arrows indicate direction of all work loops over time. Mean ± SEM, p < 0.05, n = 6 per group. RSW, regional stroke work; SAC, systolic area of contraction. hMSCs delivered on biological sutures remain in heart at least 1 week after delivery Numerous Qdot-loaded cells were detected in the seeded group. Cells were found in regions close to the biological sutures (Fig. 6), but not on the suture, suggesting migration into surrounding tissue. When stained to observe α-actinin, no Qdot-loaded cells observed exhibited positive α-actinin expression. FIG. 6. hMSCs persist in myocardium 1 week after delivery. Masson's trichrome-stained cross sections show successful delivery of biological suture [pink; black arrows in (A), white lines in (B) and (C)] to myocardium. The black boxes (B, C) show the location of serial sections stained for α-actinin and Hoechst relative to (A). The arrows in (B) and (C) indicate quantum dot-positive hMSCs that have migrated off the suture into fibrotic tissue. The cell identified by the yellow arrow in (C) is shown in the inset. Effects of implantation of unseeded and hMSC-seeded sutures on infarct size Serial sections of Masson's trichrome-stained sections were analyzed to determine infarct size, denoted as a percentage of the cross section. No infarcted tissue was detected by Masson's trichrome staining in sham-operated hearts, whereas the MI, unseeded suture, and seeded suture groups all appeared to have infarct regions (Fig. 7A). Infarcted hearts without suture delivery had the greatest infarct size of 7.0% ± 2.2% (Fig. 7B). Unseeded and hMSC-seeded groups had comparable infarct sizes of 3.8% ± 0.6% and 3.7% ± 0.8%, respectively, but were not significantly different from MI alone (p > 0.05, Fig. 7B). Due to the automated method used to calculate infarct size, the sham group had a small infarct size of 1.6% ± 0.1% (likely due to inherent collagen structure in the healthy heart). FIG. 7. Infarct size and suture area. Masson's trichrome staining shows infarcted area in blue and healthy muscle in red (A). Quantification of infarct size from serial trichrome-stained sections indicates an increase in infarct size for infarct-only group with a similar reduction in infarct size shown for seeded and unseeded groups (n = 6) (B). Quantification of persisting suture area after 1 week shows significantly smaller suture area for unseeded sutures (C). Mean ± SEM, *p < 0.05. hMSCs preserve suture area in infarct Serial sections of Masson's trichrome-stained sections were analyzed for suture area. An average suture area of 0.016 ± 0.004 mm2 and 0.033 ± 0.007 mm2 was found for unseeded and seeded groups, respectively (Fig. 7C). The group with hMSCs had a significantly greater suture area (p < 0.05), indicating that the hMSCs preserved suture area at 1 week after delivery in the infarcted heart. Discussion Cell therapy continues to hold tremendous potential to treat millions of patients living with MIs and heart failure. However, the full benefits of cell therapy have yet to be realized. The past decade yielded dozens of clinical trials aimed at improving cardiac output by examining different cell types, including bone marrow mononuclear cells,20 MSCs,4,21 and cardiosphere-derived cells,22 effect of timing on cell delivery,23–25 chronic26,27 and acute28 MI, and cell delivery methods: intracoronary,22,28 intravenous,4 intramyocardial,29 and transendocardial.12,30 Despite numerous trials investigating extensive variables, these clinical trials demonstrated limited improvements in ejection fraction without restoring cardiovascular function to baseline values. All studies failed to efficiently deliver the chosen cell type to infarcted tissue, further underscoring the value of investigating new delivery methods that improve cell retention. Current cell delivery methods to the heart exhibit retention rates of only 10–20%, which are inefficient for delivering clinically relevant number of cells.11–13 Previously, we demonstrated cell-seeded biological sutures capable of efficiently delivering cells to a targeted area (64% retention).15 In addition, we reported that delivering unseeded sutures to a healthy heart demonstrated a small fibrotic reaction that was significantly reduced with hMSC-seeded sutures.16 In terms of regional function, unseeded sutures decreased SAC, where hMSC-seeded sutures dampened the decrease in SAC.16 Similar values for RSW and SAC were reported for sham animals in the previous study and in our current study. In this study, we investigated the functional effects of delivering hMSCs through a biological suture to an infarcted heart. When combined with our method for evaluating regional mechanical function with high spatial resolution, we are uniquely able to determine changes in regional mechanical function that occur directly in the region of cell delivery. This combination allows for targeted cell delivery followed by targeted evaluation of regional mechanical function. In this study, we demonstrated that varying cell seeding concentrations and seeding times affect the quantity of hMSCs seeded onto the suture. Ideal seeding conditions were found to occur using 100,000 cells per suture seeded for 12–24 h. There was only a small increase in cell attachment between the 50,000 and 100,000 seeded groups, thus we do not believe that further increasing the number of cells seeded would lead to improved cell attachment as the surface area available for cell attachment has been maximized. Due to time considerations for animal studies, sutures seeded for 24 h with 100,000 cells were used (allowing for seeding 24 h before implantation). No changes between conditions were found for global parameters including pressure changes, diastolic relaxation constant, or heart rate. The MI created in the rat may not have been large enough to result in a detectable change in global function. In addition, rat ventricular remodeling after an MI can take 4–8 weeks to manifest such that changes on a global level may not be seen at 1 week.31 This finding further necessitates the use of HDM, which images the region of damage and cell delivery to determine on a regional level whether improvements in mechanical function were found that may have been negated at a global level. Using HDM to analyze myocardial function in infarcted rat hearts we demonstrated decreased regional mechanical function in animals with an induced MI, using metrics of RSW and SAC. Delivering unseeded sutures demonstrated improvements in regional mechanical function over the MI group; the hMSC-seeded sutures demonstrated slightly higher function, however, not statistically significant. This suggests that the biological suture alone has the ability to improve regional mechanical function in the infarct zone. Studies have demonstrated that acellular biomaterial delivery to an infarct can improve mechanical function by stabilizing the infarct, thickening the left ventricle, and decreasing infarct expansion.32–34 As both the unseeded and hMSC-seeded suture groups had similar infarct sizes, the significantly increased suture area for the hMSC-seeded group may explain the improvement in regional mechanical function for the hMSC-seeded group compared with the unseeded group. A similar trend of increased suture area for hMSC-seeded sutures was also found in studies in a healthy heart, suggesting the ability for the hMSCs to decrease in vivo degradation time of biological sutures compared with unseeded sutures.16 It has been established that MSCs act transiently and die off within days to weeks of being delivered to the infarct.8 As such, we chose not to examine cell retention as hMSC retention may be altered by the infarct environment at 1 week independent of the method used for delivery. However, hMSCs were found to persist in the myocardium for 1 week and were found in regions away from the suture, suggesting their propensity to migrate from the suture into the damaged tissue. Delivered cells were tracked using Qdots, which have previously been shown to be taken up by hMSCs and remain in the cells for several passages in vitro.35 Despite the emergence of many clinical trials focused on the delivery of MSCs to injured cardiac tissue, none of these trials have demonstrated significant findings to suggest the robust ability for MSCs to improve contractile cardiac function. Many of these studies used delivery methods with low delivery efficiencies. In this study, we describe findings using a biological suture-based method to deliver cells to infarcted tissue. Although we did not see statistically improved function when seeded sutures were delivered over infarct alone, we do remain optimistic about the use of the biological suture for cell delivery. Although hMSCs may not be the most appropriate cell to use in regeneration of contractile cardiomyocytes, they may still have beneficial effects in the infarcted heart as studies have demonstrated their immunomodulatory effect.36,37 Recent advances in induced pluripotent and embryonic stem cell technology have enabled the development of a cardiomyocyte that has demonstrated the in vitro contractile properties that may be ideal for cell replacement strategies.38,39 Future studies will focus on delivering stem cell-differentiated cardiomyocytes using our biological suture technology. Acknowledgments This study was supported by National Institutes of Health grants RO1-HL115282 (G.R.G.), the National Science Foundation DGE 1144804 (K.J.H., G.R.G.), and the American Heart Association 12PRE9020036 (J.T.F). We would like to thank Emily Abbate and Nicole Chittim for their help with surgeries and additional analysis. Author Disclosure Statement G.R.G. is cofounder and CTO for VitaThreads, LLC. J.T.F. and J.P.F. have equity interest in VitaThreads LLC. Abbreviations Used HDMhigh density mapping hMSChuman mesenchymal stem cell LADleft anterior descending MImyocardial infarction MSCmesenchymal stem cell MSCGMmesenchymal stem cell growth medium PBSphosphate-buffered saline Qdotquantum dot RSWregional stroke work SACsystolic area of contraction SEMstandard error of the mean ==== Refs References 1 Go AS , Mozaffarian D , Roger VL , et al. Heart disease and stroke statistics—2014 update: a report from the American Heart Association . Circulation . 2014 ;129 :e28 –e292 24352519 2 Boersma E , Mercado N , Poldermans D , et al. Acute myocardial infarction . Lancet . 2003 ;361 :847 –858 12642064 3 Reddy K , Khaliq A , Henning RJ Recent advances in the diagnosis and treatment of acute myocardial infarction . World J Cardiol . 2015 ;7 :243 –276 26015857 4 Hare JM , Traverse JH , Henry TD , et al. A randomized, double-blind, placebo-controlled, dose-escalation study of intravenous adult human mesenchymal stem cells (prochymal) after acute myocardial infarction . J Am Coll Cardiol . 2009 ;54 :2277 –2286 19958962 5 Meyer GP , Wollert KC , Lotz J , et al. Intracoronary bone marrow cell transfer after myocardial infarction: eighteen months' follow-up data from the randomized, controlled BOOST (BOne marrOw transfer to enhance ST-elevation infarct regeneration) trial . Circulation . 2006 ;113 :1287 –1294 16520413 6 Toma C , Pittenger MF , Cahill KS , et al. Human mesenchymal stem cells differentiate to a cardiomyocyte phenotype in the adult murine heart . Circulation . 2002 ;105 :93 –98 11772882 7 Rose RA , Jiang H , Wang X , et al. Bone marrow-derived mesenchymal stromal cells express cardiac-specific markers, retain the stromal phenotype, and do not become functional cardiomyocytes in vitro . Stem Cells . 2008 ;26 :2884 –2892 18687994 8 Laflamme MA , Murry CE Heart regeneration . Nature . 2011 ;473 :326 –335 21593865 9 Mirotsou M , Zhang Z , Deb A , et al. Secreted frizzled related protein 2 (Sfrp2) is the key Akt-mesenchymal stem cell-released paracrine factor mediating myocardial survival and repair . Proc Natl Acad Sci . 2007 ;104 :1643 –1648 17251350 10 Yagi H , Soto-Gutierrez A , Parekkadan B , et al. Mesenchymal stem cells: mechanisms of immunomodulation and homing . Cell Transplant . 2010 ;19 :667 –679 20525442 11 Hou D , Youssef EA , Brinton TJ , et al. Radiolabeled cell distribution after intramyocardial, intracoronary, and interstitial retrograde coronary venous delivery: implications for current clinical trials . Circulation . 2005 ;112 :I150 –I156 16159808 12 Vrtovec B , Poglajen G , Lezaic L , et al. Comparison of transendocardial and intracoronary CD34+ cell transplantation in patients with nonischemic dilated cardiomyopathy . Circulation . 2013 ;128 :S42 –S49 24030420 13 Laflamme MA , Murry CE Regenerating the heart . Nat Biotech . 2005 ;23 :845 –856 14 Proulx MK , Carey SP , Ditroia LM , et al. Fibrin microthreads support mesenchymal stem cell growth while maintaining differentiation potential . J Biomed Mater Res A . 2011 ;96 :301 –312 21171149 15 Guyette JP , Fakharzadeh M , Burford EJ , et al. A novel suture-based method for efficient transplantation of stem cells . J Biomed Mater Res A . 2013 ;101 :809 –818 22961975 16 Tao ZW , Favreau JT , Guyette JP , et al. Delivering stem cells to the healthy heart on biological sutures: effects on regional mechanical function . J Tissue Eng Regen Med. 2014 [Epub ahead of print]; DOI: 10.1002/term.1904 17 De Mey S , De Sutter J , Vierendeels J , et al. Diastolic filling and pressure imaging: taking advantage of the information in a colour M-mode Doppler image . Eur J Echocardiogr . 2001 ;2 :219 –233 11888816 18 Kelly DJ , Azeloglu EU , Kochupura PV , et al. Accuracy and reproducibility of a subpixel extended phase correlation method to determine micron level displacements in the heart . Med Eng Phys . 2007 ;29 :154 –162 16531092 19 Foroosh H , Zerubia JB , Berthod M Extension of phase correlation to subpixel registration . IEEE Trans Image Process . 2002 ;11 :188 –200 18244623 20 Schächinger V , Erbs S , Elsässer A , et al. Intracoronary bone marrow-derived progenitor cells in acute myocardial infarction . N Engl J Med . 2006 ;355 :1210 –1221 16990384 21 Chen SL , Fang WW , Ye F , et al. Effect on left ventricular function of intracoronary transplantation of autologous bone marrow mesenchymal stem cell in patients with acute myocardial infarction . Am J Cardiol . 2004 ;94 :92 –95 15219514 22 Malliaras K , Makkar RR , Smith RR , et al. Intracoronary cardiosphere-derived cells after myocardial infarction: evidence of therapeutic regeneration in the final 1-year results of the CADUCEUS trial (CArdiosphere-Derived aUtologous stem CElls to reverse ventricUlar dySfunction) . J Am Coll Cardiol . 2014 ;63 :110 –122 24036024 23 Traverse JH , Henry TD , Ellis SG , et al. Effect of intracoronary delivery of autologous bone marrow mononuclear cells 2 to 3 weeks following acute myocardial infarction on left ventricular function: the latetime randomized trial . JAMA . 2011 ;306 :2110 –2119 22084195 24 Traverse JH , Henry TD , Pepine CJ , et al. Effect of the use and timing of bone marrow mononuclear cell delivery on left ventricular function after acute myocardial infarction: the time randomized trial . JAMA . 2012 ;308 :2380 –2389 23129008 25 Surder D , Manka R , Lo Cicero V , et al. Intracoronary injection of bone marrow-derived mononuclear cells early or late after acute myocardial infarction: effects on global left ventricular function . Circulation . 2013 ;127 :1968 –1979 23596006 26 Mathiasen AB , Qayyum AA , Jørgensen E , et al. Bone marrow-derived mesenchymal stromal cell treatment in patients with severe ischaemic heart failure: a randomized placebo-controlled trial (MSC-HF trial) . Eur Heart J . 2015 ;36 :1744 –1753 25926562 27 Bartunek J , Behfar A , Dolatabadi D , et al. Cardiopoietic stem cell therapy in heart failure: the C-CURE (Cardiopoietic stem Cell therapy in heart failURE) multicenter randomized trial with lineage-specified biologics . J Am Coll Cardiol . 2013 ;61 :2329 –2338 23583246 28 Wollert KC , Meyer GP , Lotz J , et al. Intracoronary autologous bone-marrow cell transfer after myocardial infarction: the BOOST randomised controlled clinical trial . Lancet . 2004 ;364 :141 –148 15246726 29 Hamano K , Nishida M , Hirata K , et al. Local implantation of autologous bone marrow cells for therapeutic angiongenesis in patients with ischemic heart disease . Jpn Circ J . 2001 ;65 :845 –847 11548889 30 Perin EC , Willerson JT , Pepine CJ , et al. Effect of transendocardial delivery of autologous bone marrow mononuclear cells on functional capacity, left ventricular function, and perfusion in chronic heart failure: the focus-cctrn trial . JAMA . 2012 ;307 :1717 –1726 22447880 31 Krzeminski TF , Nozynski JK , Grzyb J , et al. Wide-spread myocardial remodeling after acute myocardial infarction in rat. Features for heart failure progression . Vasc Pharmacol . 2008 ;48 :100 –108 32 Christman KL , Vardanian AJ , Fang Q , et al. Injectable fibrin scaffold improves cell transplant survival, reduces infarct expansion, and induces neovasculature formation in ischemic myocardium . J Am Coll Cardiol . 2004 ;44 :654 –660 15358036 33 Dai W , Wold LE , Dow JS , et al. Thickening of the infarcted wall by collagen injection improves left ventricular function in rats: a novel approach to preserve cardiac function after myocardial infarction . J Am Coll Cardiol . 2005 ;46 :714 –719 16098441 34 Rane AA , Christman KL Biomaterials for the treatment of myocardial infarctiona 5-year update . J Am Coll Cardiol . 2011 ;58 :2615 –2629 22152947 35 Rosen AB , Kelly DJ , Schuldt AJ , et al. Finding fluorescent needles in the cardiac haystack: tracking human mesenchymal stem cells labeled with quantum dots for quantitative in vivo three-dimensional fluorescence analysis . Stem Cells . 2007 ;25 :2128 –2138 17495112 36 Behfar A , Yamada S , Crespo-Diaz R , et al. Guided cardiopoiesis enhances therapeutic benefit of bone marrow human mesenchymal stem cells in chronic myocardial infarction . J Am Coll Cardiol . 2010 ;56 :721 –734 20723802 37 Zhang M , Mal N , Kiedrowski M , et al. SDF-1 expression by mesenchymal stem cells results in trophic support of cardiac myocytes after myocardial infarction . FASEB J . 2007 ;21 :3197 –3207 17496162 38 Takahashi K , Yamanaka S Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors . Cell . 2006 ;126 :663 –676 16904174 39 Laflamme MA , Chen KY , Naumova AV , et al. Cardiomyocytes derived from human embryonic stem cells in pro-survival factors enhance function of infarcted rat hearts . Nat Biotechnol . 2007 ;25 :1015 –1024 17721512 Cite this article as: Hansen KJ, Favreau JT, Guyette JP, Tao Z-W, Coffin ST, Cunha-Gavidia A, D'Amore B, Perreault LR, Fitzpatrick JP, DeMartino A, Gaudette GR (2016) Functional effects of delivering human mesenchymal stem cell-seeded biological sutures to an infarcted heart, BioResearch Open Access 5:1, 249–260, DOI: 10.1089/biores.2016.0026.

PMC005xxxxxx/PMC5003012.txt

==== Front Biores Open AccessBiores Open AccessbioresBioResearch Open Access2164-78442164-7860Mary Ann Liebert, Inc. 140 Huguenot Street, 3rd FloorNew Rochelle, NY 10801USA 10.1089/biores.2016.002210.1089/biores.2016.0022Comprehensive ReviewRecent Applications of Coaxial and Emulsion Electrospinning Methods in the Field of Tissue Engineering McClellan Phillip Landis William J. *Department of Polymer Science, The University of Akron, Akron, Ohio.* Address correspondence to: William J. Landis, PhD, Department of Polymer Science, Goodyear Polymer Center, Room 1201C, The University of Akron, 171 University Avenue, Akron, OH 44325-3909, E-mail: wlandis@uakron.edu01 8 2016 2016 01 8 2016 5 1 212 227 © Phillip McClellan and William J. Landis 2016; Published by Mary Ann Liebert, Inc.2016This Open Access article is distributed under the terms of the Creative Commons License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.Abstract Electrospinning has emerged as an effective method of producing nanoscale fibers for use in multiple fields of study. One area of significant interest is nanofiber utilization for tissue engineering because the nanofibrous mats can mimic the native extracellular matrix of biological tissues. A logical next step is the inclusion of certain molecules and compounds to accelerate or increase the efficacy of tissue regeneration. Two methods are under scrutiny for their capability to encapsulate therapeutic compounds within electrospun nanofibers: emulsion and coaxial electrospinning. Both have advantages and disadvantages, which need to be taken into careful consideration when deciding to use them in a specific application. Several examples are provided here to highlight the vast potential of multilayered nanofibers as well as the emergence of new techniques to produce three-dimensional scaffolds of nanofibers for use in the field of tissue engineering. Keywords:  coaxial electrospinningdrug deliveryemulsion electrospinningtissue engineering ==== Body Introduction History and background information about electrospinning The technique known as electrospinning was discovered and described over a century ago1 as a derivative of electrospraying, a method that utilizes electrostatic forces to generate polymer droplets.2–5 At the outset of electrospinning development, at least 11 patents were issued to Formhals over the course of ∼10 years (1934–1944) for various designs providing a method to generate fine polymer filaments.6–16 In 1964, Geoffrey Taylor published research which explained the deformation of water droplets in the presence of an electric field.17 His use of mathematical treatment and imaging demonstrated the distortion of spherical drops into conical shapes17 and, in 1969, he presented another article focusing on the ejection and formation of fine jets from viscous solutions.18 Annis et al. spent time in the late 1970s investigating electrospun polyurethane for potential vascular prostheses,19 but, overall, electrospinning was not considered to have many useful applications. The technique languished for many years until the early 1990s when interest was renewed largely as a result of the need for a process which would enable the fabrication of fibers on the nanometer scale. In this regard, Reneker, Yarin, and other researchers began to examine more closely the fibers that were formed by this process and investigate the physics governing their formation.20–31 Electrospinning is possible because of an imbalance between two physical properties common to the process, surface tension, and electric field strength. A solution or melt with sufficiently high viscosity passes through an orifice, usually at the end of a needle or metal cone with a small hole at its tip, to generate a small droplet (Fig. 1). When surface tension dominates intermolecular interactions, the droplet of solution at the tip of the needle or cone is stable and remains in place. As the electric field strength increases, excess electrical charges, either positive or negative, accumulate on the surface of the suspended droplet and begin to overcome the surface tension. A Taylor cone17,18 is formed as the drop distorts and a small strand of solution, or jet, is ejected from its tip, generating a flow-modified Taylor cone. The excess electrical charges within the jet repel one another, the result of which leads to rapid elongation of the expelled strand of solution. The jet of solution travels to a collector, which is usually in the form of a flat, electrically conductive plate. As the strand of solution travels to the collector, the solvent evaporates to leave solid fibers. FIG. 1. Basic electrospinning apparatus. A polymer solution is loaded into a syringe (1) with a needle attached to a high voltage power supply (2). The nanofibers generated accumulate on the surface of a collector (3) connected to an oppositely charged electrode of the high voltage power supply. A flow-modified Taylor cone is highlighted in the upper left of the figure. Many parameters may be manipulated to optimize the electrospinning process. A few factors often altered when solvent-based methods are employed include applied voltage, distance between the electrospinning tip and collector, solvent volatility, electrical conductivity of the solution, and solution viscosity. These parameters have significant effects on the morphological characteristics of the formed nanofibers. Environmental factors such as temperature and humidity may also have a marked effect on the morphology of the fibers formed,20,21 and they have led to the creation of electrospinning designs that have tips and collectors either submerged in liquid32 or sealed in chambers with tightly controlled atmospheric conditions.33 Melt-based electrospinning is manipulated by some of the same factors as solvent-based electrospinning (applied voltage, distance between tip and collector, electrical conductivity of the melt), but it can also be modified by altering other aspects, such as the temperature of the melted polymer. Potential uses for electrospun nanofibrous mats A variety of nonwoven nanofiber mats may be generated by electrospinning. These have been utilized in applications for wound dressings,34–36 filtration devices,37 protective clothing and textiles,38 sensors,39 tissue scaffolds,40,41 and drug delivery systems.33,42,43 All of these areas benefit from the high surface area-to-mass ratio of the nanofiber mats. As an example, it is interesting to consider drug delivery in this context. A significant problem encountered in drug delivery is maintaining a therapeutic dosage to a patient over a specified period of time. Controlling the dosage rate is especially important in the case of drugs with a low therapeutic index, such as doxorubicin (dox)42 or digoxin,44 as they must be administered more often and in smaller amounts to minimize the risk of toxic side effects and remain within the window of concentration where they will be effective. However, as the frequency of required doses increases, patient compliance generally decreases. An effective way in which to ensure patients will adhere to the dosage schedule is to place the therapeutic agent within a device or chemical compound that will control the rate at which the agent is released and decrease the frequency of required doses. Insulin pumps, nicotine patches, and time-release tablets are a few examples of products currently in the consumer marketplace which utilize a form of controlled drug delivery. Hydrogels,45–47 dendrimers,48–51 liposomes,52–55 and microelectromechanical systems (MEMS)56–59 represent a small fraction of technologies under investigation for their use in drug delivery. One of the advantages of utilizing electrospun nonwoven nanofiber mats for drug delivery derives from the capability to employ them as wound dressings or tissue scaffolds for tissue engineering. The therapeutic agent can be loaded into the electrospun fibers and released at a steady rate over a period of time. Early biological applications of electrospun fibers concentrated on the field of surface wound repair for a few reasons. First, the materials used can be nonbiodegradable and consist of polymers which are already designed to form nanofiber mats.35,36 Second, the electrospun mats do not need to be subjected to the same rigorous testing protocols as a product used for implantation into the human body. Third, the morphological characteristics of the nanofibrous mats can be adjusted to allow air to pass through the mats while controlling or preventing the flow of moisture or dirt.38 Perhaps the most interesting feature of the nanofiber mats is the ability to couple them with antibacterial or regenerative agents to increase the rate of healing. Small molecules or particles may become trapped within nanofibers as the solutions are electrospun. For instance, titanium dioxide nanoparticles have been added to nanofibers of polyurethane, which demonstrated increased antibacterial activity and significant water vapor permeability, but exhibited relatively low cytotoxicity to cultured L929 (mouse fibroblast) cells. Water vapor permeability was examined because the desired application of these nanofibers was the treatment of burn wounds, which require moisture to prevent dehydration and promote tissue regeneration.35 Skin ulcers associated with Leishmaniasis were treated effectively through the use of electrospun mats which were designed to release nitric oxide upon the addition of a small amount of water.36 The primary issue encountered in utilizing nonbiodegradable materials is that they remain in place if they are implanted into the human body. To circumvent this problem, biodegradable polyesters such as polyglycolic acid (PGA), poly-l-lactic acid (PLLA), and polycaprolactone (PCL) were examined with the intent of creating biodegradable nanofibers that may be implanted into the human body and dissolved slowly over time with few, if any, deleterious effects. Tissue Scaffolds Synthetic, bioresorbable materials, such as polyethylene oxide (PEO),42 PGA,60 PCL,43 polylactic acid (PLA),42,43,60 and various copolymers of these compounds have been utilized to great effect to form tissue scaffolds of electrospun nanofibrous mats. Natural polymers, such as collagen and elastin, have also been successfully electrospun to provide nanofiber mats that closely resemble those components of the extracellular matrix of some biological systems.61 Chitosan,62–65 alginate,65,66 and hyaluronan67,68 are more natural materials, which have gained interest in light of their availability, affordability, and cellular compatibility. Whether synthetic or natural materials are used for electrospinning, living cells appear to attach and proliferate more readily on the nanofiber mats than when grown in monolayer culture.69–71 Figure 2 provides a graphical representation comparing cells grown in monolayer (spincoated poly-d,l-lactic acid [PDLLA]) and those grown on nanofibers (PDLLA fibers). The results for day 3 show a lower number of cells initially attached to the nanofibers. However, by day 14, the cell density on the nanofibrous materials has increased significantly compared with the cell density present on the spin-coated PDLLA. Figure 3 illustrates the difference in cell morphology when cells are grown in the presence of three-dimensional nanofibrous PDLLA. The cell shown in the second and third images is utilizing the nanofibers (shown in blue in the third image) as its external support structure. FIG. 2. Plot comparing the growth of MC3T3-E1 (mouse calvaria-derived osteoprogenitor) cells in monolayer and on nanofibers. Statistically significant differences from *tissue-culture polystryrene and #spin-coated PDLLA are noted. PDLLA, poly-d,l-lactic acid. (Reprinted with permission from Elsevier from Ref.69). FIG. 3. Immunofluorescent staining of adherent cells (MC3T3-E1) on spin-coated PDLLA (A) and PDLLA nanofibers (B). Green and red indicate vinculin and actin, respectively. (C) Shows the immunofluorescent staining image (B) superimposed onto a phase contrast image of PDLLA fibers (blue). (Reprinted with permission from Elsevier from Ref.69). Significant advances have been realized in recent years with regard to the utilization of electrospun nanofiber mats composed of biodegradable materials for tissue engineering. The relative simplicity of the electrospinning technique, coupled with the morphological characteristics of the fibrous mats, make it an attractive method for forming tissue scaffolds. By controlling the size, density, composition, and orientation of the fibers, it is possible to produce materials which approximate closely the fibrous structures that are present in the extracellular matrix of almost any tissue. For tissue engineering, the nanofiber mats containing a therapeutic agent can be utilized as part of a tissue scaffold. In such an application, the electrospun fibers should serve to mimic the three-dimensional environment of the extracellular matrix of the target tissue as well as release small amounts of the desired agent in a controlled fashion. The composition of the fibers may be altered to optimize their drug-loading efficiency, activity, and biocompatibility. Aligned PLLA nanofibers for neural tissue engineering With respect to a specific example, PLLA was selected as a suitable material for fabricating mats of random and aligned nanofibers in at least one case of neural tissue engineering.72 The viability of mouse neuronal stem cells was not significantly affected by the alignment of the fibers. However, the alignment of the fibers served to influence the cells to grow in a more directed fashion similar to those found within native nerve tissue. Also, the nerve cells on aligned fibers demonstrated a significant increase in neurite length. Collagen and elastin nanofibers for smooth muscle tissue Attempts to utilize natural polymers, such as collagen and elastin, have been met with success in some instances. The use of collagen/elastin mats for smooth muscle tissue engineering demonstrates promise because the fibers exhibit similar mechanical properties as the smooth muscle tissue located within blood vessel walls.61 Vascular grafts could be tailored to specific situations with only minor modifications to the nanofiber collector. Electrospun collagen/elastin mats could also be utilized to repair the outer walls of the stomach or intestinal tract. Nanofibrous mats containing calcium carbonate or hydroxyapatite nanoparticles Wutticharoenmongkol et al. generated nanofibrous mats of PCL that contained calcium carbonate or hydroxyapatite nanoparticles for bone tissue regeneration.73 The nanofibrous mats produced were referred to as a “guided bone regeneration membrane” and it was postulated that the nanoparticles would facilitate the proliferation and differentiation of osteoblasts.73 Early results indicated that the scaffolds did not exhibit any cytotoxic effects on human osteoblasts or rat fibroblasts.73 In addition, the osteoblasts grown on nanofibers containing calcium carbonate were examined under scanning electron microscopy and demonstrated the initial signs of mineralization with the appearance of small crystalline granules. A notable potential consequence of inducing and maintaining osteoblast differentiation was mentioned in the publication. The study found that the osteoblasts which had differentiated slowed considerably with respect to proliferation.73 A similar situation was mentioned by Owen et al. in a study of rat osteoblasts grown in vitro.74 Core–Sheath Nanofiber Structures Emulsion electrospinning Although the electrospinning technique is relatively simple, difficulties may occur with the method in aspects of loading therapeutic agents into the fibers. Problems may be encountered, for example, with proteins because of their size and requirements relative to their three-dimensional structure. In this regard, small conformational changes could occur as a result of the electrospinning process (solvent exposure, shear stresses) and they may render a protein inactive or possibly even toxic. To circumvent such a situation, proteins may require the use of more complex techniques, such as emulsion or coaxial electrospinning, to be loaded efficiently and effectively into nanofibers to be electrospun. In the case of both emulsion and coaxial electrospinning, the nanofibers generated consist of an outer sheath and inner core of differing composition. Emulsion electrospinning relies on chemical means of separation through the creation of an emulsion within a single solution and the subsequent organization of the emulsified droplets into two distinct phases as the solvent evaporates from the electrospun fibers.75 This technique has proven successful in encapsulating nerve growth factor (NGF),60 proteinase K,76 lysozyme,77 cytochrome C,78 and bovine serum albumin.79 Emulsion electrospinning is also being investigated because it accommodates the use of water as the solvent instead of more hazardous chemicals.80 In many cases, surfactants or other amphiphilic compounds are used to generate the emulsified droplets in water-based systems.60,79,80 Nanofibers of silk fibroin and PCL Before work with encapsulating therapeutic agents, suitable solution and electrospinning parameters should be determined. For example, core–sheath-structured nanofibers of silk fibroin (SF) and PCL have been constructed utilizing emulsion electrospinning.81 SF is similar to collagen in that it is a biopolymer obtained from natural sources and contains cell recognition peptides such as RGD (arginine-glycine-aspartic acid), which facilitate cell attachment and proliferation.82,83 The main focus in generating SF/PCL fibers was to capitalize on the mechanical strength of the PCL while exploiting the advantages of SF as a tissue engineering material. Testing in vitro using primary fibroblast cells from human skin (FEK4) demonstrated an increase in initial cell attachment as well as a slight increase in cell proliferation rate. After positive results were obtained from the SF/PCL combination, properties of the nanofibers were optimized to a greater degree by the addition of hyaluronan to the electrospinning solution.84 Hyaluronan, or hyaluronic acid (HA), is a glycosaminoglycan found in the extracellular matrix of many tissues and it can promote cellular adhesion and proliferation.67 The purpose of its inclusion into the SF/PCL emulsion was to provide some measure of control over the adsorption of proteins to the surface of the nanofibers while still enhancing infiltration of cells into the scaffold. Water swelling and contact angle measurements indicated an increase in hydrophilicity of the nanofibrous scaffolds as a result of HA addition. Also, the rate of FEK4 cell infiltration and proliferation was significantly higher after 5 days when compared with scaffolds composed only of SF and PCL.84 One potential challenge for researchers intending to capitalize on the success of this study and others like it results from the use of a fluorinated solvent, in this case, hexafluoroisopropanol (HFIP). HFIP is a corrosive, volatile solvent, which may not be suitable for biological molecules or desired therapeutic agents.85 Other solvent systems were developed to combat this issue and can be used to great effect with the emulsion electrospinning methodology.85,86 Encapsulation and controlled release of doxorubicin As outlined in the Emulsion Electrospinning section, the emulsion electrospinning technique requires a less complex apparatus than comparable coaxial arrangements to generate core–sheath-structured nanofibers. Only one syringe pump and electrospinning tip are required. The core and sheath of the nanofibers can be simply the result of the separation between organic and aqueous phases during the evaporation of the solvent as the fibers are formed. In some cases, an emulsifier such as sodium dodecyl sulfate (SDS) is required to produce a stable water-in-oil or oil-in-water emulsion for electrospinning.42 The thickness of the inner and outer layers of the nanofibers can be altered by changing solution parameters as well. When Xu et al. encapsulated dox within PLA/PEO nanofibers, they discovered that increasing the concentration of dox within the core of the fibers shrinks the diameter of the core as a result of the favorable association between dox molecules.42 The association and smaller diameter core also had the effect of significantly decreasing the release rate of the drug from 60% to only 20% released within the first 5 h. More importantly, studies such as this one demonstrate the feasibility of producing nanofibers containing aqueous and organic phases. This dual-phase nature is critical not only in the case of small molecule therapeutics, but even more so for proteins and growth factors. Altering cytochrome C release characteristics in PLLA nanofibers A study by Maretschek et al. demonstrated the encapsulation of a relatively small heme protein, cytochrome C, within nanofibers of PLLA.78 The work showed that a change of 1–3% PLLA concentration significantly slowed the initial release of cytochrome C from the core of the fibers, essentially eliminating the burst-release phenomenon.78 This investigation highlights the concept that manipulating the release rate of the therapeutic agent can be achieved through altering the sheath of the nanofibers. The medium into which the nanofiber mats are placed can exhibit a marked effect on release characteristics as well. This research group demonstrated that the release rate of cytochrome C could also be altered through the addition of hydrophilic polymers such as polyethyleneimine (PEI) or poly-l-lysine (PLL).78 By addition of one of these polymers to the emulsified electrospinning solution, the hydrophobicity of the nanofiber mats was decreased and the release rate of the protein was increased, results which also reintroduced the initial burst-release phenomenon. The capability to manipulate and control the release process shown by this study is critical in many cases for effective administration of therapeutic agents. Such control is especially important with respect to growth factor-centered treatments as it may be beneficial to release certain factors in high concentrations initially and slowly decrease the dosage rate over time. Controlling the release profile of basic fibroblast growth factor in PELA nanofibers A more thorough examination of the release profiles for encapsulated proteins was undertaken by Yang et al.34 Specifically, this group focused on the release of basic fibroblast growth factor (bFGF) sequestered within polyethylene glycol/poly-DL-lactide (PELA) nanofibers. Figure 4 shows release characteristics of bFGF over a period of 25 days in vitro. The initial 12 h after incubation (1 × phosphate-buffered saline, 37°C) of the drug-loaded nanofiber mats demonstrated the burst-release phenomenon, freeing ∼14% of the total bFGF from the nanofibers. Over the next 15 days, another 60% of the growth factor was liberated from the nanofibers during the “fast” phase of release. Following another 10 days, the release rate tapered off significantly and resulted in 15% of the protein being freed from the fibers during the “slow” release phase. FIG. 4. Release characteristics over 25 days in vitro for PELA nanofibers loaded with bFGF. Nanofibrous mats of PELA nanofiber containing bFGF (n = 3) were suspended in PBS (pH 7.4) in tubes that were placed in a shaking water bath (37°C, 100 cycles/min). PBS was collected and analyzed by ELISA at specified intervals to determine the percent of total bFGF released. bFGF, basic fibroblast growth factor; ELISA, enzyme-linked immunosorbent assay; PBS, phosphate-buffered saline; PELA, polyethylene glycol/poly-dl-lactide. (Reprinted with permission from Elsevier from Ref.34). The bFGF-loaded nanofiber mats were tested in vitro for compatibility with living cells using mouse embryo fibroblasts. For testing in vivo, the growth factor-loaded nanofiber mats were applied to dorsal wounds in diabetic rats.34 After 1 week, the wounds covered with the bFGF-loaded PELA nanofibers had already begun to heal, whereas the PELA nanofiber mats without the growth factor and containing free growth factor had not. Within 3 weeks, the wounds covered with growth factor-loaded nanofibers had completely healed. The other testing groups required an additional week to heal, and the control group still had not healed fully at the end of 4 weeks. Such marked success in wound healing was attributed by Yang et al. to three factors: encapsulation of bFGF prevented its early degradation in the proteolytic environment of the wound, the nanofibers kept the factor localized to the wound, and sustained release of bFGF was maintained over time.34 In this case, the triphasic release profile of bFGF may have been beneficial to wound healing. The initial burst of the growth factor may facilitate rapid proliferation of fibroblasts to start the healing process. The second and third phases of release follow with decreasing release rates, prolonging the presence of the growth factor within the wound while slowing the proliferation of cells as the wound shrinks and eventually closes. Coaxial electrospinning Coaxial electrospinning differs significantly from emulsion-centered methods in that it generates core–sheath fibers by physical separation through the utilization of two electrospinning tips and two solutions.72 Certain processing parameters and solution properties such as inner and outer solution flow rates, viscosities, and electrical conductivities are often taken into account when attempting to apply the coaxial technique. The coaxial electrospinning process can be utilized to create drug delivery systems of varying complexity. For instance, the core and sheath compositions of the fibers could be chosen for strength and cell attachment properties, respectively. The approach has been demonstrated by electrospinning nanofibers with a thermoplastic polyurethane core and a collagen sheath.87 Fibers such as these provide desirable mechanical strength as a result of the polyurethane core and facilitate increased cell attachment and proliferation because of the presence of the collagen sheath. Bilayer nanofibers with a polyurethane core and a collagen sheath As noted above, one of the main advantages of fabricating a tissue scaffold of nanofibers with a core–sheath structure is the capability to combine materials of varying mechanical and chemical properties. An example of such a combination was demonstrated by Chen et al., who produced nanofibers, which were composed of a polyurethane core and a collagen sheath.87 The polyurethane core contributed significant mechanical strength to the nanofiber mats, which would not have been realized using a pure collagen scaffold. Stress–strain curves (Fig. 5) generated by tensile test measurements showed that the collagen-coated polyurethane fibers were not as strong as pure urethane fibers, but they were significantly more robust than nanofibers of pure collagen. The decrease in mechanical strength was acceptable because a scaffold composed purely of synthetic polyurethane would not exhibit the same cell recognition and attachment as one fabricated with collagen. Studies in vitro utilizing pig iliac endothelial cells demonstrated a marked increase in the number of viable cells adherent to the scaffolds after 1, 3, and 5 days of cell culture when compared with polyurethane nanofibers alone, polyurethane fibers bulk-coated with collagen, and glass cover-slips.87 It is important to note that any nanofibers fabricated with collagen on their surface typically require an extra step to crosslink the collagen and preserve the structure of the nanofibers.61,87,88 Overlooking this critical step will result in dissolution of the collagen into aqueous medium. FIG. 5. Stress–strain curve for electrospun TPU, collagen, and TPU/collagen coaxial nanofibers highlighting the mixture of physical properties that results from the combination of collagen and TPU in the nanofibers. TPU, thermoplastic polyurethane. (Reprinted with permission from Elsevier from Ref.87). Production of poly(glycerol sebacate) nanofibers The coaxial technique has been employed to produce single-phase, solid nanofibers from materials which would be extremely troublesome to electrospin under normal conditions.89 Poly(glycerol sebacate) (PGS) is a degradable synthetic material which exhibits mechanical properties similar to elastin.90 Nanofibers of PGS are difficult to produce under basic electrospinning conditions because they require curing after the fibers are produced to generate the crosslinks which allow the material to display three-dimensional morphologies similar to elastin and collagen. During the curing process, the nanofibers need to remain separated from one another to prevent them from melting and merging into larger fibers or thin films. Yi and LaVan were successful in generating nanofiber mats of PGS by utilizing PLLA as a sheath to sequester the uncured PGS and ensure it would crosslink while remaining as nanofibers.90 Once the PGS was crosslinked, the PLLA was removed using dichloromethane to leave a nanofibrous mat of pure PGS. They proposed using the PGS nanofiber scaffolds for tissue engineering in microvasculature and showed the nanofibers had no significant effects on viability of human dermal microvascular endothelial cells.90 Encapsulation of NGF in PLGA nanofibers Research groups have demonstrated methods for encapsulating successfully NGF in poly(lactic-co-glycolic acid) (PLGA)91 and vascular endothelial growth factor (VEGF) within Dextran/PLGA92 as well. In both cases, the encapsulated growth factors maintained significant bioactivity and were not altered by the coaxial electrospinning process. In the instance of NGF encapsulated within PLGA, Wang et al. demonstrated a controlled, linear-release profile following an initial burst-release of 29.5%.91 Also, the nanofibrous mats were electrospun in an aligned fashion and rolled into small tubes for testing as a potential nerve guidance conduit (NGC). Results from studies in vivo indicated the NGF-loaded nanofibers were as effective as autograft nerve tissue for regenerating tissue in a rat sciatic nerve model.91 Figure 6A shows the insertion of the NGF/PLGA tube in a rat having a 13-mm defect introduced into its sciatic nerve. In Figure 6B, the NGC is shown 12 weeks after the initial implantation. The conduit is completely encapsulated by this time. The regenerated nerve is illustrated in Figure 6C, where the NGF/PLGA tube has been removed from the sciatic nerve of the rat at the 12-week time point. In addition to demonstrating the effectiveness of encapsulation of NGF within nanofibers, this study also highlighted short-term effects in vivo. No significant adverse health effects were noted throughout the course of the experiment. It appeared that the NGF was released locally around the nerve defect and did not have a systemic effect.91 FIG. 6. Surgical implantation of an aligned NGC for nerve regeneration in a rat model of a 13-mm sciatic nerve defect viewed under a microscope. (A) The NGC was used to bridge the 13-mm nerve defect; (B) the appearance of the NGC 12 weeks after implantation; (C) the regenerated nerve after removing the NGC. NGC, nerve guidance conduit. (Reprinted with permission from Taylor and Francis from Ref.90). Nanofibers of dextran/PLGA loaded with VEGF In a separate study, dextran/PLGA nanofiber mats loaded with VEGF were tested for cell proliferation and growth factor viability using human umbilical vein endothelial cells.92 It was noted that the cells began to infiltrate the pores of the VEGF-loaded nanofibrous mats as soon as day 3 after seeding and commenced secreting extracellular matrix, which began to be incorporated with the nanofibers to provide a suitable environment for additional cell growth.92 Figure 7 shows transmission and scanning electron micrographs (TEM and SEM, respectively) of coaxially spun nanofibers. The TEM micrograph (Fig. 7A) emphasizes the “hollow” or tube-like nature of the spun fibers, whereas the SEM image (Fig. 7B) demonstrates the uniformity of the nanofiber size distribution. Both of these factors are critical for controlling drug-loading and release characteristics using nanofibers. Nanofibers having nonuniform diameters and a high degree of variability with respect to the thickness of the walls of the tube could exhibit unpredictable drug-release profiles. FIG. 7. Photomicrographs from transmission (A) and scanning (B) electron microscopy of DEX(VEGF)/PLGA fibers prepared by coaxial electrospinning. DEX, dextran; PLGA, poly(lactic-co-glycolic acid); VEGF, vascular endothelial growth factor. (Reprinted with permission from Taylor and Francis from Ref.92). Another significant advantage of coaxial electrospinning for tissue scaffolds and controlled drug delivery is seen in an experiment conducted using myoblasts isolated from the skeletal muscle of mice.93 Coaxial nanofibers with a core containing VEGF and a shell composed of polyurethane were produced and demonstrated more efficient and even loading of the factor into the nanofibers when compared to simple electrospinning. Also, sustained release of the factor occurred over the course of 30 days with initial burst release occurring over the first 6 days in culture. Cell-seeded constructs were implanted into hemophilic mice. The skeletal muscle-engineered constructs demonstrated marked improvement compared with the control groups. Histological examination showed an increase in vascularity and presence of significantly more and larger lymphatic vessels.93 Comparing Emulsion and Coaxial Electrospinning Techniques and Potential Applications Emulsion, compared to coaxial, electrospinning is the more prevalent technique because it provides protein solvation within a mild solvent and separation from the harsh solvent required to dissolve the desired polymer.43,75,77 Emulsion electrospinning also has the advantage of simplicity, but it lacks well defined control over the placement of the therapeutic agent within either the core or shell of the structure. Enabling precise control over the location of the drug within the core or shell of the nanofibers is one of the advantages of using coaxial electrospinning instead of emulsion.94 Reproducibility and high-throughput are two other advantages in designing drug delivery devices using coaxial electrospinning. The major disadvantage with coaxial electrospinning lies in the complexity of its setup. A coaxial needle and multiple syringe pumps are required to control the precise formation of the core–sheath structures. The syringe pumps are relatively simple to arrange for effective delivery of polymer solutions. The design of the coaxial needle is critical to the ultimate arrangement of the core–sheath fibers. The simplest and most often utilized needle arrangement is a single core centered within the outer sheath.69,87 More complex strategies place multiple needles within the center to produce nanofibers having more than a single core.95–97 The fabrication of intricate needle designs such as these serves to add costs to the initial setup of the coaxial electrospinning apparatus and is investigated in situations where a simpler technique is not viable. Utilization of the coaxial electrospinning technique could allow creation of more complex drug delivery systems as well.97 For example, the inner layer of the electrospun fibers could be composed of PCL, loaded with osteogenic protein-1 (OP-1), whereas the outer layer could consist of collagen, loaded with bFGF. The reason for utilizing OP-1 and bFGF in this case is that OP-1 has been shown to promote differentiation of chondroblasts and osteoblasts98 and bFGF has a demonstrated capability to increase cell proliferation.99,100 Conceptually then, bFGF loaded into the outer sheath of the nanofibers would be released into the tissue environment before OP-1 encapsulated within the core of the fibers. Thus, this type of scaffold could promote delivery of factors or drugs in two stages, the first for enhancing cell proliferation and the second for ensuring that the cells differentiate into the desired cell type. A potential pitfall, which could hinder this approach, might result from the use of PCL as the material for the inner layer and the application of an organic solvent required for PCL dissolution. The organic solvent selected could alter the activity of OP-1, potentially reducing or eliminating the advantageous effects of its incorporation into the nanofibers. Release profiles of OP-1 and bFGF should be governed by the thickness of the inner core and outer sheath as well as the chemical composition of each electrospun fiber layer.77,78 As the nanofiber degradation will not be completely uniform, an overlap in the release profiles between OP-1 and bFGF may be present. It should be possible to adjust this window of overlap to maintain an optimal ratio of the two factors. Such an approach would be beneficial in many applications which require multiple drugs to be released in a sequential manner. Research has already demonstrated the capability to fabricate multidrug nanofiber mats using sequential electrospinning to generate a multilayered mesh (Fig. 8).101 The approach required two extra layers of fibers to delay the release of the second drug and produced fibers of differing morphologies in all four layers. As the fibers in every layer were composed of poly-l-lactide-co-ɛ-caprolactone (PLCL), macroscopic separation of the mesh layers did not appear to be an issue. However, the multilayered mats were being examined principally for applications in sequential chemotherapy, not in tissue engineering.101 Therefore, the meshes were not created to optimize attachment and proliferation of cells. Also, using multilayered fibers could prove to be problematic in tissue engineering applications because the cells may infiltrate all the layers and begin degrading them simultaneously, resulting in premature release of the second drug. This situation could be prevented, in concept, by generating nanofibrous mats having sufficient fiber density and small pore size to prevent rapid infiltration of cells. Small pore electrospun mats that hinder cell infiltration might be advantageous in wound healing applications, for example, in which growth factors and related molecules could penetrate a fiber mesh relatively easily, whereas cells, bacteria, and other such larger elements could not. Considerations of fiber size and porosity are important related to diffusion of various molecules and infiltration of cells into electrospun mats for tissue engineering applications. Discussion of these factors has been presented in multiple studies.69,102–105 FIG. 8. Scanning electron micrographs of layered mesh scaffolds composed of PLCL and loaded with multiple therapeutic agents. TPPS-loaded and ChroB-loaded meshes with different mesh thickness. (A–C) TPPS meshes with 23 ± 2 μm, 40 ± 3 μm, and 81 ± 4 μm thickness, respectively. (D–F) ChroB meshes with 28 ± 5 μm, 58 ± 3 μm, and 103 ± 5 μm thickness, respectively. PLCL, poly-l-lactide-co-ɛ-caprolactone; TPPS, 5,10,15,20-tetraphenyl-21H,23H-porphinetetrasulfonic acid, disulfuric acid; ChroB, chromazurol B (2,6-dichloro-4′-hydroxy-3′,3″-dimethylfuchsone-5′,5″-dicarboxylic acid disodium salt). (Reprinted with permission from Elsevier from Ref.101). In any case, the aforementioned applications of coaxial electrospinning in tissue engineering and drug delivery increase the complexity of the basic technique. Controlling the release profiles and thickness of the inner and outer electrospun fiber layers may be challenging, for instance. Any therapeutic agents loaded into the electrospinning solutions must also remain active after being subjected to the various organic solvents and the electric field in the electrospinning operation. As a beneficial side effect, the sequestration of a drug within electrospun nanofibers could prolong the amount of time a drug could be stored.106 This possibility is of particular interest for many therapeutic formulations in the pharmaceutical industry. Long-term factor or drug storage can be a significant obstacle to production of compounds, which are not required on a regular basis or require transport to locations lacking the necessary infrastructure to preserve them. Proper preparation is still important when storing or transporting therapeutic agents in this fashion. For instance, PCL, PLA, and PGA undergo hydrolytic degradation under normal environmental conditions. It would, therefore, be prudent to store drug-loaded nanofibers containing such compounds in an environment with very low humidity. While significant advances are being made in developing water-soluble107,108 and chemically stable solutions,109–111 the pharmaceutical industry will benefit greatly from the introduction of simple preservation methods to prolong the useful life and bioavailability of unstable, yet effective, formulations. Current challenges in designing drug delivery systems utilizing electrospun nanofibers Obstacles preventing the use of electrospinning for mass production of drug delivery devices obviously still exist. Establishing general parameters for successful encapsulation and release of therapeutic agents is a difficult task because of the myriad of polymers, solvents, experimental conditions, and drugs which may be utilized. Basic guidelines regarding solvent properties, solution viscosities, and experimental parameters have been set forth by multiple researchers for basic and coaxial electrospinning. It is critical to determine the specific application in which the drug-loaded nanofibers will be utilized. In the instances where the therapeutic agent is not affected by the solvent used to prepare a polymer for electrospinning, basic electrospinning may be sufficient to produce the desired result. Utilization of coaxial or emulsion electrospinning to generate core-sheath-structured nanofibers may be required to construct a suitable drug delivery device if the drug or growth factor is not suitable for basic electrospinning. The examples provided in the preceding pages represent but a small fraction of the research being conducted in the field of drug delivery using coaxial and emulsion electrospinning. In addition, multiple drug delivery technologies were combined in a number of studies to various rates of success. Sequestration of drug molecules within a carrier before production of nanofiber structures can increase the efficacy of delivery and control release profiles to great effect.112–115 This result may complicate matters further with respect to selection of appropriate solvents and electrospinning conditions. Particle size and composition may be critical to preservation of particle structure and release characteristics. In addition, the presence of particles within the fibers can affect the physical properties of the electrospun mats.116–118 Modifications to the coaxial electrospinning approaches facilitated creation of multiaxial designs and trilayer nanofibers that exhibit three distinct layers in individual fibers. Drugs may be loaded into any or all of the layers to produce systems with highly specific applications.119–121 Other Aspects of Drug-Loaded Electrospun Nanofibers All of the examples of drug-loaded nanofibers presented until now in this review rely on solvent-based methodologies to suspend therapeutic agents within solutions that are then transformed into nanofibers by the electrospinning process. Melt electrospinning is another potential technique that could be employed to include drug formulations within nanofibers. In fact, Nagy et al. demonstrated such a process when this research group loaded carvedilol into nanofibers composed of EUDRAGIT® E.122 A solvent-based methodology was compared with melt-based electrospinning, and one finding was that the structure of carvedilol was retained in its amorphous form using both methods. The primary advantage of employing the melt-based technique developed from the lack of a solvent, thus simplifying the process and saving the cost and difficulty of dealing with various solvents. Also, the high surface area of the nanofibers permitted more rapid release of carvedilol into solution when compared with the bulk, crystalline form of the drug. An extensive review of melt electrospinning was published recently by Brown et al., exploring many general and specific applications of melt electrospinning techniques.123 To date, there have not been many documented attempts to include therapeutics within nanofibers fabricated using melt electrospinning. The process offers a unique concept and approach for accomplishing such an advance as it removes the requirement of a solvent, but it introduces new challenges as well. Elevated temperatures necessary for the melt electrospinning process could result, for example, in complete denaturation or deactivation of certain proteins and growth factors. For compounds that retain their activity after exposure to high temperature, melt electrospinning could be a viable option for developing nanofiber-based drug-delivery systems. Additionally, the literature is limited with respect to reports that describe the incorporation of solution-based, coaxial electrospinning methodology into melt-based electrospinning technology. McCann et al. detailed production of phase-change nanofibers consisting of a composite sheath and long-chain hydrocarbon cores that demonstrated the capability of absorbing significant quantities of thermal energy.124 Li et al. generated coaxial nanofibers using a melt-based system as well.125 The core of the nanofibers contained thermochromic compounds, which resulted in a color change to the nanofibers in response to increases or decreases in temperature. These reports could offer insight into developing coaxial, melt electrospinning systems that include growth factors and other therapeutic agents within core–shell nanofibers. Examples have been noted in the previous paragraphs that only touch on the use of drug/growth factor-loaded nanofibers as simple, flat sheets of material. Flat materials provide some data on cellular interaction with nanofibrous sheets, but they lack the capability of producing complex, three-dimensional structures. Attempts to produce electrospun materials with greater three-dimensional complexities have met with some success in recent years. Controlled deposition of nanofibers affords the possibility of generating patterned structures in two dimensions, which could be layered to form more complex three-dimensional structures.123 However, the simplest examples of these experiments are seen in a number of reports detailing methods of forming small tubes composed of electrospun nanofibers with the intent of utilizing them to tissue engineer blood vessels.19,126–129 In the 1970s, Annis et al. detailed the creation of vascular prostheses of electrospun polyurethane nanofibers that were implanted into minipigs for up to a year.19 In addition to significant tissue regeneration, they noted that the mechanical properties of the synthetic vascular graft prevented formation of aneurysms. Recent attempts to create synthetic blood vessels highlight multilayering techniques as well as incorporation of natural polymers into the nanofibrous structure.126–128 Three-dimensional nanofibrous scaffolds Expansion of nanofibers to more complex three-dimensional structures is problematic. In addition to blood vessels, Gaudio et al. demonstrated fabrication of a trileaflet heart valve composed of PCL nanofibers.129 While the overall structure of the valve was three-dimensional, it was still composed of a two-dimensional sheet of fibers. Shim et al. showed it is possible to modify sheets of nanofibers post-production to expand them into three dimensions.130 Specifically, this group used a small metal comb to pull the nanofibers apart to decrease the overall nanofiber density and increase porosity of the PLLA scaffolds. Blakeney et al. approached the dimensionality issue by modifying the nanofiber collector, altering the structure of a PCL nanofiber scaffold as the nanofibers were deposited.131 The scaffolds in this instance were structured similar to a cotton ball and were called Focused, Low density, and Uncompressed nanoFibrous (FLUF) mesh. Jha et al. produced three-dimensional, aligned nanofiber scaffolds for nerve regeneration using a two-pole air gap technique, another modification to the nanofiber collection method.132 Cai et al. manipulated the conductivity of the polymer solution to create three-dimensional nanofiber structures.133 Hybrid scaffolds were also produced, incorporating electrospun nanofibers as a thin coating over the surfaces of pre-existing materials.134 An interesting extension of these approaches would incorporate drug-loaded nanofibers into more intricate, three-dimensional structures. Such systems would increase the complexity and potentially complicate determination of therapeutic agent loading and release in some instances, but the benefits may outweigh these issues for certain cases. Concluding Remarks Ultimately, the intended application and its requisite components drive the development of electrospinning methods to enclose, protect, and/or control release rates for a particular drug delivery system or other purpose. This review has noted many such approaches. Over a century has passed since Boys wrote, “Fibres spun by the electrical method are so brittle that they do not seem to be of any practical use.”1 However, that statement was a bit premature as new applications for electrospun nanofibers are being discovered constantly in many research institutions around the globe. Author Disclosure Statement No competing financial interests exist. Abbreviations Used bFGFbasic fibroblast growth factor HAhyaluronic acid HFIPhexafluoroisopropanol NGCnerve guidance conduit NGFnerve growth factor PCLpolycaprolactone PDLLApoly-d,l-lactic acid PELApolyethylene glycol/poly-dl-lactide PEOpolyethylene oxide PGApolyglycolic acid PGSpoly(glycerol sebacate) PLApolylactic acid PLGApoly(lactic-co-glycolic acid) PLLApoly-l-lactic acid SFsilk fibroin VEGFvascular endothelial growth factor ==== Refs References 1 Boys CV On the production, properties, and some suggested uses of the finest threads . Proc Phys Soc Lond. 1887 ;9 :8 –19 2 Zeleny J The electrical discharge from liquid points, and a hydrostatic method of measuring the electric intensity at their surfaces . Phys Rev. 1914 ;3 :69 –91 3 Zeleny J Instability of electrified liquid surfaces . Phys Rev. 1917 ;10 :1 –6 4 Cooley JF Apparatus for electrically dispersing fluids . US Patent No. 692,631, 1902 5 Morton WJ Method of dispersing fluids . US Patent No. 705,691, 1902 6 Formhals A Apparatus for producing artificial filaments from material such as cellulose acetate . US Patent No. 1,975,504, 1934 7 Formhals A Production of artificial fibers . US Patent No. 2,077,373, 1937 8 Formhals A Artificial fiber construction . US Patent No. 2,109,333, 1938 9 Formhals A Method and apparatus for the production of fibers . US Patent No. 2,116,942, 1938 10 Formhals A Method and apparatus for the production of fibers . US Patent No. 2,123,992, 1938 11 Formhals A Method for producing artificial fibers . US Patent No. 2,158,415, 1939 12 Formhals A Method and apparatus for the production of artificial fibers . US Patent No. 2,158,416, 1939 13 Formhals A Method and apparatus for spinning . US Patent No. 2,160,962, 1939 14 Formhals A Artificial thread and method of producing the same . US Patent No. 2,187,306, 1940 15 Formhals A Production of artificial fibers from fiber forming liquids . US Patent No. 2,323,025, 1943 16 Formhals A Method and apparatus for spinning . US Patent No. 2,349,950, 1944 17 Taylor G Disintegration of water drops in an electric field . Proc R Soc Lond A Math Phys Sci. 1964 ;280 :383 –397 18 Taylor G Electrically driven jets . Proc R Soc Lond A Math Phys Sci. 1969 ;313 :453 –475 19 Annis D , Bornat A , Edwards RO , et al. An elastomeric vascular prosthesis . Trans Am Soc Artif Intern Organs. 1978 ;24 :209 –214 716001 20 Reneker DH , Yarin RL Electrospinning jets and polymer nanofibers . Polymer. 2008 ;40 :2387 –2425 21 Reneker DH , Yarin AL , Fong H , et al. Bending instability of electrically charged liquid jets of polymer solutions in electrospinning . J Appl Phys. 2000 ;87 :4531 –4547 22 Reneker DH , Yarin AL , Zussman E , et al. Electrospinning of nanofibers from polymer solutions and melts . Adv Appl Mech. 2007 ;41 :43 –195 23 Koombhongse S , Liu W , Reneker DH Flat polymer ribbons and other shapes by electrospinning . J Polym Sci B Polym Phys. 2001 ;39 :2598 –2606 24 Thompson CJ , Chase GG , Yarin AL , et al. Effects of parameters on nanofiber diameter determined from electrospinning model . Polymer. 2007 ;48 :6913 –6922 25 Zussman E , Theron A , Yarin AL Formation of nanofiber crossbars in electrospinning . Appl Phys Lett. 2003 ;82 :973 –975 26 Doshi J , Reneker DH Electrospinning process and applications of electrospun fibers . J Electrostat. 1995 ;35 :151 –160 27 Tan S , Huang X , Wu B Some fascinating phenomena in electrospinning processes and applications of electrospun fibers . Polym Int. 2007 ;56 :1330 –1339 28 Feng JJ The stretching of an electrified non-Newtonian jet: a model for electrospinning . Phys Fluids. 2002 ;14 :3912 –3926 29 Lukas D , Sarkar A , Pokorny P Self-organization of jets in electrospinning from free liquid surface: a generalized approach . J Appl Phys. 2008 ;103 :084308 30 Nurfaizey AH , Stanger J , Tucker N , et al. Manipulation of electrospun fibres in flight: the principle of superposition of electric fields as a control method . J Mater Sci. 2012 ;47 :1156 –1163 31 Vaquette C , Cooper-White J The use of an electrostatic lens to enhance the efficiency of the electrospinning process . Cell Tissue Res. 2012 ;347 :815 –826 22287045 32 Townsend-Nicholson A , Jayasinghe SN Bio-protocols for directly forming active encapsulations containing living primary cells . Soft Matter. 2008 ;4 :1219 –1229 33 Mitchell SB , Sanders JE A unique device for controlled electrospinning . J Biomed Mater Res. 2006 ;78A :110 –120 34 Yang Y , Xia T , Zhi W , et al. Promotion of skin regeneration in diabetic rats by electrospun core-sheath fibers loaded with basic fibroblast growth factor . Biomaterials. 2011 ;32 :4243 –4254 21402405 35 Yan L , Si S , Chen Y , et al. Electrospun in-situ hybrid polyurethane/nano-TiO2 as wound dressings . Fiber Polym. 2011 ;12 :207 –213 36 Silva SY , Rueda LC , Lopez M , et al. Double blind, randomized controlled trial, to evaluate the effectiveness of a controlled nitric oxide releasing patch versus meglumine antimoniate in the treatment of cutaneous leishmaniasis [NCT00317629] . Trials. 2006 ;7 :14 –24 16700912 37 Qin XH , Wang SY Filtration properties of electrospinning nanofibers . J Appl Polym Sci. 2006 ;102 :1285 –1290 38 Obendorf SK , Lee S Use of electrospun nanofiber web for protective textile materials as barriers to liquid penetration . Text Res J. 2007 ;77 :696 –702 39 Zheng Y , Wang J , Yao P Formaldehyde sensing properties of electrospun NiO-doped SnO2 nanofibers . Sensor Actuat B-Chem. 2011 ;156 :723 –730 40 Song JH , Kim HE , Kim HW Electrospun fibrous web of collagen-apatite precipitated nanocomposite for bone regeneration . J Mater Sci Mater Med. 2008 ;19 :2925 –2932 18360796 41 Meng ZX , Wang YS , Ma C , et al. Electrospinning of PLGA/gelatin randomly-oriented and aligned nanofibers as potential scaffold in tissue engineering . Mat Sci Eng C. 2010 ;30 :1204 –1210 42 Xu X , Chen X , Ma P , et al. The release behavior of doxorubicin hydrochloride from medicated fibers prepared by emulsion-electrospinning . Eur J Pharm Biopharm. 2008 ;70 :165 –170 18472256 43 Xiaoqianga L , Yana S , Shuipinga L , et al. Encapsulation of proteins in poly(L-lactide-co-caprolactone) fibers by emulsion electrospinning . Colloid Surface B. 2010 ;75 :418 –424 44 Smith TW , Haber E Digoxin intoxication: the relationship of clinical presentation to serum digoxin concentration . J Clin Invest. 1970 ;49 :2379 –2386 45 Hirose M , Tachibana A , Tanabe T Recombinant human serum albumin hydrogel as a novel drug delivery vehicle . Mat Sci Eng C. 2010 ;30 :664 –669 46 Luo Y , Kirker KR , Prestwich GD Cross-linked hyaluronic acid hydrogel films: new biomaterials for drug delivery . J Control Release. 2000 ;69 :169 –184 11018555 47 Vinogradov SV , Bronich TK , Kabanov AV Nanosized cationic hydrogels for drug delivery: preparation, properties, and interactions with cells . Adv Drug Deliv Rev. 2002 ;54 :135 –147 11755709 48 Kolhe P , Khandare J , Pillai O , et al. Preparation, cellular transport, and activity of polyamidoamine-based dendritic nanodevices with a high drug payload . Biomaterials. 2006 ;27 :660 –669 16054211 49 Namazi H , Adeli M Dendrimers of citric acid and poly (ethylene glycol) as the new drug-delivery agents . Biomaterials. 2005 ;26 :1175 –1183 15451637 50 Tomalia DA , Reyna LA , Svenson S Dendrimers as multi-purpose nanodevices for oncology drug delivery and diagnostic imaging . Biochem Soc Trans. 2007 ;35 :61 –67 17233602 51 Svenson S Dendrimers as versatile platform in drug delivery applications . Eur J Pharm Biopharm. 2009 ;71 :445 –462 18976707 52 Geers B , Lentacker I , Sanders NN , et al. Self-assembled liposome-loaded microbubbles: the missing link for safe and efficient ultrasound triggered drug delivery . J Control Release. 2011 ;152 :249 –256 21362448 53 Gregoriadis G Drug entrapment in liposomes . FEBS Lett. 1973 ;36 :292 –296 4763309 54 Blume G , Cevc G Liposomes for the sustained drug release in vivo . Biochim Biophys Acta. 1990 ;1029 :91 –97 2223816 55 Allen TM Liposomal drug delivery . Curr Opin Colloid Interface Sci. 1996 ;1 :645 –651 56 Grayson ACR , Shawgo RS , Li Y , et al. Electronic MEMS for triggered delivery . Adv Drug Deliver Rev. 2004 ;56 :173 –184 57 Elman NM , Ho Duc HL , Cima MJ An implantable MEMS drug delivery device for rapid delivery in ambulatory emergency care . Biomed Microdevices. 2009 ;11 :625 –631 19169826 58 Park JH , Allen MG , Prausnitz MR Biodegradable polymer microneedles: fabrication, mechanics, and transdermal drug delivery . J Control Release. 2005 ;104 :51 –66 15866334 59 Lo R , Li PY , Saati S , et al. A passive MEMS drug delivery pump for treatment of ocular diseases . Biomed Microdevices. 2009 ;11 :959 –970 19396548 60 Liao Y , Zhang L , Gao Y , et al. Preparation, characterization, and encapsulation/release studies of a composite nanofiber mat electrospun from an emulsion containing poly(lactic-co-glycolic acid) . Polymer. 2008 ;49 :5294 –5299 19907634 61 Buttafoco L , Kolkman NG , Engbers-Buijtenhuijis P , et al. Electrospinning of collagen and elastin for tissue engineering applications . Biomaterials. 2006 ;27 :724 –734 16111744 62 Geng X , Kwon OH , Jang J Electrospinning of chitosan dissolved in concentrated acetic acid solution . Biomaterials. 2005 ;26 :5427 –5432 15860199 63 Bhattarai N , Edmondson D , Veiseh O , et al. Electrospun chitosan-based nanofibers and their cellular compatibility . Biomaterials. 2005 ;26 :6176 –6184 15885770 64 Suh JK , Matthew HW Application of chitosan-based polysaccharide biomaterials in cartilage tissue engineering: a review . Biomaterials. 2000 ;21 :2589 –2598 11071608 65 Li Z , Ramay HR , Hauch KD , et al. Chitosan-alginate hybrid scaffolds for bone tissue engineering . Biomaterials. 2005 ;26 :3919 –3928 15626439 66 Bonino CA , Krebs MD , Saquing CD , et al. Electrospinning alginate-based nanofibers: from blends to crosslinked low molecular weight alginate-only systems . Carbohydr Polym. 2011 ;85 :111 –119 67 Laurent TC , Fraser JRE Hyaluronan . FASEB J. 1992 ;6 :2397 –2404 1563592 68 Brenner EK , Schiffman JD , Thompson EA , et al. Electrospinning of hyaluronic acid nanofibers from aqueous ammonium solutions . Carbohydr Polym. 2012 ;87 :926 –929 69 Badami AS , Kreke MR , Thompson MS , et al. Effect of fiber diameter on spreading, proliferation, and differentiation of osteoblastic cells on electrospun poly(lactic acid) substrates . Biomaterials. 2006 ;27 :596 –606 16023716 70 Xu CY , Inai R , Kotaki M , et al. Aligned biodegradable nanofibrous structure: a potential scaffold for blood vessel engineering . Biomaterials. 2004 ;25 :877 –886 14609676 71 Liao S , Li B , Ma Z , et al. Biomimetic electrospun nanofibers for tissue regeneration . Biomed Mater. 2006 ;1 :R45 –R53 18458387 72 Yang F , Murugan R , Wang S , et al. Electrospinning of nano/micro scale poly(L-lactic acid) aligned fibers and their potential in tissue engineering . Biomaterials. 2005 ;26 :2603 –2610 15585263 73 Wutticharoenmongkol P , Sanchavanakit N , Pavasant P , et al. Preparation and characterization of novel bone scaffolds based on electrospun polycaprolactone fibers filled with nanoparticles . Macromol Biosci. 2006 ;6 :70 –77 16374772 74 Owen TA , Aronow M , Shalhoub V , et al. Progressive development of the rat osteoblast phenotype in vitro: reciprocal relationships in expression of genes associated with osteoblast proliferation and differentiation during formation of the bone extracellular matrix . J Cell Physiol. 1990 ;143 :420 –430 1694181 75 Yarin RL Coaxial electrospinning and emulsion electrospinning of core-shell fibers . Polym Adv Technol. 2011 ;22 :310 –317 76 Li X , Zhang H , Li H , et al. Encapsulation of proteinase K in PELA ultrafine fibers by emulsion electrospinning: preparation and in vitro evaluation . Colloid Polym Sci. 2010 ;288 :1113 –1119 77 Yang Y , Li X , Qi M , et al. Release pattern and structural integrity of lysozyme encapsulated in core-sheath structured poly(DL-lactide) ultrafine fibers prepared by emulsion electrospinning . Eur J Pharm Biopharm. 2008 ;69 :106 –116 18078743 78 Maretschek S , Greiner A , Kissel T Electrospun biodegradable nanofiber nonwovens for controlled release of proteins . J Control Release. 2008 ;127 :180 –187 18314212 79 Liu S , Su Y , Chen Y Fabrication, surface properties, and protein encapsulation/release studies of electrospun gelatin nanofibers . J Biomater Sci Polym Ed. 2011 ;22 :945 –955 20566066 80 Agarwal S , Greiner A On the way to clean and safe electrospinning—green electrospinning: emulsion and suspension electrospinning . Polym Adv Technol. 2011 ;22 :372 –378 81 Li L , Li H , Qian Y , et al. Electrospun poly(ɛ-caprolactone)/silk fibroin core-sheath nanofibers and their potential applications in tissue engineering and drug release . Int J Biol Macromol. 2011 ;49 :223 –232 21565216 82 Bhardwaj N , Kundu SC Chondrogenic differentiation of rat MSCs on porous scaffolds of silk fibroin/chitosan blends . Biomaterials. 2012 ;33 :2848 –2857 22261099 83 Kambe Y , Yamamoto K , Kojima K , et al. Effects of RGDS sequence genetically interfused in the silk fibroin light chain protein on chondrocyte adhesion and cartilage synthesis . Biomaterials. 2010 ;31 :7503 –7511 20643479 84 Li L , Qian Y , Jiang C , et al. The use of hyaluronan to regulate protein adsorption and cell infiltration in nanofibrous scaffolds . Biomaterials. 2012 ;33 :3428 –3445 22300743 85 Dong B , Arnoult O , Smith ME , et al. Electrospinning of collagen nanofiber scaffolds from benign solvents . Macromol Rapid Comm. 2009 ;30 :539 –542 86 Meng L , Arnoult O , Smith M , et al. Electrospinning of in situ crosslinked collagen nanofibers . J Mater Chem. 2012 ;22 :19412 –19417 87 Chen R , Huang L , Ki Q , et al. Preparation and characterization of coaxial electrospun thermoplastic polyurethane/collagen compound nanofibers for tissue engineering applications . Colloid Surface B. 2010 ;79 :315 –325 88 Jiang H , Hu Y , Zhao P , et al. Modulation of protein release from biodegradable core-shell structured fibers prepared by coaxial electrospinning . J Biomed Mater Res B. 2006 ;79B :50 –57 89 Gao J , Crapo P , Nerem R , et al. Co-expression of elastin and collagen leads to highly compliant engineered blood vessels . J Biomed Mater Res A. 2008 ;85A :1120 –1128 18412137 90 Yi F , LaVan DA Poly(glycerol sebacate) nanofiber scaffolds by core/shell electrospinning . Macromol Biosci. 2008 ;8 :803 –806 18504802 91 Wang CY , Liu JJ , Fan CY , et al. The effect of aligned core-shell nanofibres delivering NGF on the promotion of sciatic nerve regeneration . J Biomater Sci Polym Ed. 2012 ;23 :167 –184 21192836 92 Jia X , Zhao C , Li P , et al. Sustained release of VEGF by coaxial electrospun dextran/PLGA fibrous membranes in vascular tissue engineering . J Biomater Sci Polym Ed. 2011 ;22 :1811 –1827 20961491 93 Liao IC , Leong KW Efficacy of engineered FVIII-producing skeletal muscle enhanced by growth factor-releasing co-axial electrospun fibers . Biomaterials. 2011 ;32 :1669 –1677 21084118 94 Zhao Y , Cao X , Jiang L Bio-mimic multichannel microtubes by a facile method . J Am Chem Soc. 2007 ;129 :764 –765 17243804 95 Chen H , Wang N , Di J , et al. Nanowire-in-microtube structured core/shell fibers via multifluidic coaxial electrospinning . Langmuir. 2010 ;26 :11291 –11296 20337483 96 Sharma U , Pham Q , Marini J , et al. Electrospinning process for manufacture of multi-layered structures . US Patent No. 20120193836, 2012 97 Su Y , Su Q , Liu W , et al. Dual-drug encapsulation and release from core-shell nanofibers . J Biomater Sci Polym Ed. 2012 ;23 :861 –871 21418751 98 Asahina I , Sampath TK , Hauschka PV Human osteogenic protein-1 induces chondroblastic, osteoblastic, and/or adipocytic differentiation of clonal murine target cells . Exp Cell Res. 1996 ;222 :38 –47 8549671 99 Makino T , Jinnin M , Muchemwa FC , et al. Basic fibroblast growth factor stimulates proliferation of human dermal fibroblasts via the ERK ½ and JNK pathways . Br J Dermatol. 2010 ;162 :717 –723 19995368 100 Yuan Q , Kubo T , Doi K , et al. Effect of combined application of bFGF and inorganic polyphosphate on bioactivities of osteoblasts and initial bone regeneration . Acta Biomater. 2009 ;5 :1716 –1724 19251495 101 Okuda T , Tominaga K , Kidoaki S Time-programmed dual release formulation by multilayered drug-loaded nanofiber meshes . J Control Release. 2010 ;143 :258 –264 20074599 102 Pham QP , Sharma U , Mikos AG Electrospun poly(ɛ-caprolactone) microfiber and multilayer nanofiber/microfiber scaffolds: characterization of scaffolds and measurement of cellular infiltration . Biomacromolecules. 2006 ;7 :2796 –2805 17025355 103 Min BM , Lee G , Kim SH , et al. Electrospinning of silk fibroin nanofibers and its effect on the adhesion and spreading of normal human keratinocytes and fibroblasts in vitro . Biomaterials. 2004 ;25 :1289 –1297 14643603 104 Woodfield TBF , van Blitterswijk CA , de Wijn J , et al. Polymer scaffolds fabricated with pore-size gradients as a model for studying the zonal organization within tissue-engineered cartilage constructs . Tissue Eng. 2005 ;11 :1297 –1311 16259586 105 Yang S , Leong KF , Du Z , et al. The design of scaffolds for use in tissue engineering. Part I. Traditional factors . Tissue Eng. 2001 ;7 :679 –689 11749726 106 Liu S , Li X , Su Y , et al. Study of electrospun PLLACL nanofibrous mats for drug delivery systems . Mater Sci Forum. 2009 ;610 :1319 –1322 107 Ahuja N , Katare OP , Singh B Studies on dissolution enhancement and mathematical modeling of drug release of a poorly water-soluble drug using water-soluble carriers . Eur J Pharm Biopharm. 2007 ;65 :26 –38 16962750 108 Nakamura J , Nakajima N , Matsumura K , et al. Water-soluble taxol conjugates with dextran and targets tumor cells by folic acid immobilization . Anticancer Res. 2010 ;30 :903 –910 20393013 109 Patravale VB , Date AA , Kulkarni RM Nanosuspensions: a promising drug delivery strategy . J Pharm Pharmacol. 2004 ;56 :827 –840 15233860 110 Baba K , Pudavar HE , Roy I , et al. New method for delivering a hydrophobic drug for photodynamic therapy using pure nanocrystal form of the drug . Mol Pharm. 2007 ;4 :289 –297 17266331 111 Zhang X , Xia Q , Gu N Preparation of all-trans retinoic acid nanosuspensions using a modified precipitation method . Drug Dev Ind Pharm. 2006 ;32 :857 –863 16908423 112 Pouton CW Formulation of poorly water-soluble drugs for oral administration: physicochemical and physiological issues and the lipid formulation classification system . Eur J Pharm Sci. 2006 ;29 :278 –287 16815001 113 Dong B , Smith ME , Wnek GE Encapsulation of multiple biological compounds within a single electrospun fiber . Small. 2009 ;5 :1508 –1512 19384883 114 Beck-Broichsitter M , Thieme M , Nguyen J , et al. Novel ‘nano in nano’ composites for sustained drug delivery: biodegradable nanoparticles encapsulated into nanofiber non-wovens . Macromol Biosci. 2010 ;10 :1527 –1535 20824691 115 Jo E , Lee S , Kim KT , et al. Core-sheath nanofibers containing colloidal arrays in the core for programmable multi-agent delivery . Adv Mater. 2009 ;21 :968 –972 116 Hamlett CAE , Jayasinghe SN , Preece JA Electrospinning nanosuspensions loaded with passivated Au nanoparticles . Tetrahedron. 2008 ;64 :8476 –8483 117 Zakaria SM , Zein SHS , Othman MR , et al. Hydroxyapatite nanoparticles: electrospinning and calcination of hydroxyapatite/polyvinyl butyral nanofibers and growth kinetics . J Biomed Mater Res A. 2013 ;101 :1977 –1985 23225849 118 Yang Q , Sui G , Shi YZ , et al. Osteocompatibility characterization of polyacrylonitrile carbon nanofibers containing bioactive glass nanoparticles . Carbon. 2013 ;56 :288 –295 119 Liu W , Ni C , Chase DB , et al. Preparation of multilayer biodegradable nanofibers by triaxial electrospinning . ACS Macro Lett. 2013 ;2 :466 –468 120 Yu DG , Li XY , Wang X , et al. Nanofibers fabricated using triaxial electrospinning as zero order drug delivery systems . ACS Appl Mater Interfaces. 2015 ;7 :18891 –18897 26244640 121 Han D , Steckl AJ Triaxial electrospun nanofiber membranes for controlled dual release of functional molecules . ACS Appl Mater Interfaces. 2013 ;5 :8241 –8245 23924226 122 Nagy ZK , Balogh A , Dravavolgyi G , et al. Solvent-free melt electrospinning for preparation of fast dissolving drug delivery system and comparison with solvent-based electrospun and melt extruded systems . J Pharm Sci. 2013 ;102 :508 –517 23161110 123 Brown TD , Dalton PD , Hutmacher DW Melt electrospinning today: an opportune time for an emerging polymer . Prog Polym Sci. 2016 ;56 :116 –166 124 McCann JT , Marquez M , Xia Y Melt coaxial electrospinning: a versatile method for the encapsulation of solid materials and fabrication of phase change nanofibers . Nano Lett. 2006 ;6 :2868 –2872 17163721 125 Li F , Zhao Y , Wang S , et al. Thermochromic core–shell nanofibers fabricated by melt coaxial electrospinning . J Appl Polym Sci. 2009 ;112 :269 –274 126 Ju YM , Choi JS , Atala A , et al. Bilayered scaffold for engineering cellularized blood vessels . Biomaterials. 2010 ;31 :4313 –4321 20188414 127 Wang S , Zhang Y , Yin G , et al. Electrospun polylactide/silk fibroin-gelatin composite tubular scaffolds for small-diameter tissue engineering blood vessels . J Appl Polym Sci. 2009 ;113 :2675 –2682 128 Vaz CM , van Tuijl S , Bouten CVC , et al. Design of scaffolds for blood vessel tissue engineering using a multi-layering electrospinning technique . Acta Biomater. 2005 ;1 :575 –582 16701837 129 Gaudio CD , Bianco A , Grigioni M Electrospun bioresorbable trileaflet heart valve prosthesis for tissue engineering: in vitro functional assessment of a pulmonary cardiac valve design . Ann Ist Super Sanita. 2007 ;44 :178 –186 18660567 130 Shim IK , Jung MR , Kim KH , et al. Novel three-dimensional scaffolds of poly(L-lactic acid) microfibers using electrospinning and mechanical expansion: fabrication and bone regeneration . J Biomed Mater Res B Appl Biomater. 2010 ;95B :150 –160 20725960 131 Blakeney BA , Tambralli A , Anderson JM , et al. Cell infiltration and growth in a low density, uncompressed three-dimensional electrospun nanofibrous scaffold . Biomaterials. 2011 ;32 :1583 –1590 21112625 132 Jha BS , Colello RJ , Bowman JR , et al. Two pole air gap electrospinning: fabrications of highly aligned, three-dimensional scaffolds for nerve reconstruction . Acta Biomater. 2011 ;7 :203 –215 20727992 133 Cai S , Xu H , Jiang Q , et al. Novel 3D electrospun scaffolds with fibers oriented randomly and evenly in three dimensions to closely mimic the unique architectures of extracellular matrices in soft tissues: fabrication and mechanism study . Langmuir. 2013 ;29 :2311 –2318 23390966 134 McClellan P , Landis WJ Three-dimensional coating of nanofibers on surfaces of poorly conductive objects . Polymer. 2013 ;54 :6702 –6708 Cite this article as: McClellan P, Landis WJ (2016) Recent applications of coaxial and emulsion electrospinning methods in the field of tissue engineering, BioResearch Open Access 5:1, 212–227, DOI: 10.1089/biores.2016.0022.

PMC005xxxxxx/PMC5003016.txt

==== Front Acoust AustAcoust AustAcoustics Australia0814-6039Springer Singapore Singapore 275826154010.1007/s40857-015-0040-5Review PaperThe Efficacy of Anti-vibration Gloves Hewitt Sue +44 1298 21 8386susan.hewitt@hsl.gsi.gov.uk 1Dong Ren 2McDowell Tom 2Welcome Daniel 21 grid.9984.cHealth and Safety Executive, Harpur Hill, Buxton, SK17 9JN UK 2 grid.416809.20000000404230663Engineering & Control Technology Branch, Health Effects Laboratory Division, National Institute for Occupational Safety and Health, Morgantown, WV 26505 USA 3 2 2016 3 2 2016 2016 44 1 121 127 4 11 2015 23 12 2015 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Anyone seeking to control the risks from vibration transmitted to the hands and arms may contemplate the use of anti-vibration gloves. To make an informed decision about any type of personal protective equipment, it is necessary to have performance data that allow the degree of protection to be estimated. The information provided with an anti-vibration glove may not be easy to understand without some background knowledge of how gloves are tested and does not provide any clear route for estimating likely protection. Some of the factors that influence the potential efficacy of an anti-vibration glove include how risks from hand–arm vibration exposure are assessed, how the standard test for a glove is carried out, the frequency range and direction of the vibration for which protection is sought, how much hand contact force or pressure is applied and the physical limitations due to glove material and construction. This paper reviews some of the background issues that are useful for potential purchasers of anti-vibration gloves. Ultimately, anti-vibration gloves cannot be relied on to provide sufficient and consistent protection to the wearer and before their use is contemplated all other available means of vibration control ought first to be implemented. Keywords Anti-vibration glovesHand–arm vibrationHand–arm vibration syndromePersonal protective equipmentissue-copyright-statement© Australian Acoustical Society 2016 ==== Body Introduction The connection between use of vibrating power tools and the associated health effects referred to as hand–arm vibration syndrome (HAVS) has been known for around a century. In the modern workplace, health effects associated with power tool use are still commonly reported and there are hundreds of new cases reported every year in the UK [1]. When attempting to manage exposure to hand–arm vibration in the workplace, and having exhausted all the other possible approaches to managing the problem, the question of personal protective equipment (PPE) will inevitably arise. Anti-vibration gloves are available, which are typically made from materials such as resilient gel, foam or rubber-like material or an array of air bladders. This paper considers the issues that surround the selection and use of anti-vibration gloves as PPE for hand–arm vibration. In the UK, the Health and Safety Executive produced guidance in 2005 on the control of risks from hand–arm vibration [2]. Part six of the guidance contains technical information on anti-vibration gloves and explains the main points to be considered very succinctly. The guidance concludes that employers should not assume that anti-vibration gloves reduce vibration exposures unless specific data are available for the particular combination of glove and tool used. This paper provides further background and updated information; however, the guidance remains unchanged. To understand the technical considerations relating to the prospective use of anti-vibration gloves, it is necessary to know how exposure to hand–arm vibration is assessed according to current international standards and how a vibration-reducing glove is tested before it can be marketed as an anti-vibration glove in Europe and USA. Assessment of Hand–Arm Vibration Exposure The international standard for measurement and assessment of exposure to hand–arm vibration is ISO 5349:2001, parts 1 and 2 [3, 4]. These standards define how the vibration to which an individual is exposed is measured and evaluated in terms of the frequency-weighted vibration total value at or near to the gripping zone. The hand–arm frequency weighting defined in ISO 5349-1:2001 is shown in Fig. 1.Fig. 1 ISO 5349-1:2001 [3] hand–arm frequency weighting, Wh The hand–arm frequency weighting gives most weight to low frequencies between 6.3 and 25 Hz. This is relatively low compared with most types of power tool, which typically have a main operating frequency in the range of 25–150 Hz [5]. The frequency-weighted vibration total value is a single figure relating to the vibration on the surface of the machine that the operator is in contact with. It combines the measures of the vibration in three orthogonal directions; axes: x,y and z. Figure 2 shows the x,y and z axes as described in a recent paper [6] as they relate to the axes used for assessment of the performance of an anti-vibration glove. The hand–arm frequency weighting is applied to the vibration in each of the three axes, before they are combined to give the vibration total value. When assessing vibration exposure according to ISO 5349-1:2001, the vibration total value is combined with information on the duration of the exposure to vibration to give a daily vibration exposure, A(8),  expressed in units of m/s2. Fig. 2 Illustration of x,y and z axes used for testing according to the thenar region-based biodynamic coordinate system from Dong et al. [6] Assessment of Anti-vibration Glove Performance The current international standard that should be applied to a glove before it can be marketed as an anti-vibration glove is ISO 10819:2013 [7]. The test described in this standard involves applying a defined signal to a vibrating handle and then measuring how much of that vibration is transmitted through the glove to the palm of the wearer. To achieve this, the vibration is measured simultaneously on the surface of the handle and in the palm of the hand using an adaptor. This enables the vibration transmitted through the glove to be calculated. The test uses a band-limited random vibration signal. The vibration magnitude used for the test is defined in all the one-third octave frequency bands from 25 to 1600 Hz. This range is selected based on the possible effective frequency of anti-vibration gloves and the frequency range of concern for hand–arm vibration exposure defined in ISO 5349-1:2001. The values that are produced by application of the test are referred to as transmissibilities. The transmissibility values are calculated using hand–arm frequency-weighted vibration magnitudes to determine whether the glove reduces the vibration that is transmitted to the wearer. When the overall results of the glove transmissibility measurements are calculated, they are expressed as two values: T¯(M), the average result from the 25 Hz one-third octave band to the 200 Hz one-third octave band and T¯(H), the average result from the 200 Hz one-third octave band to the 1250 Hz one-third octave band. A transmissibility of 1.0 means that all of the vibration is transmitted through the glove material to the wearer. If the transmissibility is less than 1.0, it indicates that the glove is reducing the amount of vibration that is being transmitted. If the transmissibility is more than 1.0, it indicates that the glove is amplifying the vibration.Fig. 3 Example transmissibilities of an air bladder glove at the palm (from Dong and colleagues [9]) To be CE marked and marketed as an anti-vibration glove in Europe, a glove must first satisfy both the criteria for the transmissibility set in ISO 10819:2013: T¯(M)≤0.90,andT¯(H)≤0.60. These criteria mean that the glove must provide on average at least 10 % reduction in the frequency-weighted vibration between 25 and 200 Hz and must provide on average at least a 40 % reduction in the frequency-weighted vibration between 200 and 1250 Hz. The ISO 10819:2013 standard also specifies the maximum thickness of the material in the palm of an anti-vibration glove and also at the fingers, although the transmissibility of the glove is only tested at the palm. Gloves that satisfy the transmissibility criteria and also satisfy the requirements for thickness of the material can be given a CE mark and sold as an anti-vibration glove in the EU. Factors Affecting the Apparent Performance of Anti-vibration Gloves ISO 5349 Hand–Arm Frequency Weighting Most anti-vibration gloves do not provide much reduction in vibration transmission at frequencies below 25 Hz at the palm of the hand and below 160 Hz at the fingers [8, 9]. A typical anti-vibration glove may in fact cause slight amplification of vibration at low frequencies [8, 9]. Figure 3 shows examples of the transmissibilities of an air bladder glove measured in the x-, y- and z-axes at the palm of the hand [10]. Figure 3 shows that the typical glove transmissibilities are close to 1.0 at low frequencies where the hand–arm frequency weighting is at its peak. The main reductions in transmissibility also depend on the vibration axis and the applied hand forces [10, 11]. The gloves can usually become more effective with the reduction of the applied hand forces [11] but certain hand forces are required to control a tool. At more than 160 Hz, the ISO 5349-1:2001 hand–arm frequency weighting reduces the vibration signal to less than one tenth of its actual value. This, in combination with the fact that most machines have operating frequencies below 160 Hz, makes it difficult for any glove to provide very much reduction of the frequency-weighted vibration for the fingers; the frequency weighting limits the contribution made by the higher frequencies to the overall vibration magnitude. The ISO 5349-1:2001 hand–arm frequency weighting is currently under review [12], although it is unlikely to be changed in the near future. The current hand–arm frequency weighting is a single weighting used to assess vibration exposure for all possible health effects that might be associated with hand–arm vibration, but the origins of the frequency weighting are not related to health. The underlying research was based on equal vibration sensation contours of the entire hand–arm system [13]. As there are many different health effects encompassed by the term HAVS, it is possible that the current hand–arm frequency weighting may be more appropriate for some of these health effects than for others. Evidence from studies of health effects and biodynamic modelling [14, 15] indicates that the current hand–arm weighting may be most suited to health effects in the palm–wrist–arm substructures. Further evidence relating to health effects at the fingers [16, 17] indicates that gloves may be more beneficial than is predicted by the limited reductions in frequency-weighted vibration exposure that gloves provide. Other studies of the neurological health effects of hand–arm vibration exposure [18, 19] also indicate that high frequencies may be more damaging. These findings point to the possibility that the frequency weighting is inadequate for assessing the risk of developing some of the health effects associated with hand–arm vibration exposure. The question of the contribution of higher frequencies to the development of health effects and the issues relating to the hand–arm frequency weighting are discussed in more detail by Hewitt et al. [20]. Ultimately however, because the exact mechanism or mechanisms of damage for vascular and neurological finger symptoms have not been clearly identified, and the exposure–response relationship for HAVS remains ill-defined [21], it is difficult to establish a suitable frequency weighting or weightings to predict the different health effects of vibration. In the absence of any strong evidence to support alternatives, the current frequency weighting is unlikely to be changed at any time in the near future [12] and consequently the technique for assessment of the performance anti-vibration gloves is also unlikely to be changed. Limitations of the Standardised Glove Test Transmissibility of a Glove in Different Directions The frequency-weighted vibration total value is a combination of the vibration measured in the three orthogonal axes as shown in Fig. 2. ISO 10819:2013 only specifies measurement of the performance of a glove in the z-axis with the material acting in compression. An example of the test set-up for z-axis testing is shown in Fig. 4. The biodynamic models developed by Dong et al. [15] demonstrate how the transmission of vibration through the glove is affected by the physical characteristics of the individual components of the finger and the palm–wrist–arm structure. The transmissibility can be very different in shear, which is in a direction through the hand as represented by the y-axis in Fig. 2 and shown in Fig. 5.Fig. 4 A glove test set-up, testing in compression in the z-axis Fig. 5 A glove test set-up, testing in shear in the y-axis Fig. 6 Example transmissibilities of an air bladder glove at the fingers (from Welcome and colleagues [23]) The examples of transmissibilities in the three axes, Tx,Ty and Tz measured at the palm of the hand for an air bladder glove are shown in Fig. 3. The differences at higher frequencies are not only due in part to the difference in the properties and behaviour of the glove material in shear, but are also due to the differences in the biodynamics of the hand–arm system. It is clear then that the transmissibility of an anti-vibration glove is both direction and posture specific. Because the effective mass at the palm is usually the highest along the z-axis, the vibration reduction of a glove is usually most effective in this direction, but because the standard glove test is only applied in this direction, the results will usually overestimate the total effectiveness of the glove. Transmissibility of a Glove at the Fingers The vibration transmissibility of a glove at the fingers has been shown to be much different from the transmissibility at the palm of the hand [22]. ISO 10819:2013 sets requirements for distribution and thickness of the vibration-reducing material at the fingers, for compliance with the standard test criteria. These requirements do not actually increase the glove effectiveness, but they may make production of gloves more difficult [22]. Furthermore, the actual performance of a glove is only measured at the palm. Figure 6 shows an example of air bladder glove transmissibilities measured at the fingers in the y-axis and in the combined x- and z-axes. The transmissibilities for the x- and z-axes are combined because it is difficult to reliably separate the two directions when considering the fingers [22]. The vibration transmissibility of a glove at the fingers has been shown to be generally much higher than at the palm of the hand, meaning that gloves are much less effective at reducing vibration transmitted to the fingers than to the palm of the wearer [23]. This is mainly due to differences in the apparent mass of the fingers compared with the rest of the hand–arm system [23]. Estimates of the protection afforded to the fingers have shown that anti-vibration gloves would actually have little value for reducing finger-transmitted weighted vibration, except in some special cases [20]. Vibration Spectrum from the Tool The performance of an anti-vibration glove will depend on the acceleration spectrum and direction on the power tool handle. A transfer function method has been used to estimate the vibration reduction potential of a glove at the palm and at the fingers for a variety of different power tool spectra [10, 24]. The data show that for the vibration transmissibility at the fingers, any reduction in vibration is not significant for most tools. When the transmissibility at the palm is considered, the performance of the glove is expected to depend heavily on the main operating frequency of the tool [10]. For tools that work at low frequencies, such as sand rammers, very little vibration reduction is predicted. For tools operating at medium frequencies, around 30–50 Hz, such as chipping hammers, the predicted reductions are typically between 5 and 20 %. In one example for an impact drill, which has a large amount of high frequency content in the vibration spectrum for each axis, the glove is predicted to reduce the vibration by more than 30 %. These data are similar to results from an earlier study [8] which showed the estimated range of performance for one type of gel and foam glove was from 3 % amplification when applied to the vibration spectra from an angle grinder with a main operating frequency of 100 Hz, to 30 % reduction for a multi-use sanding tool with a main operating frequency of 315 Hz. These estimates are, however, theoretical and have not been corroborated by measurement. In practice, a valid measurement of the transmissibility of a glove on a real tool is very difficult to achieve [8] due to the influence of factors such as the mounting of the transducers as well as due to changes in grip and push forces affecting measured vibration magnitudes. Design Limitations of Anti-vibration Gloves The effectiveness of an anti-vibration glove depends on both the material properties of the glove and the effective mass of the hand–arm system [23]. While the glove materials can vary substantially, the natural dynamic properties of the hand–arm system cannot be substantially changed. This is one of the major factors that limit the effectiveness of the anti-vibration gloves. The design of anti-vibration gloves is also limited by the need for gloves to be wearable and safe. Thicker, softer materials will be more effective at reducing transmissibility, but increasing softness introduces issues with safe operation and adequate control. Thicker gloves may also result in the need for increased grip force to hold and control the tool, which may also potentially result in operator fatigue [25]. Influences of Varying Forces and Individuals The influence of applied forces is known to affect the amount of vibration transmitted through a glove [11, 26]. When a power tool is used for a real work task, the grip and push forces and working postures may be highly variable across a wide range of forces. The measurements of transmissibilities of anti-vibration gloves according to ISO 10819:2013 are carried out under controlled laboratory conditions. This includes controlling the amount of grip and push force applied during any measurement made as well as the posture adopted. In the standard test conditions in ISO 10819:2013, measurements of transmissibility at the palm are made using a grip force controlled to 30 N and the push force controlled to 50 N. How applicable transmissibility data measured at one specific level of force might actually be for the real work situation has not been well established, but grip and push forces are bound to vary considerably in real work situations. The overall transmissibility of a glove at the palm of the hand has been shown to vary by as much as 20 % from individual to individual even under controlled laboratory test conditions [27]. When transmissibilities are measured, they are typically averaged across operators and this limits their applicability to the general population depending on the number and physical characteristics of those used as test subjects. The number of test subjects required for the ISO 10819:2013 test has been increased from 3 to 5 to take this into account, but even this may not be sufficient in some cases [20]. It is difficult to reach a consensus for further increasing the number of subjects, as the increase will largely increase the cost of the test. Inter-relationship Between Influencing Factors and Prediction of Performance Many factors such as the vibration spectrum of a power tool, the main direction of the vibration, the transmissibility of the glove in that direction, the physical characteristics of the wearer and the posture and amount of force applied by the wearer to the vibrating surface will all be combined to define a level of transmissibility which is specific to that set of circumstances. The tool vibration spectrum and biodynamic responses themselves are also influenced by the hand forces, vibration direction, operating styles, working materials and individuals. Therefore, the factors influencing the assessment, performance and effectiveness of an anti-vibration glove are interrelated and their interactions may be complex. The large number of influencing factors and their interactions make it very difficult to accurately predict or measure the actual individual performance of a glove. They also make anti-vibration gloves unreliable as a form of PPE. Summary and Conclusions There are many factors that influence the measured transmissibility of an anti-vibration glove and the potential that a glove has to provide protection to the wearer. These factors include the effect of different directions and different frequencies of vibration and how they interact, the differences in transmissibility between the palm and the finger, and the variations due to different forces applied to the glove and due to different physical characteristics of the wearers. Anti-vibration gloves can reduce vibration components at very high frequencies (≥500Hz), especially when a low hand coupling force is applied. However, the hand–arm frequency weighting defined in ISO 5349-1:2001 required to evaluate the exposure for risk assessment restricts the measured efficacy of an anti-vibration glove. Other ways of controlling vibration exposure, such as eliminating the need for the exposure, using low-vibration machinery and minimising exposure times are far more likely to be effective and ought first to be adopted. Thicker gloves are more likely to be effective at reducing vibration transmission, but may increase the grip forces needed to safely operate the machine and reduce manual dexterity, so the pros and cons of anti-vibration gloves ought to be carefully balanced if their use is to be considered. This publication and the work it describes were funded by the Health and Safety Executive (HSE) and the National Institute for Occupational Safety and Health (NIOSH). Its contents, including any opinions and/or conclusions expressed, are those of the authors alone and do not necessarily reflect HSE and NIOSH policies and their official position. ==== Refs References 1. Health and Safety Executive: Hand Arm Vibration (HAV) in Great Britain. HSE (2013). http://www.hse.gov.uk/Statistics/causdis/vibration/index.htm. Accessed 19 Oct 2015 2. HSE: Hand-arm vibration. The Control of Vibration at Work Regulations 2005. Health and Safety Executive, HSE Books, Great Britain (2005) 3. International Organization for Standardization: ISO 5349-1:2001 Mechanical Vibration and Shock—Evaluation of Human Exposure to Hand-Transmitted Vibration—Part 1 General Requirements. ISO, Geneva (2001) 4. International Organization for Standardization: ISO 5349-2:2001 Mechanical Vibration— Measurement and Evaluation of Human Exposure to Hand-Transmitted Vibration—Part 2: Practical Guidance for Measurement at the Workplace. ISO, Geneva (2001) 5. Griffin MJ Measurement, evaluation and assessment of occupational exposures to hand-transmitted vibration Occup. Environ. Med. 1997 54 73 89 10.1136/oem.54.2.73 9072014 6. Dong RG Sinsel EW Welcome DE Warren C Xu XS McDowell TW Wu JZ Review and evaluation of hand–arm coordinate systems for measuring vibration exposure, biodynamic responses and hand forces Saf. Health Work 2015 6 159 173 10.1016/j.shaw.2015.05.003 26929824 7. International Organization for Standardization: ISO 10819:2013 Mechanical Vibration and Shock—Hand–Arm Vibration—Measurement and Evaluation of the Vibration Transmissibility of Gloves at the Palm of the Hand. ISO, Geneva (2013) 8. Hewitt, S.M.: Triaxial Measurements of the Performance of Anti-vibration Gloves. HSE Research Report RR795. HSE Books. www.hse.gov.uk/research/rrpdf/rr795.pdf (2010). Accessed 2 Feb 2016 9. McDowell TW Dong RG Welcome DE Warren C Xu XS Vibration reducing gloves: transmissibility at the palm of the hand in three orthogonal directions Ergonomics 2013 56 1823 1840 10.1080/00140139.2013.838642 24160755 10. Dong RG Welcome DE Peterson DR Xu XS McDowell TW Warren C Asaki T Kudernatsch S Brammer AJ Tool-specific performance of vibration-reducing gloves for attenuating palm-transmitted vibration in three orthogonal directions Int. J. Ind. Ergon. 2014 44 827 839 10.1016/j.ergon.2014.09.007 26726275 11. Dong RG McDowell TW Welcome DE Barkley J Warren C Washington B Effects of hand tool coupling conditions on the isolation effectiveness of air bladder anti-vibration gloves J Low Freq. Noise Vib. Active Control 2009 23 4 231 248 12. Pitts PM Mason HM Poole KA Relative performance of frequency weighting Wh and candidates for alternative frequency weightings when used to predict the occurrence of hand–arm vibration induced injuries Ind. Health 2012 50 388 396 10.2486/indhealth.MS1381 23060252 13. Miwa T Evaluation methods for vibration effect. Part 3: measurements of threshold and equal sensation contours on hand for vertical and horizontal sinusoidal vibrations Ind. Health 1967 5 213 220 10.2486/indhealth.5.213 14. Malchaire J Piette A Cock N Associations between hand–wrist musculoskeletal and sensorineural complaints and biomechanical and vibration work constraints Ann. Occup. Hyg. 2001 45 479 491 10.1093/annhyg/45.2.123 11513798 15. Dong RG Welcome DE McDowell TW A proposed theory on biodynamics frequency weighting for hand-transmitted vibration exposure Ind. Health 2012 50 412 424 10.2486/indhealth.MS1380 23060254 16. Mahbub MH Yokoyama K Laskar S Assessing the influence of anti-vibration glove on digital vascular responses to acute hand–arm vibration J. Occup. Health 2007 49 165 171 10.1539/joh.49.165 17575396 17. Jetzer T Haydon P Reynolds D Effective intervention with ergonomics antivibration gloves and medical surveillance to minimize hand–arm vibration hazards in the workplace J. Occup. Environ. Med. 2003 45 1312 1317 10.1097/01.jom.0000099981.80004.c9 14665818 18. Barregard L Ehrenström L Marcus K Hand–arm vibration syndrome in Swedish car mechanics Occup. Environ. Med. 2003 60 287 294 10.1136/oem.60.4.287 12660377 19. Cherniak M Brammer AJ Nilsson T Nerve conduction and sensorineural function in dental hygienists using high frequency ultrasound handpieces Am. J. Ind. Med. 2006 49 313 326 10.1002/ajim.20288 16570257 20. Hewitt, S.M., Dong, R.G., Welcome, D.E., McDowell, T.W.: Anti-vibration gloves? Ann. Occup. Hyg. 1–15 (2014). doi:10.1093/annhyg/meu089 21. Hewitt, S.M., Mason, H.M.: A Critical Review of Evidence Related to Hand–Arm Vibration Syndrome and the Extent of Exposure to Vibration. HSE Books. http://www.hse.gov.uk/research/rrhtm/rr1060.htm (2015). Accessed 2 Feb 2016 22. Welcome DE Dong RG Xu XS The effects of vibration-reducing gloves on finger vibration Int. J. Ind. Ergon. 2014 44 45 49 10.1016/j.ergon.2013.10.003 26543297 23. Dong RG McDowell TW Welcome DE Analysis of anti-vibration gloves mechanism and evaluation methods J. Sound Vib. 2009 321 435 453 10.1016/j.jsv.2008.09.044 24. Welcome, D.E., Dong, R.G., Xu, X.S., Warren, C., McDowell, T.W.: Tool-specific performance of vibration-reducing gloves at the fingers. In: Proceedings of the 5th American Conference on Human Vibration, Guelph, Ontario, Canada (2014) 25. Wimer BM McDowell TW Xu XS Welcome DE Warren C Dong RG Effects of gloves on total grip strength applied to cylindrical handles Int. J. Ind. Ergon. 2010 40 574 583 10.1016/j.ergon.2010.05.004 26. Laszlo, H.E., Griffin, M.J.: The transmission of vibration through gloves: effects of push force, vibration magnitude and intersubject variability. Ergonomics 54, 488–496 (2011) 27. Welcome DE Dong RG McDowell TW An evaluation of the revision of ISO 10819 Int. J. Ind. Ergon. 2012 42 143 155 10.1016/j.ergon.2011.09.003

PMC005xxxxxx/PMC5003020.txt

==== Front Front Aging NeurosciFront Aging NeurosciFront. Aging Neurosci.Frontiers in Aging Neuroscience1663-4365Frontiers Media S.A. 10.3389/fnagi.2016.00204NeuroscienceOriginal ResearchCognitive Decline and Reorganization of Functional Connectivity in Healthy Aging: The Pivotal Role of the Salience Network in the Prediction of Age and Cognitive Performances La Corte Valentina 123*†Sperduti Marco 12†Malherbe Caroline 456Vialatte François 7Lion Stéphanie 4Gallarda Thierry 28Oppenheim Catherine 4Piolino Pascale 12391Laboratory of Memory and Cognition, Institute of Psychology, University Paris DescartesParis, France2INSERM UMR S894, Center of Psychiatry and Neurosciences, University Paris DescartesParis, France3IDEX ‘Dynamique du Vieillir’, Sorbonne Paris Cité, University Paris DiderotParis, France4INSERM U894, Center of Psychiatry and Neurosciences, Department of Radiology, University Paris DescartesParis, France5Department of Computational Neuroscience, University Medical Center EppendorfHamburg, Germany6Clinic and Polyclinic of Neurology, University Medical Center EppendorfHamburg, Germany7Brain Plasticity Unit, CNRS U8249, ESPCI ParisTechParis, France8Laboratory of Physiopathology of Psychiatric Diseases, Center of Psychiatry and NeurosciencesParis, France9University Institute of France, IUFParis, FranceEdited by: Ronald Cohen, University of Florida, USA Reviewed by: Claire O’Callaghan, University of Cambridge, UK; Mihai Moldovan, University of Copenhagen, Denmark *Correspondence: Valentina La Corte, valentina.lacorte@gmail.com†These authors are co-first authors. 29 8 2016 2016 8 20415 3 2016 09 8 2016 Copyright © 2016 La Corte, Sperduti, Malherbe, Vialatte, Lion, Gallarda, Oppenheim and Piolino.2016La Corte, Sperduti, Malherbe, Vialatte, Lion, Gallarda, Oppenheim and PiolinoThis is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.Normal aging is related to a decline in specific cognitive processes, in particular in executive functions and memory. In recent years a growing number of studies have focused on changes in brain functional connectivity related to cognitive aging. A common finding is the decreased connectivity within multiple resting state networks, including the default mode network (DMN) and the salience network. In this study, we measured resting state activity using fMRI and explored whether cognitive decline is related to altered functional connectivity. To this end we used a machine learning approach to classify young and old participants from functional connectivity data. The originality of the approach consists in the prediction of the performance and age of the subjects based on functional connectivity by using a machine learning approach. Our findings showed that the connectivity profile between specific networks predicts both the age of the subjects and their cognitive abilities. In particular, we report that the connectivity profiles between the salience and visual networks, and the salience and the anterior part of the DMN, were the features that best predicted the age. Moreover, independently of the age of the subject, connectivity between the salience network and various specific networks (i.e., visual, frontal) predicted episodic memory skills either based on a standard assessment or on an autobiographical memory task, and short-term memory binding. Finally, the connectivity between the salience and the frontal networks predicted inhibition and updating performance, but this link was no longer significant after removing the effect of age. Our findings confirm the crucial role of episodic memory and executive functions in cognitive aging and suggest a pivotal role of the salience network in neural reorganization in aging. resting staters-fMRIepisodic memoryautobiographical memoryexecutive functionsfunctional connectivitymachine learningaging ==== Body Introduction The cognitive and neural changes accompanying healthy aging are a crucial topic in cognitive neuroscience. The age-related cognitive decline has emerged as a major societal concern given the increase in the elderly population. Nevertheless, not all cognitive domains are equally affected by age, and not all cognitive processes show age-related decline. There is compelling evidence that executive functions and memory are the most severely impaired cognitive domains in this population (Salthouse et al., 2003). Executive functions are seen as high-level cognitive processes responsible for flexible and adaptive behavior (Miyake et al., 2000). Thus, they play a fundamental role in dealing with complex situations in everyday life. Moreover, they largely contribute to the effective functioning of other cognitive processes, such as memory. Notably, some authors have proposed that the central deficit responsible for the general cognitive decline in aging is linked to inefficient executive functioning (West, 1996; Salthouse et al., 2003). At the neural level, this decline may be accounted by the functional and structural reorganization of the frontal lobes with aging (Moscovitch and Winocur, 1995; Cabeza, 2002; Grady et al., 2005; Fjell and Walhovd, 2010). The cognitive domain that has received the greatest attention in normal aging is memory. Many older adults complain of increased memory lapses as they age and a major focus of research has been trying to distinguish memory decline attributable to normal aging from that related to pathological aging, in particular in Alzheimer’s disease. Within the framework of long-term memory, dissociation between spared semantic memory (i.e., general knowledge about the world, words and concept) and impaired episodic memory (i.e., memory for personally experienced events that occurred in a particular place at a specific time) has been reported in aging. The episodic memory decline in older adults may result from a parallel impairment of strategic control processes involved in encoding and memory retrieval. Accordingly, several studies using laboratory tests of episodic memory have highlighted a reduction in the use of effortful encoding strategies, which are mainly related to prefrontal brain regions (Hara and Naveh-Benjamin, 2015). In the same line, considerable evidence points to deficits in effortful retrieval in older adults. In particular several studies have shown impaired free recall along with normal cued recall or recognition (Ward and Maylor, 2005). These findings show that memory decline in cognitive aging is strongly related to executive functions. Moreover, a large number of studies have investigated cognitive aging changes in episodic performance via autobiographical memory, which is defined as the memory for personal experiences that underlies the personal identity and the temporal continuity of an individual (Conway and Pleydell-Pearce, 2000). A distinction between an episodic and a semantic component has also been proposed in this domain. The former refers to memory for personal events situated in a specific spatiotemporal context, while the latter refers to general knowledge about one’s own past and about oneself. Again, dissociation between spared semantic and impaired episodic autobiographical memory has been documented in the elderly (Levine et al., 2002; Piolino et al., 2002, 2006; St. Jacques and Levine, 2007; Martinelli et al., 2013a). The deficit of the episodic component of autobiographical memory has been linked to a reduced availability of executive resources (Conway and Pleydell-Pearce, 2000; Conway, 2005; Piolino et al., 2010; Coste et al., 2011), and to a reduced recruitment of the underlying brain structures (Martinelli et al., 2013b). Functional magnetic imaging (fMRI) has been widely used in order to link age-related cognitive decline with patterns of altered brain function. A consistent finding in the fMRI literature is that healthy old adults present higher brain activation in a wide range of cognitive tasks (Cabeza, 2002). On the other hand some studies have highlighted a reduced brain activity in cognitive aging (Damoiseaux et al., 2008). More recently, an increasing number of investigations have focused on the study of the relationship between cognitive functions and functional connectivity mainly derived from resting state fMRI (rs-fMRI). rs-fMRI is the study of the interregional correlation of spontaneous fluctuation in brain activity while subjects are not engaged in any specific cognitive task. It represents a promising candidate for studying the complex neural organization underlying cognition and its modification due to different conditions (normal aging, psychiatric and neurodegenerative disorders) without task-specific confounds. The use of rs-fMRI to study functional connectivity has allowed the identification of a set of networks named resting state networks (RSNs). These networks are commonly identified across young healthy subjects (Damoiseaux et al., 2006) and have shown high reproducibility (Guo et al., 2012). The most widely studied RSN is the default mode network (DMN), composed of regions that are deactivated during the performance of goal-directed tasks and that show high levels of activity at rest. Buckner et al. (2008) defined the core regions associated with the brain’s default network: the ventral/dorsal medial prefrontal cortex (PFC), the posterior cingulate and retrosplenial cortex, the inferior parietal lobule and the hippocampal formation (including the entorhinal cortex and parahippocampal cortex). Beside the DMN, other networks of intrinsic brain connectivity have been described in healthy populations. These findings indicate that the human brain has a network-based organization even at rest. In recent years, a consistent number of investigations have focused on the salience network (Menon, 2015; Metzler-Baddeley et al., 2016). The salience network is an intrinsically connected large-scale network anchored in the anterior insula and the dorsal anterior cingulate cortex. With the anterior insula as its dynamic hub, the salience network contributes to a variety of complex brain functions through the integration of sensory, emotional and cognitive information (Menon, 2015). Recently, a direct link between inter-individual variability in functional connectivity measured at rest in specific networks and cognitive functions has been documented. For example, Cole et al. (2012) reported that the global connectivity of the lateral prefrontal cortex (LPFC) predicted individual differences in fluid intelligence. A correlation between the strength of the connectivity between the two major nodes of the DMN, the ventral medial prefrontal cortex (vMPFC) and the posterior cingulate cortex (PCC), and working memory abilities (Hampson et al., 2006), or episodic memory performances (Tambini et al., 2010) has been reported. This approach has proved fruitful in describing the neural reorganization in aging. Several studies have reported reduced connectivity between the two major nodes of the DMN, the vMPFC and the PCC (Andrews-Hanna et al., 2007; Damoiseaux et al., 2008; Balsters et al., 2013; Mevel et al., 2013). Other networks with reduced connectivity are the fronto-parietal attentional (Andrews-Hanna et al., 2007; Balsters et al., 2013; Vergun et al., 2013), the sensorimotor (Meier et al., 2012) and the salience networks (Meier et al., 2012; Onoda et al., 2012). In particular, the connectivity profile in the salience network has been shown to be the best feature to classify young and old participants using a machine learning approach (Meier et al., 2012), and that internetwork connectivity between the salience and the visual and the auditory networks is reduced in aging (Onoda et al., 2012). Moreover, a direct link between reduced network connectivity and impaired cognitive functions has been reported in aging. In particular, decreased connectivity between the anterior and the posterior node of the DMN correlated with a composite measure of memory (Andrews-Hanna et al., 2007). Concerning autobiographical memory, a correlation has been reported between the strength of connectivity between the posterior node of the DMN and middle temporal structures, comprising the hippocampus, and an episodic autobiographical fluency, and between semantic autobiographical fluency and the connectivity between the anterior node of the DMN and the ventral anterior cingulate cortex (Mevel et al., 2013). Additionally, reduced connectivity within the DMN and the salience network has been related to a decline in executive functions in aging (Damoiseaux et al., 2008; Onoda et al., 2012). Taken together, these findings highlight the pertinence of using rs-fMRI to explain the complex neuronal reorganization linked to the cognitive decline observed in aging (Andrews-Hanna et al., 2007; Damoiseaux et al., 2008; Onoda et al., 2012; Sala-Llonch et al., 2015). The principal aim of this study was to further characterize the brain functional reorganization related to cognitive aging in order to shed light on the network reorganization related to cognitive decline in older adults, in particular linked to episodic memory and executive functions. The originality of the study consisted in using a machine-learning approach to predict age and cognitive performance from functional connectivity patterns. Gottlieb (2012) recently proposed that a closer integration of machine learning in cognitive neuroscience has the potential to answer fundamental questions about cognitive functions. Such an approach has already proven its validity in recent investigations of neuropsychological features in neurology or psychiatry (Costafreda et al., 2011; Quintana et al., 2012). Developed from a connectionist approach, this modeling strategy has several advantages over computationalist methods: it can be easily applied to multi-modal data analysis, and in addition it is not constrained by a priori assumptions or abstractions on the data. The model is built using the input feature vectors (e.g., multimodal recordings of cognitive tasks) and matching this vector with expected outputs (e.g., prediction of cognitive variables). Once the model has been built, it is then confronted to a new independent test dataset to estimate its validity. Therefore, we used multivariate statistical techniques to classify young and old participants using a machine learning approach (Meier et al., 2012). We hypothesized that aging would disrupt not only DMN but also the salience network (Onoda et al., 2012) and that this pattern of modifications at the functional level would be related to cognitive changes in particular in episodic memory and executive functions. Materials and Methods Subjects Twenty-seven healthy participants, 17 young adults (YA: nine females, mean age 28.75 ± 4.62) and 10 old adults (OA: four females, mean age 70 ± 5.01) took part in the study. These participants represent a subgroup of an fMRI activation study whose data have already been published elsewhere (Martinelli et al., 2013b). All participants gave their informed written consent and the study was approved by the local ethics committee of Sainte Anne Hospital (CPP Ile de France 3 n°2687). All subjects were right-handed (according to the Edinburgh Handedness Inventory; Oldfield, 1971), and native French speakers. Medical, demographic, and psychometric data were obtained prior to the scanning session. All participants were unmedicated, living at home and rigorously screened for uncontrolled hypertension and cerebrovascular risk factors. Exclusion criteria included presence of a history of alcohol or substance abuse, head trauma, major disease affecting brain function, neuropsychiatric disorders (tested with the Mini-International Neuropsychiatric Interview, Sheehan et al., 1998), depression [tested with the Geriatric Depression Scale, Yesavage et al., 1983, cut-off score > 10; YA: 2.65 ± 2.67; OA: 3.4 ± 2.91; student t-test: t(25) = 0.68, p = 0.5], abnormal general cognitive functioning as assessed by the Mattis scale (Mattis, 1976, cut-off score < 136; young adults: 142.50 ± 1.26 and old adults: 139.90 ± 3.04). The two groups were matched according to their verbal abilities and crystallized intelligence as assessed by the Mill Hill test [Deltour, 1993; percentile score for YA: 54.38 ± 26.83 and OA: 53.83 ± 30.82, student t-test: t(25) = 0.04, p = 0.97]. Procedure The whole experimental session comprised three phases (pre-scanning, scanning, post-scanning). During the first phase (pre-scanning interview), participants were tested for exclusion and inclusion criteria, they underwent a medical examination, neuropsychological assessment and completed the Taste and Interest Questionnaire (TIQ) that was employed to create personal cues used for the autobiographical memory tasks during the scanning and post-scanning sessions. During the scanning session, participants were first trained in the autobiographical task outside the scanner, and then a high-resolution 3D structural image was acquired as well as a resting state functional session. Subsequently they participated in an activation protocol during which they performed the autobiographical task from personal cues (other than those used for training). After the fMRI protocol, during the post-scanning session (debriefing) subjects were asked to re-evoke their autobiographical memories from the same cues seen under the scan. Here we will mainly focus on the measures of neuropsychological assessment and autobiographical memories at debriefing and the rs-fMRI (for details on the activation protocol results see Martinelli et al., 2013c). Behavioral Measures Autobiographical Memory In the pre-scanning interview, exclusion and inclusion criteria were verified by means of a clinical assessment and psychometric tests, and then neuropsychological tests and the TIQ were submitted to subjects. The aim of the TIQ was mainly to collect information so as to create personalized specific event cues for each participant. Twenty-four activities or interests for episodic autobiographical memory (EAM) were selected from the TIQ (for a complete description of personal cue elaboration see Martinelli et al., 2013b; Sperduti et al., 2013). The participants were first invited to take part in a training session before the fMRI scanning. Participants received detailed explanations on the nature of the task and participated in a brief simulation of the experiment on a laptop. For the two EAM tasks (mental retrieval under the scanner and aloud retrieval at debriefing) we gave the following explanations: - EAM was defined as a memory of a single event that occurred at a specific time and place, of short duration, lasting less than 24 h. Participants were instructed to mentally relive personal episodes prompted by cues and to try to retrieve spatiotemporal, affective and perceptual details (such as time, location, perceptions, feelings, scenery, and people present in the scene) (e.g., “a unique memory linked to a trip to New York”). After the scanning session, in order to score the memories retrieved in the scanner, participants were asked to recall each memory again. EAMs were rated for richness and specificity on standard scales (Levine et al., 2002; Piolino et al., 2009; Martinelli et al., 2013a). More precisely, the presence of a sense of remembering with recall of specific spatial and temporal details, and other contextual and phenomenological details in each evocation was noted (1 point per type of detail, maximum 4; e.g., “I remembered my visit to the Palace of Tokyo in Paris, in August 2009, as if I was still there. I was with Chiara in a room at the exhibition on the first floor in the dark to see the TV reports and talk with other visitors, it was 6 pm and still very warm, but it was worth it!, after that we went to the restaurant of the outdoor museum on the bank of the Seine...”). For each participant we computed a global ratio of specificity (EAM score) totaling up the sum of spatiotemporal, other contextual and phenomenological details divided by the sum depending on the number of recalls. Episodic Memory The Free and Cued Selective Reminding test (FCRT) was used to assess episodic memory capacities (Grober and Buschke, 1987; French version Van der Linden et al., 2004). Different studies have shown the validity of this tool to discriminate healthy old adults from prodromal Alzheimer’s disease (AD) patients (Lemos et al., 2015; Papp et al., 2015). The test begins with a study phase designed to control attention and cognitive processing to identify memory impairment that it is not secondary to other cognitive deficits. During the encoding phase, subjects have to identify words in response to category cues (fruits, clothing, etc.). In the test phase, subjects are asked to recall the items they learned (free recall). The category cues are used to prompt recall of items not retrieved by free recall to generate a score called cued recall. We calculated the sum of free and cued recall termed EPI total recall. Executive Functions For the assessment of the executive functions, we used the Trail Making Test (TMT; Reitan, 1958) and verbal fluency (Cardebat et al., 1990) as a measure of behavioral and cognitive flexibility respectively. For the TMT we computed the difference in execution time between part B and part A (TMTB-A score). Concerning the verbal fluency we added the total number of words for the lexical (number of words starting with the letter P) and the semantic (number of words belonging to the semantic category “animals”) fluency (FLU score). As a measure of inhibition we used the Victoria STROOP test (Stroop, 1935). In particular we computed the difference between the time of denomination of the interference part and the denomination part (INHIB score). For up-dating (UP-D score) in working memory we used the running span (Quinette et al., 2003). For visuo-spatial working memory we used a battery assessing the visuo-spatial span (VSS score) forward and backward task (sum of the two spans), and the short-term binding (STB score) ability using a visuo-spatial binding task (Picard et al., 2012). fMRI Data Acquisition All data were acquired with a 3 T scanner (Discovery MR 750, General Electric Healthcare). The anatomical scan used an inversion recovery 3-D T1-weighted gradient-echo sequence of images (TE = 4.3 ms, TR = 11.2 ms, TI = 400 ms, matrix = 384 × 384, slice thickness = 1.2 mm). Functional resting state images were acquired using a gradient echo echoplanar (EPI) sequence (TE = 30 ms, TR = 2000 ms, flip angle = 90°, matrix = 64 × 64, slice thickness = 3 mm, 42 contiguous sections). The functional scan lasted 5 min. fMRI Data Analysis Extraction of Networks and Regions of Interest We extracted resting state networks from an independent set of resting state data available on the 1000 Functional Connectome Project1. This dataset contains functional scans of 86 subjects (45 females, age 19–85 years) acquired with a 3T scanner with the following parameters: TR = 2 s, 23 slices, acquisition type = sequential ascending. All data were processed using SPM5 software (Statistical Parametric Mapping 5, Welcome Department of Cognitive Neurology, UK2). Standard pre-processing procedures were applied to functional data. EPI volumes were corrected for slice timing, subject’s rigid motion and spatially smoothed using an isotropic Gaussian kernel filter of 5 mm full-width half-maximum. After preprocessing, resting state networks were extracted using group spatial independent component analysis (sICA) as implemented in the Network Detection using ICA (NEDICA) software (Perlbarg et al., 2008). Networks were first extracted for each subject in her/his native space. The spatial ICs obtained for each subject were then normalized in the MNI standard space and clustered into classes representative of the population. To do so we used a hierarchical clustering algorithm (Hartigan, 1975). All normalized spatial maps in each class were then averaged and thresholded at p < 0.05 using t-test statistics corrected for multiple comparisons using a false discovery rate (FDR) approach (Genovese et al., 2002). Threshold maps were visually inspected to select maps exhibiting a known spatial organization. These maps are referred to as functional networks (for a similar procedure, see Malherbe et al., 2014). Then, regions of interest were extracted from the functional networks obtained by selecting the maxima of connectivity peaks of the group functional networks. All regions of interest were defined as a sphere of 10 voxels in the Montreal Neurological Institute space (voxel size: 3.5 mm × 3.5 mm × 3.5 mm). These regions of interest were then used to extract the time course of our functional resting state data after applying the same preprocessing steps described above. The mean time series were calculated across all voxels within each region of interest in the MNI space, for each subject. The motion parameters, as well as signals from white matter and CSF and linear and quadratic drifts were then used as covariates of no interest in a general linear model for the mean time courses in each region considered in the analysis. Regions were then grouped in networks. For each network, the time course was obtained by meaning all time courses of implicated regions. Finally, a correlation matrix of the average time course between each pair of networks was computed (see Table in Appendix 1). This measure was used in the following steps of data analysis. Machine Learning Analyses In the present study, using a machine learning approach, we performed two analyses: supervised feature selection, and supervised regression. All the analyses were performed after volume correction: all variables were orthogonalized according to the cortical volume. Feature selection was performed using the orthogonal forward regression (OFR) algorithm (Chen et al., 1989), which was used to select the best network activities to predict either the age of a subject, or one of the cognitive variables. All the descriptors were considered as vectors (one fMRI was considered as one vector), and we analyzed iteratively the best set of features to model one expected output at a time. Given the feature vectors f_i,i∈[1..N] and the output vector Ω, the OFR feature selection approach follows three steps: (I) All descriptors are ranked according to their distance to the output. The distance is computed as the cosine of the angle 𝜃 between the vector and the output: 𝜃 = cos(f_i,Ω). (II) The descriptor with the lowest absolute angle (maximum cosine) is ranked first. All remaining descriptors and the output are projected into the null-space of the best descriptor. (III) The selected descriptor is stored and removed from the set, and the algorithm iterates on the remaining orthogonalized features. In order to control for the relevance of the selected features, we used a probe variables approach (Stoppiglia et al., 2003). We inserted in the feature set 100 randomly drawn vectors. The rank distribution of these probes indicated the risk of a descriptor containing information that could be explained by chance. We fixed a threshold of 5% of probes in our investigation, and selected only descriptors above that threshold. The variables were analyzed after volumetric correction (the correction was performed by orthogonalizing all variables to the null-space of the volumetry measure). Supervised regression was performed using multilayer feedforward artificial neural networks. Multilayer perceptrons are universal approximators: when good care is taken to control their complexity, they can provide better fitting than classical polynomial regressions (Haykin, 2009). We used a 2-layer perceptron, with a non-linear (sigmoid) hidden layer and a linear output. The network inputs were the two best features selected using OFR. Regression was performed using a second order gradient descent approach, with the Levenberg–Marquart algorithm (Pujol, 2007). Performances were estimated using a leave-one-out approach: (I) One sample was taken out of the database. (II) The network was then trained on the remaining samples, and afterwards tested on the excluded sample. (III) The same estimation was performed iteratively for all samples of the database. The overall classification of the excluded samples is the leave-one-out error, which is a good estimate of the generalization error. We tuned the network complexity by manipulating the number of hidden units (from 0 to 5), according to the leave-one-out error (Dreyfus, 2005). Data analyses were performed using Matlab 2013a (Mathworks®). Results Behavioral Results We performed independent sample t-tests on all the measures of interest. We found significant differences for all measures in favor of better performance in young adults than in older adults, except for FLU [t(25) = 1.71, p = 0.1]: EAM [t(25) = 7.56, p < 0.001], EPI [t(25) = 3.34, p < 0.01], TMTB-A [t(25) = 4.24, p < 0.001], INHIB [t(25) = 6.72, p < 0.001], UP-D [t(25) = 2.41, p < 0.05], VSS [t(25) = 4.57, p < 0.001], and STB [t(25) = 3.07, p < 0.01]. fMRI Results Clustering The best silhouette was obtained using six clusters. The within-cluster mean of distance to centroid was below 0.25 for all clusters except for cluster 6. Cluster 1 is particularly large and contains 16 networks (Table 1). Using the distance matrices between the cluster elements, we extracted the cluster hierarchy (Figure 1), which shows three blocks in these six clusters, composed of cluster 3, cluster 5, and the remaining clusters. Table 1 Clusters of pairs of networks (cosine distance measure), ordered according to their homogeneity and dimension. clusters 1 2 3 4 5 6 homogeneity 0.18 0.18 0.18 0.20 0.21 0.35 networks Lvattfr-dmfr Mot-sal Mot-dmps Dmfr-dmps Lvattfr-lvattps Mot-dmfr Lvattfr-dmps Mot-lvattfr Mot-vis Dmfr-dmtemp Lvattfr-rvattfr Mot-dmtemp Lvattfr-dmtemp Mot-lvattps Dmfr-vis Dmps-dmtemp Lvattfr-rvattps Mot-front Lvattfr-front Mot-rvattfr Dmps-vis Dmps-front Lvattps-rvattfr Sal-lvattfr Lvattps-dmfr Mot-rvattps Dmtemp-vis Dmtemp-front Rvattfr-rvattps Sal-lvattps Lvattps-dmps Sal-vis Front-vis Dmfr-front Sal-rvattfr Lvattps-dmtemp Lvattfr-vis Sal-rvattps Lvattps-front Lvattps-rvattps Sal-dmfr Rvattfr-dmfr Lvattps-vis Sal-dmps Rvattfr-dmps Rvattfr-vis Sal-dmtemp Rvattfr-dmtemp Sal-front Rvattfr-front Rvattps-dmfr Rvattps-dmps Rvattps-dmtemp Rvattps-front Lvattfr, left ventral attentional frontal; Lvattps, left ventral attentional posterior; Rvattfr, right ventral attentional frontal; Rvattps, right ventral attentional posterior; dmfr, default mode frontal; dmps, default mode posterior; dmtemps, default mode temporal; front, frontal; Mot, motor; Sal, salience; vis, visual.FIGURE 1 Dendrogram of the clusters, computed from their cosine distance matrices. Prediction of Age The best two pairs of networks to predict age according to the OFR algorithm were sal-vis and sal-dmfr, corresponding to clusters 2 and 6. These two clusters belong to the same block on the dendrogram. A Pearson correlation R2 of 0.61 was obtained (p = 2.07∗10-5) using these two variables. The best neural network architecture selected had four hidden units. A leave-one-out mean-squared error of 0.06 was achieved for the age prediction, corresponding to a generalization error of 6.69 ± 5.3 years (the prediction error reached 9.10-7 on the training set) (Figure 2). Given the pivotal role of the salience network in predicting age and cognitive variables, we conducted post hoc analyses on the within network connectivity. Interestingly, we found that the connectivity between the cingulate gyrus and insula (which are the main hubs of the salience network) was reduced in the group of old adults (t = -2.52, p < 0.05). Moreover the connectivity between these two brain regions showed a negative correlation with age in the old adults (R = -0.7, p < 0.05), but not in the young adults group (R = -0.03, p = 0.9). Nevertheless, when accounting for within network connectivity, we were still able to significantly predict the age from the between network connectivity between the sal-dmnf and the sal-vis (R = 0.368, p = 4.06 E-03). FIGURE 2 Prediction of the age of the subjects from their rs-fMRI inter network activity. (A) Linear regression, each circle represents a subject, the colored plane represents the linear regression. (B) Non-linear regression based on the multilayer perceptron, on the variables after volume correction, on the leave-one-out validation set. Each dot represents a subject; the dashed line represents the optimum. Prediction of Cognitive Variables Seven cognitive variables were successfully predicted from rs-fMRI pairs of networks (see Table 2): EPI total, STB, EAM, FLU, TMT B-A, and VSS Scores. The best prediction was obtained with EPI total, which had the highest linear correlation p-values and a very low leave-one-out generalization error (6%). Moreover, three of these variables were regressed efficiently using a neural network, with a satisfactory leave-one-out error (below 15%). These three variables (EPI total, STB, EAM) were predicted from the same clusters (clusters 6 and 2), which are the same as the age predicting clusters. The four remaining variables had poorer leave-one-out errors (above 20%). Three variables (FLU, TMT B-A, VSS) had cluster 5 in common. Within those three, the last two variables (TMT B-A, and VSS) shared the exact same networks (belonging to cluster 5 and cluster 6). Two other variables showed the same network (cluster 6), but did not show significant correlations after age correction (p > 0.05): INHIB score, UP-D score (Table 2; Figure 3). Table 2 Cognitive variables prediction from fMRI pairs of networks. Variable Networks Clusters Significance Corrected significance Learning error EPI total Sal-front 6, 2 R2 0.59 R2 0.46 0.06 Mot-lvattps p = 3.1∗10-5 p = 7.9∗10-4 STB Sal-dmps 6, 2 R2 0.31 R2 0.44 0.14 Rvattps-vis p = 1.3∗10-2 p = 1.2∗10-3 EAM Sal-dmtemp 6, 2 R2 0.60 R2 0.30 0.14 Sal-vis p = 2.4∗10-5 p = 0.015 FLU Rvattfr-dmfr 1, 5 R2 0.44 R2 0.35 0.21 Lvattps-Rvattfr p = 1.3∗10-3 p = 6.6∗10-3 TMT B-A Lvattps-Rvattfr 5, 6 R2 0.49 R2 0.38 0.21 Sal-rvattps p = 4.4∗10-4 p = 3.9∗10-3 VSS Lvattps-Rvattfr 5, 6 R2 0.38 R2 0.33 0.29 Sal-rvattps p = 4.3∗10-3 p = 9.9∗10-3 INHIB Sal-front 6 R2 0.56 R2 0.22 0.11 p = 8.9∗10-5 p = 0.053 UP-D Sal-dmtmp 6 R2 0.34 R2 0.16 0.19 Sal-rvattps p = 8.3∗10-3 p = 0.14 The network column indicates the two best pairs of networks selected using OFR algorithm. The cluster column indicates the corresponding clusters. The significance column indicates Pearson correlation results using the selected networks. Corrected significance is the result of Pearson correlation analysis after correction by the age variable. Learning error is the leave-one-out generalization error on the normalized output. Grey color indicated satisfactory leave-one-out error (below 15%). EPI total, total score episodic memory; STB, short term binding; EAM, episodic autobiographical memory; FLU, verbal fluency; TMTB-A, difference of time execution between the part B and the part A; VSS, visuo-spatial span; INHIB, difference between the time of denomination of the interference part and the denomination part; UP-D, updating in working memory.FIGURE 3 Prediction of EPI total (verbal episodic memory) of the subjects from their rs-fMRI inter network activity. (A) Linear regression on the corrected variables (correction by volume and age). Each circle represents a subject, the colored plane represents the linear regression. (B) Non-linear regression based on the multilayer perceptron, on the variables after volume and age correction, on the leave-one-out validation set. Each dot represents a subject; the dashed line represents the optimum (obtained with a linear perceptron without hidden units). Furthermore by combining the measures of distances and the variable predictions, it was possible to draw a general graph of the relationships between the cognitive activation clusters and the predicted variables independently of age (Figure 4). On that graph, we can identify three functional blocks (clusters 2 + 6; clusters 6 + 5, clusters 5 + 1). One can see the central importance of cluster 6, which is involved in all but one variable prediction (verbal fluency FLU) and was identified as a general cluster of cognitive decline: this cluster also predicts age, and in addition most cognitive variables when excluding age effects. The functional block of verbal fluency FLU is the only one that is not related to cluster 6. Cluster number 4 was not associated directly to any of the investigated variables (but as it is close to clusters 2 and 6, it could nevertheless be used as a replacement cluster for the prediction of the variables Age, EPI total, STB, and EAM). FIGURE 4 Cognitive network of clusters. The distances between each cluster represents the dendrogram distance already illustrated in Figure 1. Each ellipse indicates groups of variables successfully predicted by pairs of the same activation networks. EPI total, total score verbal episodic memory; STB, short term binding; EAM, episodic autobiographical memory; FLU, verbal fluency; TMTB-A, difference of time execution between the part B and the part A; VSS, visuo-spatial span. Discussion In this work, we showed that the connectivity profile between specific RSN networks predicts both the age of the subjects and their cognitive abilities. The originality of the study consisted in using a machine-learning approach to predict age and cognitive performance from the functional connectivity patterns characterized in the brain. In particular we reported that the connectivity between the salience and visual networks, and the salience and the anterior part of the DMN, were the best features in predicting the age of the subjects. Moreover, connectivity between the salience and different specific networks predicted the episodic performance (Sal-front), the short-term binding (Sal-dmps) and the episodic autobiographical score (Sal-dmtemp, Sal-vis), independently of the age of the subject. Finally, the connectivity between the salience and the frontal networks predicted inhibition and updating performance, but this correlation was no longer significant after removing the effect of age. Our findings suggest a pivotal role of the salience network in the neural reorganization in aging. The connectivity profile of this network was not only the best feature to predict age, but was also involved in the prediction of several cognitive functions, such as verbal episodic memory, short-term binding and episodic autobiographical memory. Nevertheless, these scores were also predicted independently of the age of the subject, thus the variability in the strength of connectivity between these networks seems more linked to the variability in cognitive functions per se than to the effect of aging. On the contrary, the prediction of inhibitory and updating performances dropped when age was taken into account, suggesting that the connectivity profile of the network predicting inhibition and updating in working memory is particularly sensitive to the effect of aging. These findings are in line with studies that have related cognitive deficits in the elderly to a reduction in inhibitory control (Hasher et al., 1999). The central role of the salience network reported here is coherent with recent findings showing that the connectivity profile of this network was one of the best predictors of age (Meier et al., 2012), and that the connectivity between the salience and the visual networks and the salience and the temporal networks was correlated with age (Onoda et al., 2012). In accordance with the latter study we showed that one of the best features predicting age was the connectivity between the salience and the visual network. On the contrary, while Onoda et al. (2012) did not report robust alteration of the default mode with age, we found that another feature involved in age prediction was the connectivity between the salience and the anterior portion of the default mode. The salience network, composed of the anterior cingulate cortex (ACC) and the insula, is thought to code behaviorally relevant information (Seeley et al., 2007). One recent proposal is that this network, in particular the insular cortex, may promote the dynamic switch between other large scale networks (e.g., the default mode and the central executive network) in order to ensure adaptive behavior via flexible cognitive control mechanisms (Sridharan et al., 2008; Menon and Uddin, 2010). Recent studies have reported an altered salience network in normal aging. In particular He et al. (2014) showed that functional and structural impairment of the salience network may occur early in normal aging and that functional disconnection between this network and the central executive network and the DMN may also be associated with normal aging and Alzheimer’s disease. Moreover, the connectivity between the salience network and specific networks predicted different cognitive functions. In particular, the connectivity with the frontal networks predicted episodic memory performance. This finding is in line with the role of the frontal cortex in both encoding and retrieval of episodic memory (e.g., Spaniol et al., 2009). Concerning episodic autobiographical memory, we reported that performance was predicted by the connectivity between the salience network and the temporal component of the DMN, comprising the hippocampus. The role of the hippocampus in episodic autobiographical memory is well established (see for example two meta-analyses: Svoboda et al., 2006; Martinelli et al., 2013c). A recent investigation by Grady et al. (2015) pointed out that the salience network is also engaged during recall failures. In particular these findings suggest that the dedifferentiation of functional connectivity within the salience network across memory conditions and the reduction in functional coupling between it and the PFC may indicate weak inter-network communication either while retrieval is attempted or when monitoring takes place after retrieval has failed. In addition, supplementary results showed that the connectivity between crucial hubs of the salience network, such as cingulate gyrus and insula, was reduced in elderly subjects and that the connectivity between these two regions showed a negative correlation with age only in the old adults group. These findings highlight the involvement of the principal hubs of the Salience Network in neurocognitive aging. Moreover, this reduction of connectivity correlated with age only in the elderly group. Nevertheless, we were still able to predict the age of the subjects from between networks connectivity when within network connectivity was taken into account. These findings suggest that while age is accompanied by an alteration of the intrinsic dynamic of the salience network; inter networks connectivity seems to represent more robust predictors of age. Finally, the connectivity between the posterior portion of the DMN, comprising the temporo-parietal junction (TPJ) and the precuneus/posterior cingulate cortex, and the salience network predicted short-term binding in working memory. The temporo-parietal junction, beyond attentional and social functions (Scholz et al., 2009), has been linked to working memory processes (Anticevic et al., 2010). Moreover, interestingly, a recent study reported a direct involvement of the TPJ in visual feature binding (Pollmann et al., 2014). Thus, the role of this structure seems coherent with the cognitive demands of our short-term binding task. Taken together these findings suggest that the salience network allocates the necessary cortical resources to other networks that are specialized in the task at hand. Moreover, the link between the connectivity of these networks and the corresponding cognitive functions does not seem to be particularly sensitive to aging, since correlations remain significant even after the effect of age is taken into account. On the contrary, the correlation between the connectivity of the salience and the frontal network and inhibition performance was affected by age. The link between both the ACC, one of the nodes of the salience network, and the PFC and inhibition, especially during the Stroop task, is well documented (Laird et al., 2005; Nee et al., 2007). Moreover, a recent study showed that performance on the Stroop task was associated with the integrity of fiber tracts connecting these structures in aging, even when controlling for general processing speed (Wolf et al., 2014). Interestingly another recent study has shown that the role of the salience network changes over the life span, which may have implications for the early detection of pathophysiology in elderly populations (Archer et al., 2016). Conclusion The present study highlights the crucial role of the salience network in cognitive aging related to specific cognitive decline in particular in episodic memory and executive functions. This network is situated at the interface of the cognitive, motivational and affective system of the human brain. It plays a crucial role in identifying the most biologically and cognitively relevant endogenous and external stimuli in order to adaptively guide behavior (Menon, 2015). Indeed it can be considered as a key brain system for integrating cognition, action and feelings. Further research on normal aging and pathological populations is needed to better characterize the role of disrupted connectivity in the preclinical phase of neurodegenerative disease. Within this context the early detection of functional connectivity abnormalities may be helpful for early diagnosis of the diseases with the aim of characterizing a pathological signature of the reorganization of brain networks in pathological aging. Author Contributions VLC, MS, FV, and PP wrote the article. MS and SL did the neuroimaging exams. CM, FV did the data processing and data analyses. PP, TG, and CO conceptualized the experiment. All the authors contributed to the final draft of the article. 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 authors gratefully acknowledge support from the National Hospital Clinical Research Program NEMAUVI (to TG and PP) and the Institut Universitaire de France (to PP) and the Excellent Initiative of Sorbonne Paris Cité, IDEX “Dynamics of Aging” University Paris Diderot, Paris, France (VLC postdoctoral contract). We thank all volunteers for their participation in this study and the neuroimaging staff of the Center of Psychiatry and Neuroscience at Sainte Anne Hospital, especially A. D. Devauchelle, P. Martinelli and Marion Delhommeau for their help in the neuroimaging and neuropsychological exams. 1 http://www.nitrc.org/frs/shownotes.php?release_id=916 2 www.fil.ion.ucl.ac.uk/spm Supplementary Material The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fnagi.2016.00204 Click here for additional data file. ==== Refs References Andrews-Hanna J. R. Snyder A. Z. Vincent J. L. Lustig C. Head D. Raichle M. E. (2007 ). Disruption of large-scale brain systems in advanced aging. Neuron 56 924 –935 . 10.1016/j.neuron.2007.10.038 18054866 Anticevic A. Repovs G. Shulman G. L. Barch D. M. (2010 ). When less is more: TPJ and default network deactivation during encoding predicts working memory performance. Neuroimage 49 2638 –2648 . 10.1016/j.neuroimage.2009.11.008 19913622 Archer J. A. Lee A. Qiu A. Chen S. A. (2016 ). A compréhensive analysis of connectivity and aging over adukt life span. Brain Connect. 6 169 –185 .26652914 Balsters J. H. O’Connell R. G. Galli A. Nolan H. Greco E. Kilcullen S. M. (2013 ). Changes in resting connectivity with age: a simultaneous electroencephalogram and functional magnetic resonance imaging investigation. Neurobiol. Aging 34 2194 –2207 . 10.1016/j.neurobiolaging.2013.03.004 23608113 Buckner R. L. Andrews-Hanna J. R. Schacter D. L. (2008 ). The brain’s default network: anatomy, function, and relevance to disease. Ann. N. Y. Acad. Sci. 1124 1 –38 . 10.1196/annals.1440.011 18400922 Cabeza R. (2002 ). Hemispheric asymmetry reduction in older adults: the HAROLD model. Psychol. Aging 17 85 –100 . 10.1037/0882-7974.17.1.85 11931290 Cardebat D. Doyon B. Puel M. Goulet P. Joanette Y. (1990 ). [Formal and semantic lexical evocation in normal subjects. Performance and dynamics of production as a function of sex, age and educational level]. Acta Neurol. Belg. 90 207 –217 .2124031 Chen S. Billings S. A. Luo W. (1989 ). Orthogonal least squares methods and their application to non-linear system identification. Int. J. Control 50 1873 –1896 . 10.1080/00207178908953472 Cole M. W. Yarkoni T. Repovš G. Anticevic A. Braver T. S. (2012 ). Global connectivity of prefrontal cortex predicts cognitive control and intelligence. J. Neurosci. 32 8988 –8999 . 10.1523/JNEUROSCI.0536-12.2012 22745498 Conway M. A. (2005 ). Memory and the self. J. Mem. Lang. 53 594 –628 . 10.1016/j.jml.2005.08.005 Conway M. A. Pleydell-Pearce C. W. (2000 ). The construction of autobiographical memories in the self-memory system. Psychol. Rev. 107 261 –88 . 10.1037/0033-295X.107.2.261 10789197 Costafreda S. G. Fu C. H. Picchioni M. Toulopoulou T. McDonald C. Kravariti E. (2011 ). Pattern of neural responses to verbal fluency shows diagnostic specificity for schizophrenia and bipolar disorder. BMC Psychiatry 11 :18 10.1186/1471-244X-11-18 Coste C. Agar N. Petitfour E. Quinette P. Guillery-Girard B. Azouvi P. (2011 ). Exploring the roles of the executive and short-term feature-binding functions in retrieval of retrograde autobiographical memories in severe traumatic brain injury. Cortex 47 771 –786 . 10.1016/j.cortex.2010.07.004 20817159 Damoiseaux J. S. Beckmann C. F. Arigita E. S. Barkhof F. Scheltens P. Stam C. J. (2008 ). Reduced resting-state brain activity in the “default network” in normal aging. Cereb. Cortex 18 1856 –1864 . 10.1093/cercor/bhm207 18063564 Damoiseaux J. S. Rombouts S. A. R. B. Barkhof F. Scheltens P. Stam C. J. Smith S. M. (2006 ). Consistent resting-state networks across healthy subjects. Proc. Natl. Acad. Sci. U.S.A. 103 13848 –13853 . 10.1073/pnas.0601417103 16945915 Deltour J. J. (1993 ). Echelle de Vocabulaire de Mill Hill de J. C. Raven. Braine-le-Château : Éditions l’Application des Techniques Modernes SPRL . Dreyfus G. (2005 ). Neural Networks: Methodology and Applications. Berlin : Springer-Verlag . Fjell A. M. Walhovd K. B. (2010 ). Structural brain changes in aging: courses, causes and cognitive consequences. Rev. Neurosci. 21 187 –222 . 10.1515/REVNEURO.2010.21.3.187 20879692 Genovese C. R. Lazar N. A. Nichols T. (2002 ). Thresholding of statistical maps in functional neuroimaging using the false discovery rate. Neuroimage 15 870 –878 . 10.1006/nimg.2001.1037 11906227 Gottlieb J. (2012 ). Attention, learning, and the value of information. Neuron 76 281 –295 . 10.1016/j.neuron.2012.09.034 23083732 Grady C. L. McIntosh A. R. Craik F. I. (2005 ). Task-related activity in prefrontal cortex and its relation to recognition memory performance in young and old adults. Neuropsychologia 43 1466 –1481 .15989937 Grady C. L. St-Laurent M. Burianovà H. (2015 ). Age differences in brain activity related to unsuccessful déclarative memory retrieval. Brain Res. 1612 30 –47 .25541365 Grober E. Buschke H. (1987 ). Genuine memory deficits in dementia. Dev. Neuropsychol. 3 13 –36 . 10.1080/87565648709540361 Guo C. C. Kurth F. Zhou J. Mayer E. A. Eickhoff S. B. Kramer J. H. (2012 ). One-year test–retest reliability of intrinsic connectivity network fMRI in older adults. Neuroimage 61 1471 –1483 . 10.1016/j.neuroimage.2012.03.027 22446491 Hampson M. Driesen N. R. Skudlarski P. Gore J. C. Constable R. T. (2006 ). Brain connectivity related to working memory performance. J. Neurosci. 26 13338 –13343 . 10.1523/JNEUROSCI.3408-06.2006 17182784 Hara Y. Naveh-Benjamin M. (2015 ). The role of reduced working memory storage and processing resources in the associative memory deficit of older adults : simulation studies with younger adults. Neuropsychol. Dev. Cogn. B Aging Neuropsychol. Cogn. 22 129 –154 .24617835 Hartigan J. A. (1975 ). Clustering Algorithms. New York, NY : Wiley . Hasher L. Zacks R. T. Rahhal T. A. (1999 ). Timing, instructions, and inhibitory control: some missing factors in the age, and memory debate. Gerontology 45 355 –357 . 10.1159/000022121 10559658 Haykin S. (2009 ). Neural Networks and Learning Machines, 3rd Edn. Upper Saddle River, NJ : Pearson Education . He X. Qin W. Liu Y. Zhang X. Duan Y. Song J. (2014 ). Abnormal salience network in normal aging and in amnestic mild cognitive impairement and Alzheimer disease. Hum. Brain Mapp. Jul. 35 3446 –3464 . 10.1002/hbm.22414 Laird A. R. McMillan K. M. Lancaster J. L. Kochunov P. Turkeltaub P. E. Pardo J. V. (2005 ). A comparison of label-based review and ALE meta-analysis in the stroop task. Hum. Brain Mapp. 25 6 –21 . 10.1002/hbm.20129 15846823 Lemos R. Maroco J. Simoes M. R. Santiago B. Tomas J. Santana I. (2015 ). The free and cued sélective reminding test for predictng progression to Alzheimer’s disease in patients with mild cognitive impairement: a prosective longitudinal study. J. Neuropsychol. 10.1111/jnp.12075 [Epub ahead of print] . Levine B. Svoboda E. Hay J. F. Winocur G. Moscovitch M. (2002 ). Aging and autobiographical memory: dissociating episodic from semantic retrieval. Psychol. Aging 17 677 –689 . 10.1037/0882-7974.17.4.677 12507363 Malherbe C. Messé A. Bardinet E. Pélégrini-Issac M. Perlbarg V. Marrelec G. (2014 ). Combining spatial independent component analysis with régression to identify the subcortical components of resting state FMRI functional networks. Brain Connect 4 181 –192 .24575752 Martinelli P. Anssens A. Sperduti M. Piolino P. (2013a ). The influence of normal aging and Alzheimer’s disease in autobiographical memory highly related to the self. Neuropsychology 27 69 –78 . 10.1037/a0030453 23148495 Martinelli P. Sperduti M. Devauchelle A. D. Kalenzaga S. Gallarda T. Lion S. (2013b ). Age-related changes in the functional network underlying specific and general autobiographical memory retrieval: a pivotal role for the anterior cingulate cortex. PLoS ONE 8 :e82385 10.1371/journal.pone.0082385 Martinelli P. Sperduti M. Piolino P. (2013c ). Neural substrates of the self-memory system: new insights from a meta-analysis. Hum. Brain Mapp. 34 1515 –1529 . 10.1002/hbm.22008 22359397 Mattis S. (1976 ). “Mental Status examination for organic mental syndrome in the elderly patient ,” in Geriatric Psychiatry , eds Bellack L. Karusu T. B. (New York, NY : Grune and Stratton ), 77 –121 . Meier T. B. Desphande A. S. Vergun S. Nair V. A. Song J. Biswal B. B. (2012 ). Support vector machine classification and characterization of age-related reorganization of functional brain networks. Neuroimage 60 601 –613 . 10.1016/j.neuroimage.2011.12.052 22227886 Menon V. (2015 ). “Salience network ,” in Brain Mapping: An Encyclopedic Reference Vol. 2 ed. Toga A. W. (Cambridge, MA : Academic Press ), 597 –611 . Menon V. Uddin L. Q. (2010 ). Saliency, switching, attention and control: a network model of insula function. Brain Struct. Funct. 214 655 –667 . 10.1007/s00429-010-0262-0 20512370 Metzler-Baddeley C. Caeyenberghs K. Foley S. Jones D. K. (2016 ). Task complexity, and location specific changes of cortical thickness in exécutive and salience networks after working memory training. Neuroimage. 130 48 –62 . 10.1016/j.neuroimage.2016.01.007 26806288 Mevel K. Landeau B. Fouquet M. La Joie R. Villain N. Mézenge F. (2013 ). Age effect on the default mode network, inner thoughts, and cognitive abilities. Neurobiol. Aging 34 1292 –1301 . 10.1016/j.neurobiolaging.2012.08.018 23084083 Miyake A. Friedman N. P. Emerson M. J. Witzki A. H. Howerter A. Wager T. D. (2000 ). The unity and diversity of executive functions and their contributions to complex “frontal lobe” tasks: a latent variable analysis. Cognit. Psychol. 41 49 –100 . 10.1006/cogp.1999.0734 10945922 Moscovitch M. Winocur G. (1995 ). Frontal lobes, memory and aging. Ann. N. Y. Acad. Sci. 769 119 –150 . 10.1111/j.1749-6632.1995.tb38135.x 8595020 Nee D. E. Wager T. D. Jonides J. (2007 ). Interference resolution: insights from a meta- analysis of neuroimaging tasks. Cogn. Affect. Behav. Neurosci. 7 1 –17 . 10.3758/CABN.7.1.1 17598730 Oldfield R. C. (1971 ). The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia 9 97 –113 . 10.1016/0028-3932(71)90067-4 5146491 Onoda K. Ishihara M. Yamaguchi S. (2012 ). Decreased functional connectivity by aging is associated with cognitive decline. J. Cogn. Neurosci. 24 2186 –2198 . 10.1162/jocn_a_00269 22784277 Papp K. V. Amariglio R. E. Mormino E. C. Hedden T. Dekhytar M. Johnson K. A. (2015 ). Free and cued Memory in relation to biomarker-defined abnormalities in clinically normal older adults and those at risk for Alzheimer disease. Neuropsychologia 73 169 –175 . 10.1016/j.neuropsychologia.2015.04.034 26002757 Perlbarg V. Marrelec G. Doyon J. Pélégrini-Issac M. Lehericy S. Benali H. (2008 ). “NEDICA: detection of group functional networks in fMRI using spatial independent component analysis ,” in Proceedings of the 5th IEEE International Symposium on Biomedical Imaging: From Nano to Macro, 2008 (Rome : IEEE ), 1247 –1250 . Picard L. Cousin S. Guillery-Girard B. Eustache F. Piolino P. (2012 ). How do the different components of episodic Memory develop ? Role of exécutive functions and short-term feature-binding abilities. Child Dev. 83 1037 –1050 . 10.1111/j.1467-8624.2012.01736.x 22364311 Piolino P. Coste C. Martinelli P. Macé A. L. Quinette P. Guillery-Girard B. (2010 ). Reduced specificity of autobiographical memory and aging: Do the executive and feature binding functions of working memory have a role? Neuropsychologia 48 429 –440 . 10.1016/j.neuropsychologia.2009.09.035 19804792 Piolino P. Desgranges B. Benali K. Eustache F. (2002 ). Episodic and semantic remote autobiographical memory in ageing. Memory 10 239 –257 . 10.1080/09658210143000353 12097209 Piolino P. Desgranges B. Clarys D. Guillery-Girard B. Taconnat L. Isingrini M. (2006 ). Autobiographical memory, autonoetic consciousness, and self-perspective in aging. Psychol. Aging 21 510 10.1037/0882-7974.21.3.510 Piolino P. Desgranges B. Eustache F. (2009 ). Episodic autobiographical memories over the course of time: cognitive, neuropsychological and neuroimaging findings. Neuropsychologia 47 2314 –2329 . 10.1016/j.neuropsychologia.2009.01.020 19524095 Pollmann S. Zinke W. Baumgartner F. Geringswald F. Hanke M. (2014 ). The right temporo-parietal junction contributes to visual feature binding. Neuroimage 101 289 –297 . 10.1016/j.neuroimage.2014.07.021 25038438 Pujol J. (2007 ). The solution of nonlinear inverse problems and the Levenberg-Marquardt method. Geophysics 72 W1 –W16 . 10.1190/1.2732552 Quinette P. Guillery B. Desgranges B. de la Sayette V. Viader F. Eustache F. (2003 ). Working memory and executive functions in transient global amnesia. Brain 126 1917 –1934 . 10.1093/brain/awg201 12876141 Quintana M. Guàrdia J. Sánchez-Benavides G. Aguilar M. Molinuevo J. L. Robles A. (2012 ). Using artificial neural networks in clinical neuropsychology: high performance in mild cognitive impairment and Alzheimer’s disease. J. Clin. Exp. Neuropsychol. 34 195 –208 . 10.1080/13803395.2011.630651 22165863 Reitan R. M. (1958 ). Validity of the trail making test as an indicator of organic brain damage. Percept. Mot. Skills 1958 271 –276 . 10.2466/PMS.8.7.271-276 Sala-Llonch R. Bartres-Faz D. Junqué C. (2015 ). Reorganization of bran networks in aging: a review of functional connectivity studies. Front. Psychol. 6 :663 10.3389/fpsyg.2015.00663 Salthouse T. A. Atkinson T. M. Berish D. E. (2003 ). Executive functioning as a potential mediator of age-related cognitive decline in normal adults. J. Exp. Psychol. Gen. 132 566 –594 . 10.1037/0096-3445.132.4.566 14640849 Scholz J. Triantafyllou C. Whitfield-Gabrieli S. Brown E. N. Saxe R. (2009 ). Distinct regions of right temporo-parietal junction are selective for theory of mind and exogenous attention. PLoS ONE 4 :e4869 10.1371/journal.pone.0004869 Seeley W. W. Menon V. Schatzberg A. F. Keller J. Glover G. H. Kenna H. (2007 ). Dissociable intrinsic connectivity networks for salience processing and executive control. J. Neurosci. 27 2349 –2356 . 10.1523/JNEUROSCI.5587-06.2007 17329432 Sheehan D. V. Lecrubier Y. Sheehan K. H. Amorim P. Janavs J. Weiller E. (1998 ). The Mini-International Neuropsychiatric Interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J. Clin. Psychiatry 59 22 –33 .9881538 Spaniol J. Davidson P. S. Kim A. S. Han H. Moscovitch M. Grady C. L. (2009 ). Event-related fMRI studies of episodic encoding and retrieval: meta-analyses using activation likelihood estimation. Neuropsychologia 47 1765 –1779 . 10.1016/j.neuropsychologia.2009.02.028 19428409 Sperduti M. Martinelli P. Kalenzaga S. Devauchelle A. D. Lion S. Malherbe C. (2013 ). Don’t be too strict with yourself ! Rigid négative self-representation in healthy subjects mimics the neurocognitive profile of dépression for autobiographical Memory. Front. Behav. Neurosci. 21 :41 10.3389/fnbeh.2013.00041 Sridharan D. Levitin D. J. Menon V. (2008 ). A critical role for the right fronto-insular cortex in switching between central-executive and default-mode networks. Proc. Natl. Acad. Sci. U.S.A. 105 12569 –12574 . 10.1073/pnas.0800005105 18723676 St. Jacques P. L. Levine B. (2007 ). Ageing and autobiographical memory for emotional and neutral events. Memory 15 129 –144 . 10.1080/09658210601119762 17534107 Stoppiglia H. Dreyfus G. Dubois R. Oussar Y. (2003 ). Ranking a random feature for variable and feature selection. J. Mach. Learn. Res. 3 1399 –1414 . Stroop J. R. (1935 ). Studies of interference in serial verbal reactions. J. Exp. Psychol. 18 643 –662 . 10.1037/h0054651 Svoboda E. McKinnon M. C. Levine B. (2006 ). The functional neuroanatomy of autobiographical memory: a meta-analysis. Neuropsychologia 44 2189 –2208 . 10.1016/j.neuropsychologia.2006.05.023 16806314 Tambini A. Ketz N. Davachi L. (2010 ). Enhanced brain correlations during rest are related to memory for recent experiences. Neuron 65 280 –290 . 10.1016/j.neuron.2010.01.001 20152133 Van der Linden M. Coyette F. Poitrenaud J. Kalafat M. Calicis F. Wyns C. (2004 ). “L’épreuve de rappel libre/rappel indicé à 16 items (RL/RI-16) ,” in L’Évaluation des Troubles de la Mémoire: Présentation de Quatre Tests de Mémoire Épisodique Avec Leur Étalonnage , eds Van der M. Linden S. Adam A. Agniel, et les membres du GREMEM (Marseille : SOLAL ). Vergun S. Deshpande A. S. Meier T. B. Song J. Tudorascu D. L. Nair V. A. (2013 ). Characterizing functional connectivity differences in aging adults using machine learning on resting state fMRI data. Front. Comput. Neurosci. 7 :38 10.3389/fncom.2013.00038 Ward G. Maylor E. A. (2005 ). Age –related déficits in free recall : the rôle of rehearsal. Q. J. Exp. Psychol. A 58 98 –119 . 10.1080/02724980443000223 15881293 West R. L. (1996 ). An application of prefrontal cortex function theory to cognitive aging. Psychol. Bull. 120 272 –292 . 10.1037/0033-2909.120.2.272 8831298 Wolf D. Zschutschke L. Scheurich A. Schmitz F. Lieb K. Tuscher O. (2014 ). Age-related increases in Stroop interférence: délinéation of général slowing based on behavioral and whitematter analyses. Hum. Brain Mapp. 35 2448 –2458 .24038539 Yesavage J. A. Brink T. L. Rose T. L. Lum O. Huang V. Adey M. (1983 ). Development and validation of a geriatric depression screening scale: a preliminary report. J. Psychiatr. Res. 17 37 –49 . 10.1016/0022-3956(82)90033-4 7183759

PMC005xxxxxx/PMC5003041.txt

==== Front Front Plant SciFront Plant SciFront. Plant Sci.Frontiers in Plant Science1664-462XFrontiers Media S.A. 10.3389/fpls.2016.01227Plant ScienceOriginal ResearchCassava Breeding I: The Value of Breeding Value Ceballos Hernán 1*Pérez Juan C. 12Joaqui Barandica Orlando 1Lenis Jorge I. 1Morante Nelson 1Calle Fernando 1Pino Lizbeth 1Hershey Clair H. 11International Center for Tropical AgricultureSantiago de Cali, Colombia2Corporación Colombiana de Investigación AgropecuariaSanta Marta, ColombiaEdited by: Soren K. Rasmussen, University of Copenhagen, Denmark Reviewed by: Guillaume Jean Bauchet, Boyce Thompson Institute, USA; Paul Gibson, Makerere University, Uganda *Correspondence: Hernán Ceballos h.ceballos@cgiar.orgThis article was submitted to Crop Science and Horticulture, a section of the journal Frontiers in Plant Science 29 8 2016 2016 7 122705 5 2016 02 8 2016 Copyright © 2016 Ceballos, Pérez, Joaqui Barandica, Lenis, Morante, Calle, Pino and Hershey.2016Ceballos, Pérez, Joaqui Barandica, Lenis, Morante, Calle, Pino and HersheyThis is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.Breeding cassava relies on several selection stages (single row trial-SRT; preliminary; advanced; and uniform yield trials—UYT). This study uses data from 14 years of evaluations. From more than 20,000 genotypes initially evaluated only 114 reached the last stage. The objective was to assess how the data at SRT could be used to predict the probabilities of genotypes reaching the UYT. Phenotypic data from each genotype at SRT was integrated into the selection index (SIN) used by the cassava breeding program. Average SIN from all the progenies derived from each progenitor was then obtained. Average SIN is an approximation of the breeding value of each progenitor. Data clearly suggested that some genotypes were better progenitors than others (e.g., high number of their progenies reaching the UYT), suggesting important variation in breeding values of progenitors. However, regression of average SIN of each parental genotype on the number of their respective progenies reaching UYT resulted in a negligible coefficient of determination (r2 = 0.05). Breeding value (e.g., average SIN) at SRT was not efficient predicting which genotypes were more likely to reach the UYT stage. Number of families and progenies derived from a given progenitor were more efficient predicting the probabilities of the progeny from a given parent reaching the UYT stage. Large within-family genetic variation tends to mask the true breeding value of each progenitor. The use of partially inbred progenitors (e.g., S1 or S2 genotypes) would reduce the within-family genetic variation thus making the assessment of breeding value more accurate. Moreover, partial inbreeding of progenitors can improve the breeding value of the original (S0) parental material and sharply accelerate genetic gains. For instance, homozygous S1 genotypes for the dominant resistance to cassava mosaic disease (CMD) could be generated and selected. All gametes from these selected S1 genotypes would carry the desirable allele and 100% of their progenies would be resistant. Only half the gametes produced by the heterozygous S0 progenitor would carry the allele of interest. For other characteristics, progenies from the S1 genotypes should be, at worst, similar to those generated by the S0 progenitors. within-family genetic variationpartial inbreedinggenetic gainsrecurrent selectionadditive genetic effectsnon-additive genetic effects ==== Body Introduction Most cassava breeding programs started in the 1970s or later. Ceballos et al. (2012) proposed that the initial progress was actually to finalize the domestication of the crop, i.e., to move from a crop adapted almost exclusively to rustic, low management conditions to one that responds well to more intensive management for productivity. By the 1990s officially released varieties had shown a significant increase (Kawano et al., 1998; Kawano, 2003) in fresh root yield (FRY) and dry matter content (DMC). An outstanding example is KU50, a variety released in Thailand in 1992, and still grown on more than 1 million ha annually in several countries in SE Asia. This variety, along with others released at about the same period, had a significant impact in the livelihoods of millions of resource-limited farmers (Kawano and Cock, 2005; Fu et al., 2014). It has been recently reported that selection alters the relationship between FRY and DMC. The selection process favors genotypes with high dry matter productivity through either high FRY or high DMC, but it is very difficult to find genotypes that are outstanding simultaneously for both traits (Ceballos and Hershey, 2016). However, the impressive genetic progress achieved from 1975 to 1995 has slowed down considerably in the last two decades (1995–2015). Combined analyses of different reports from cassava breeding in Thailand indicate that gains from 1995 to 2015 are at best half of those observed in the previous two decades for FRY and DMC (CIAT, 2007; Ceballos and Hershey, 2016). Similar trends can be observed in Colombia and Brazil. It was expected that biotechnology tools, such as marker assisted selection, would help recover the rate of genetic gains. Molecular biology has been successful in diagnostics for cassava diseases and their genetic diversity (Restrepo and Verdier, 1997; Hernández Pérez et al., 1999; Monger et al., 2001a,b; Álvarez et al., 2003, 2009; Calvert et al., 2008; Legg et al., 2011); gene expression studies in host-pathogen interactions (Hong and Stanley, 1995; Fregene et al., 2004; Kemp et al., 2004, 2005); introgression of resistance to cassava mosaic disease (CMD) in Latin American germplasm (Egesi et al., 2006; Okogbenin et al., 2007), or dissection of the pathway leading to post-harvest physiological deterioration in cassava roots (Reilly et al., 2007). The first molecular map of cassava was first published two decades ago (Fregene et al., 1997). Yet, the only successful applied experience of marker assisted selection in cassava breeding to date has been for resistance to CMD (Fregene et al., 2000; Akano et al., 2002; Rabbi et al., 2014), while impact on increasing FRY has been limited. In spite of these advances in breeding tools, the slowing down in genetic gains for FRY has not been reversed. Breeders continue to aim for high yield, but have also shifted attention to other value-added traits that are easier to breed such as nutritional quality (Ceballos et al., 2013; Maziya-Dixon and Dixon, 2015), starch functional properties (Aiemnaka et al., 2012) or resistance to CMD (Rabbi et al., 2014). Cassava breeders typically apply phenotypic recurrent selection, as is common for clonally propagated crops (Burton, 1992; Grüneberg et al., 2009; Lebot, 2010; Quero-García et al., 2010; Ceballos et al., 2012). Because of the low multiplication rate of cassava from stem cuttings, it takes several years to have enough planting material available for replicated multi-location evaluations, under conventional propagation systems (Ceballos et al., 2004, 2012). A typical selection cycle requires 2 years to produce the progeny (botanical seeds) of planned crosses and 6 consecutive years of field evaluation. Initial phenotypic evaluations are based on unreplicated trials grown in one or, at most, two locations. Critical selection decisions need to be taken during this lengthy process: breeders try to reconcile the practical need to reduce the large number of genotypes in the early stages of selection with the awareness that selection based on unreplicated trials is prone to large experimental errors. Ceballos and co-workers suggested the possibility of using breeding value (e.g., general combining ability) for cassava genetic enhancement based on promising results they had observed using phenotypic data (Ceballos et al., 2004). Falconer (1981) defined breeding value of an individual as the mean value of its progeny, a simple yet powerful concept in plant and animal breeding. The breeding value is the deviation of the progeny generated by a given progenitor from the average of a reference population. Breeding value depends on the average performance of the reference population as well as on the value of the alleles that each progenitor can transfer to its progeny (Falconer, 1981). Typically, breeding value is related to additive genetic effects, although some dominance effects (e.g., a single dominant source of resistance to a given disease or pest) can influence breeding values. Best linear unbiased prediction (BLUP) was originally developed for more accurate estimation of breeding values in animal breeding and has now been widely used in many areas of research including different crops (Henderson, 1975; Pander and Allen, 1995; Bernardo, 2002). However, it seems that it has not gained the same popularity in plant breeding (Piepho et al., 2008). Genomic selection currently under pilot testing in cassava brings hope of a positive impact for enhanced productivity (de Oliveira et al., 2012; Wolfe et al., 2016) and evolved from earlier applications of BLUP (Heffner et al., 2009). Genomic selection is a form of marker assisted selection that sorts individuals out, based on genomic estimated breeding values (Nakaya and Isobe, 2012). Genomic selection relies on the estimation of breeding values for quantitative traits based on whole genome genotypes through the simultaneous estimation of marker effects in a single step (Heslot et al., 2012). The current study consolidates phenotypic data from 14 years of successive trials in a sub-humid tropical environment of Colombia, from more than 20,000 genotypes initially evaluated in single row trials—SRT. The data consolidated, curated and organized for analysis can be accessed at http://dx.doi.org/10.7910/DVN/QB9FUW. The original raw data is also available at https://www.cassavabase.org. The main objectives were, (i) to estimate breeding values of progenitors of the more than 20,000 genotypes initially evaluated; (ii) assess the usefulness of these breeding values for predicting which genotypes eventually reach the most advanced stage of selection (uniform yield trials—UYT), grown in several locations and years, and (iii) attempt to identify factors that affect the probability of clone(s) from a given progenitor to reach the UYT stage. Materials and methods Breeding objectives and selection criteria Breeders apply a wide range of objectives in cassava in response to the diversity of production environments, management practices, and end uses. However, only a few are broadly accepted as common key traits for improvement: FRY; high and stable DMC; suitable plant architecture, and resistance to locally or regionally relevant pests and diseases. At CIAT, in addition to individual ratings, breeders integrate plant architecture and resistance to biotic/abiotic stresses into a single score indicating overall desirability of the above-ground plant appearance (plant type score or PTS) where 1 is very good and 5 is very poor. Because of the low heritability of FRY in early stages of selection, cassava breeders for many years have applied indirect selection for yield by using correlated traits with higher heritabilities, such as harvest index (HIN) (Kawano et al., 1998). CIAT generally applies a selection index (SIN) that integrates these four relevant variables, assigning them best-judgment weight (in italics in the formula below) established by the breeder's experience (Ceballos et al., 2012): SIN=(FRY * 10)+(DMC * 10)-(PTS * 5)+(HIN * 3) In the case of PTS the desired target is a lower score. Therefore, a negative sign is assigned to the respective term in the SIN equation. Evaluation and selection process We obtained botanical seed by controlled (full sibs) or open (half sibs) crossing among outstanding progenitors (all cassava genotypes currently used in breeding are heterozygous). Seed was germinated, seedlings grown for about 2 months in a greenhouse, and then transplanted to the field. The seedling plants (F1) were grown in Palmira, Valle del Cauca, Colombia (CIAT headquarters), which offers fertile soils, moderate temperatures and availability of irrigation—ideal for high cassava productivity. Selection and harvest of plants took place at 9–10 months after transplanting. The only selection criterion applied was the capacity of the plant to produce eight vegetative cuttings (20 cm stem pieces) for the following stage of selection. This step initiated the long process of phenotypic recurrent selection as described below (Figure 1). Figure 1 Illustration of the different stages of a typical evaluation process in cassava breeding. Plants from germinated seed (seedling plants) are grown in the field and used as the source of clonal planting material (left side). The first evaluation takes place in single row trials (SRT), followed by preliminary (PYT) and advanced (AYT) yield trials. The first multi-location evaluation is in the uniform yield trials (UYT), or sometimes earlier, in the AYTs. Size of plots in UYT has been slightly modified to illustrate the effect of different environments on the growth of cassava. Clonal evaluation trials or single row trials (SRTs) This is the first stage where selection for agronomic performance takes place in the sub-humid environment (Caribbean coast of Colombia). The region is characterized by moderate rainfall (800–1200 mm annually) and a long dry season (3–4 months), typical of many cassava-growing regions of the world. Trials usually include about 1000–2000 genotypes, each represented by six to eight plants in a single row (1–2 ha), in a single location. About 150–250 genotypes are selected for the next stage of evaluation. An important feature of SRTs is that, being the first stage of the selection process, information from all progenies (selected or not) is available, thus providing unbiased information about the progenitors used to generate them. Preliminary yield trials (PYTs) Each genotype is represented in three repetitions with 10-plant plots (two rows of five plants). A randomized complete block design is used in all remaining type of trials. All plants in each plot (except the front plant in each row) are harvested. PYTs are planted in a single location. Advanced yield trials (AYT) Plots consist of four (or five) rows and five plants per row, with, three replications. The six (or nine) central plants are harvested to generate the data used in the selection process. AYTs are usually planted in a single location. Uniform yield trials (UYT) This is the final stage in the CIAT-managed evaluation and selection process. Plot size, number of repetitions and planting arrangement is the same as those for AYTs. UYTs are planted for 2 consecutive years in 5–10 locations. Typically UYTs will have 20–25 experimental clones and 5–8 local or commercial checks. Farmer and end user criteria are used during each step of selection, and they are invited to participate for more intensive input and interaction with breeders during the harvest of AYTs and UYTs. In addition, planting material of the most promising clones is shared with key farmers for semi-commercial evaluation. In general, varieties are released by national programs only after successful performance (according to the farmers' and end-users' criteria) in these semi-commercial evaluations (0.5–1.0 ha). Data analysis Data from evaluation trials conducted from 2000 through 2013 were used. Target growing conditions included various sites within the sub-humid environment, the most important cassava growing region in Colombia and in most of the world. This large database was prepared for analysis with SAS (2008). The first step was to consolidate data from different trials grown during successive years into a large megafile. A total of 1038 full- or half-sib families were evaluated in these trials involving a total of 20,229 genotypes evaluated in the SRTs (9108 from full-sib families and 11,221 from half-sib families). Four variables (FRY, DMC, PTS, and HIN) were considered for the analysis and used to estimate a selection index (SIN) for each genotype, using the same weights as in the formula described above. SAS Proc Means procedure was used to obtain the family averages for every trait, including selection index. From the initial number of genotypes evaluated in SRT, 2652 were selected and evaluated in PYT, 567 in AYT and only 114 in UYT. This study concentrates on the data from the first and last stages of selection (SRT and UYT, respectively) and will not consider the intermediate stages. Data from all the individual genotypes belonging to a given full- or half-sib family was consolidated to obtain the respective averages and other statistical parameters for the key variables: FRY, DMC, PTS, HIN, and SIN. Since progenitors are used to generate more than one family, averages for each progenitor across all the families that it had generated were estimated. The phenotypic average of all the progenies (across different families) generated by a given progenitor will be considered as the breeding value of that progenitor. Phenotypic data from the 9108 full-sib genotypes was used twice: for the estimation of breeding values of the progenitor used as female, and for the breeding value when used as a male. Data from each progenitor was not balanced because of lack of a uniform number of progenies evaluated from each progenitor. The number of crosses (e.g., full- or half-sib families) generated from each progenitor was also variable, as was the number of years in which progenies from a given progenitor were involved. It is acknowledged therefore that breeding value as estimated in this study is not as accurate as that obtained, for instance, from a diallel study. However, the estimated breeding values fully agree with the original concept in Falconer described earlier, and are based on actual data generated by an ongoing breeding process. Progenitors represented by fewer than 50 genotypes among the progeny, across all families in which they had been used, were discarded from the analysis. A sample size of <50 individuals was considered too small to properly represent the breeding value of the respective progenitor. The initial number of progenitors (297) was therefore reduced to 107. Results A large dataset was consolidated from the different trials conducted from 2000 to 2013. A total of 20,229 genotypes were evaluated in SRT. Table 1 provides a general description of the 107 progenitors analyzed in this study (after discarding those represented by progenies with fewer than 50 clones). The average size of the progenies from the 107 progenitors was 255. There was wide variation in the sample size for each progenitor (ranging from the minimum required of 50 progenies all evaluated in a single year, through 1350 progenies evaluated across the 14 SRTs). This variation in the number of progenies from each progenitor relates to the highly variable flowering behavior of different cassava genotypes (Ceballos et al., 2012). Some genotypes may flower 3–4 times during a year, whereas others flower only once. In few cases plants may have to be grown for more than a year for them to flower for the first time. Table 1 Progenitors (107) selected in the study and number of progenies (#) derived from them. Progenitor # Progenitor # Progenitor # Progenitor # Progenitor # Progenitor # SM1565-17 1350 SM2773-32 472 C-4 233 SM2775-4 126 GM273-57 92 CM6756-15 68 TAI8 1255 SM1521-10 445 CM4843-1 232 SM1068-10 125 SM1600-4 89 SM2615-25 67 SM1665-2 915 CM4365-3 437 SM737-38 230 CM7514-7 124 SM2619-6 87 SM2772-8 67 CM8027-3 906 KU50 425 SM1427-1 215 SM2621-4 118 VEN167 85 BRA496 65 SM1411-5 861 SM2081-34 418 CM9456-12 211 CM7389-9 117 SM3058-29 79 CM7395-5 65 SM1219-9 795 SM2546-40 403 CM8475-4 210 SM2621-29 115 SM2545-20 78 BRA1107 64 SM805-15 760 SM2772-5 403 SM2619-4 210 SM2546-52 114 SM2923-3 78 SM1152-13 64 CM6754-8 735 SM2620-1 334 SM1778-45 206 SM2769-11 114 SM494-2 77 SM2772-2 63 SM1433-4 718 CM7985-24 326 CM9560-1 204 GM290-50 109 GM259-167 75 C-243 62 CM7514-8 667 SM1789-20 295 SM1210-10 180 CM3306-4 108 SM1282-2 75 CM4574-7 58 CM6758-1 663 SM2546-32 294 TAI1 175 SGB765-2 108 R90 73 COL945 58 CM9067-2 657 SM2780-17 291 SM1637-22 165 SM2775-2 105 SM2546-54 73 CT20-2 54 SM1511-6 647 SM1759-29 290 SM1656-7 159 GM462-4 103 CG1141-1 72 SM1650-7 54 CM523-7 584 SM2629-36 272 SM890-9 155 SM1669-5 103 CM9924-6 72 CM3372-4 52 SM2192-6 536 NGA19 266 SM1201-5 152 SM1669-7 98 CM6756-13 71 SM2619-1 51 SM2782-4 519 SM1657-12 264 SGB765-4 151 SMB2446-2 98 SM2623-1 71 VEN25 51 SM1438-2 489 SM2545-22 240 SM1422-4 129 SM1973-25 93 CM3555-6 69 CM9912-107 50 CM2772-3 481 C-18 235 SM643-17 128 SM2603-9 93 CM7951-5 69 A total of 114 genotypes were evaluated in different UYTs in the sub-humid environment during the 2000–2013 period. A key objective of this study was to identify factors that influence the probability of clone(s) from a given progenitor to reach the UYT stage, taking into consideration that, in vegetatively propagated crops, breeding values can be measured across generations with the same genotypes. Progenitors of the 114 clones that reached UYTs were therefore identified. Only three progenitors (CM4919–1, CM681–2, and SM1565–15) of clones in UYTs were not included in the study because they were represented by fewer than 50 progenies. The progenitors of clones reaching the UYTs that are analyzed in this study are listed in Table 2, along with the number of clones derived from them which reached that stage. There was a large variation in the number of clones in UYTs representing different progenitors. Twenty clones in UYTs had been derived from SM1411–5, suggesting that this progenitor has excellent breeding value. Similarly SM 1665–2, CM 8027–3, CM 9067–2, and CM 7514–8 were progenitors of at least 10 genotypes evaluated in UYTs. On the other hand, 12 progenitors were represented only once by their progenies in UYTs and 66 progenitors were not represented in UYTs at all. Results suggest, therefore, that there were strong differences in the probabilities of progenies from a given progenitor reaching the last stage of selection (Table 2). From the breeding point of view, it would be very useful to explain why progenies from SM1411–5, for example, had a higher chance of reaching the last stage of selection and, conversely, why so many progenitors failed to contribute with any clones in UYTs. Table 2 Progenitors (41) of clones that reached the UYT and were represented by more than 50 progenies in the SRT; and number of clones (#) from each of these progenitors. Progenitor # Progenitor # Progenitor # Progenitor # Progenitor # Progenitor # SM1411-5 20 TAI8 7 SM1759-29 4 KU50 2 SM2629-36 2 SM1210-10 1 SM1665-2 15 SM1521-10 6 SM890-9 4 SM1422-4 2 SM737-38 2 SM1637-22 1 CM8027-3 12 SM1438-2 5 CM523-7 3 SM1511-6 2 CG1141-1 1 SM1657-12 1 CM9067-2 12 SM2192-6 5 SM1219-9 3 SM1656-7 2 CM4574-7 1 SM2081-34 1 CM7514-8 10 SM805-15 5 SM1565-17 3 SM1669-5 2 CM7395-5 1 SM2773-32 1 SM1433-4 8 CM6754-8 4 CM7985-24 2 SM1778-45 2 NGA19 1 SM643-17 1 CM4365-3 7 SM1427-1 4 CM8475-4 2 SM1789-20 2 SM1201-5 1 Selection from SRT, through PYT, AYT, and UYT is based on the SIN that integrates the information of four key variables (FRY, DMC, HIN, and PTS). If the selection index is formulated well, average SIN for the progenies of each progenitor should be the parameter most closely associated with the true breeding value of each progenitor measurable at SRT. Table 3 presents the best and worst ten genotypes, based on the average SIN of their progenies from SRTs. Data from SRTs was used because it takes into consideration information from all progeny derived from a given progenitor, regardless of whether or not they were selected. Average SIN (≈breeding value) of the progenies from SM1411-5 was ranked third-best among the 107 genotypes analyzed and was represented by 861 progenies (a very robust progeny size). Figure 2 presents the relationship between average SIN from each progenitor and the size of their respective progenies. Smaller samples tend to show more extreme variation (e.g., ranging from very high to very poor breeding values), compared to larger samples. This is not surprising as standard deviations of the mean and sample sizes are inversely associated (Steel and Torrie, 1988). The information presented in Figure 2 indicates that breeding value (estimated as average SIN for the progenies of each progenitor) is heavily influenced by the size of the progenies rather that the genetic merit of each progenitor: extreme cases (positive or negative) were only found for progenitors represented by fewer than 200 progenies. Table 3 Average selection index (SIN) of the 10 best and 10 worst progenitors, minimum and maximum SIN, as well as the size of their respective progenies. Progenitor Size SIN (Average) SIN (Minimum) SIN (Maximum) R90 73 27.5 −3.9 58.3 SM2545-20 78 10.6 −31.8 52.8 SM1411-5 861 8.3 −76.7 46.1 SM2780-17 291 8.3 −47.5 53.0 GM462-4 103 7.6 −55.8 43.3 C-18 235 7.3 −45.9 52.6 SM2546-54 73 7.1 −41.0 36.8 SM1521-10 445 7.0 −54.7 48.6 SM2619-1 51 6.3 −26.7 30.2 SM1656-7 159 6.2 −58.8 52.8 CM2772-3 481 −8.2 −63.9 33.1 SM2615-25 67 −8.4 −51.7 19.9 SM494-2 77 −8.4 −51.7 22.4 SM1778-45 206 −9.7 −69.4 36.0 CM6756-13 71 −10.3 −48.0 31.6 SM2623-1 71 −10.4 −53.0 40.0 BRA496 65 −11.1 −46.0 23.0 TAI1 175 −14.7 −68.9 35.4 SM2772-2 63 −17.4 −86.6 27.7 COL945 58 −17.6 −48.1 17.7 Average 255 −0.46 −55.49 41.30 The list is ordered from higher to lower average SIN. Figure 2 Relationship between average selection index (SIN) of each progenitor with the respective number of clones representing them at UYT. Extreme average SIN were observed for progenitors represented by fewer than 200 progenies. Figure 3 illustrates the relationship between average SIN for each progenitor and the number of their respective progenies reaching the UYT stage. The performance of SM 1411–5 is worth highlighting because it was a progenitor in about 20% of the clones reaching the UYT and its average SIN was 8.3, suggesting an association between high and positive SIN and success in deploying progenies in UYT. On the other hand, several progenitors with average SIN above 10 had no clones representing them in UYTs. The regression of number of clones in UYT on average SIN for each progenitor (Figure 3) shows a negligible r2 = 0.05, indicating that breeding value (e.g., average SIN for each progenitor) is not a good predictor of the probabilities of a clone from a given progenitor reaching the UYT. Figure 3 Relationship between average selection index (SIN) of each progenitor with the respective number of clones representing them at UYT (arrow identifies SM 1411-5). In addition to the average SIN values, Table 3 provides the maximum and minimum SIN for the individual clones derived from each progenitor. Maximum SIN values are very relevant because they identify the best genotypes which should be, ultimately, those reaching UYT. One of the problems cassava breeders face is the huge within-family variation arising from the fact that progenitors are heterozygous. That variability (illustrated by the wide range of variation of individual SINs in Table 3) weakens the identity of families and supports the idea that outstanding hybrids can be obtained basically from each and every family (Losada Valle, 2015). The plots presented in Figure 4 describe the relationship between number of families (Figure 4A) and progenies (Figure 4B) per progenitor against the number of clones derived from each progenitor reaching UYT. The r2 value from the regression analysis of number of progenies from a progenitor reaching UYT on the total number of progenies per progenitor (0.48) was considerably better than the same parameter from a regression based on average SIN in Figure 3 (0.05). Number of families generated by each progenitor was also a better predictor (r2 = 0.40) than average SIN of the probability of its progeny reaching UYT. Figure 4 Number of families (A) and progenies (B) per progenitor against the number of clones derived from each progenitor reaching UYT (arrow identifies SM 1411-5). It is clear that family size, as expected, strongly influences the results of this study. The initial analysis arbitrarily set a minimum family size n = 50. This number was a reasonable starting point (it was rendered to be large enough to properly represent the breeding potential of each progenitor, but not too large to reduce the total number of progenitors analyzed) but, nonetheless it was arbitrary. Therefore, an exercise was made to analyze the relationship between average SIN at SRT and the probability of progenies reaching UYTs using different family sizes. Figure 5 presents the results of this exercise. The coefficient of determination increased linearly from negligible (when family size < 50) to values larger than 0.25 (when family size > 250). Family size > 300 provided much larger coefficient of determination (>0.45). Results presented in Figure 5 make sense: larger samples of progenies from a given progenitor are expected to provide more reliable information than smaller samples. It is not surprising that a large family size (e.g., 250 genotypes) is required to somewhat predict the chances of one of its members reaching the UYT stage. This is a reflection of the large within-family genetic variability generated from the heterozygous progenitors used in cassava breeding (Ceballos et al., 2015). Families larger than 400 were not considered as the number of progenitors that met this requirement would have been drastically reduced. Figure 5 Coefficient of determination for the regression of average SIN at SRT on number of progenies reaching the UYT stage considering different family sizes. Discussion This study focuses on data from the extremes of the selection process—from the earliest (SRT) to the last stage (UYT). Between these two steps, however, are the PYT and AYT stages. It has been suggested that the phenotypic performance of individual genotypes may “evolve” through the different stages of selection. Epigenetic effects and the impact that biotic and abiotic factors have in the quality of planting material may affect differentially the performance of different genotypes through time (Ceballos and Hershey, 2016; Joaqui et al., in review). This can partially explain the poor association between average SIN at SRT and probabilities of a progenitor being represented in UYT depicted in Figure 3. The large within-family genetic variation in cassava is another factor explaining that poor association (Supplementary Image). The implementation of new genomic tools can contribute to our understanding of the differences in breeding values suggested by data in Tables 2, 3. For the implementation of genomic selection, however, it would be advisable to use phenotypic data from later stages of selection, once the phenotypic performance of each genotype has “stabilized.” Phenotypic recurrent selection in cassava has the advantage that the cloned genotypes can be evaluated and selected many times in different locations and growing seasons. The gradual selection, through four different stages (SRT, PRY, AYT, and UYT) allows the selection of genotypes that have shown consistently outstanding performances. Data from SRT is of particular relevance because it offers unbiased information about the progenitors, i.e., data from all progenies, selected or not. Although SRT data is prone to large experimental errors (single plot at one location and usually large environmental variation in the evaluation sites), averages across many genotypes tend to provide more robust information. Selection of progenitors based on their general combining ability or breeding value in cassava, originally proposed by Ceballos et al. (2004) is further supported by the large variability in number of clones at UYT representing each progenitor (Table 2). It is clear that certain genotypes are better progenitors than others. The fact that 20 out of 114 clones reaching UYTs were derived from SM 1411–5 is a convincing evidence for this statement. SM 1665–2; CM 8027–3; CM 9067–2; CM 7514–8, SM 1433–4, CM 4365–3; and MTAI 8 also were well represented by their progenies in UYTs. However, the average SIN from these progenitors was not outstanding (they were not among the best 10 progenitors), except for SM 1411–5 (Table 3). On the other hand, the average number of progenies from all these genotypes was 802 (ranging from 437 to 1255), well above the average across all progenitors (255). The best predictor for the probability of the progeny of a given clone to reach the UYT seems to be the number of progenies derived from it that are evaluated in SRTs. This is of little help for breeders. It is recognized that the large variation in the number of progenies evaluated from each of the progenitor in this study is a weakness. On the other hand, this reflects the dynamics in any cassava breeding program. It is easy to obtain botanical seed from certain clones and difficult from others. The reproductive biology of cassava will prevail over efforts made to balance the number of progenies from each genotype. The ongoing research to develop a protocol for the induction of flowering (Next Generation Cassava Breeding project, www.nextgencassava.org) will facilitate achieving a more balanced number of progenies from each progenitor. The idea that “good hybrids can be obtained from almost every family” (assuming that parents are basically adapted to the broad biotic and abiotic conditions of the target environment) arises from the large within-family segregations that breeders observe in their nurseries, particularly for traits such as FRY. It is this large within-family variation, however, that weakens the usefulness of breeding value in cassava. It is the best clone(s) within each family that may eventually reach UYTs and it is the identification of that particular clone that is difficult and expensive. The use of homozygous progenitors in cassava would lead to a reduction of within-family genetic variation, in fact to zero unless there existed some residual heterozygosity (Ceballos et al., 2015). However, it is currently difficult to produce inbred genotypes in cassava. Successive self-pollinations are time consuming and favor the selection of early flowering genotypes with profuse branching architecture. Progress in the development of a protocol for the production of doubled haploids has been made (Perera et al., 2013) but is not yet routinely feasible. In the meantime, an alternative option is the use of partially inbred progenitors (e.g., S1 or S2 genotypes). This approach would reduce considerably the within-family genetic variation and in turn help breeders to more easily identify the true breeding value of these progenitors. Inbreeding depression is prevalent for FRY but not so much for traits such as plant height and traits related to above ground biomass (Rojas et al., 2009; Kawuki et al., 2011; de Freitas et al., 2016). Partial inbreeding would not only contribute to identifying more clearly the breeding value of progenitors but it could also be the way to improve it (Kaweesi et al., 2016). Figure 6 illustrates this concept. For example, resistance to CMD has been linked to a single dominant factor (Rabbi et al., 2014). If an S1 genotype homozygous for the resistance to CMD was used (CC in Figure 6) instead of the (putatively) heterozygous S0 progenitor from which it was derived, its breeding value would double (e.g., 100% of the progenies rather than 50% of the progenies would be resistant to CMD). This concept is described on the left side of Figure 6. In addition to homozygous resistance to CMD, segregating S1 genotypes would be selected for agronomic performance as well. Similarly a “complementary” breeding population may be developed for increased levels of DMC (right side of Figure 6). The idea of “complementary” populations has been successfully implemented in commercial vegetables breeding (Knapp, personal communication). One population for example can be the source for defensive traits, while the other would provide desirable quality traits to the resulting hybrids. Figure 6 Illustration of the way breeding value could be consistently improved in a stepwise fashion in two “complementary” breeding populations. Squares are used for S0 genotypes, whereas circles are used for partial inbreds. On the left, selections are made for resistance to CMD. Molecular markers can be used to distinguish homozygous [CC] from heterozygous [Cc] genotypes (❶). In addition to homozygous resistance to CMD, segregating S1 genotypes are selected for agronomic performance (❷). Diameters of the circles (or size of squares for S0) in both left and right diagrams represent levels of DMC (larger circles or squares, higher DMC). On the right, selections in the “complementary” population are made for increased dry matter content (❸). This population does not carry resistance to CMD so the genotype for this trait [cc] has not been included in every genotype. The selected products (S1 genotypes) from these first steps of selection are shaded. Both products, however, are susceptible to a target herbicide. In a parallel process (perhaps from a partner), S1 genotypes homozygous for a recessive source for tolerance to a herbicide have been generated (❹). The S1 genotypes selected for resistance to CMD or high DMC are then crossed with the source of tolerance to herbicides. The resulting crosses will be heterozygous for monogenic traits and intermediate for DMC. Self-pollination of the resulting crosses will allow the recovery of S1 genotypes that are homozygous for CMD and for tolerance to the herbicide (left side), or have improved levels of DMC combined with tolerance to the herbicide (right side). The second-step products are also shaded. Crossing the second-step products generate progenies that are 100% resistant to CMD [Cc], and tolerant to the herbicide [hh] and have excellent levels of DMC. Let's assume that a new recessive source of tolerance to a given herbicide has been identified. The source of tolerance is already partially inbred and homozygous (hh) for tolerance to the herbicide. The initial products (e.g., S1 genotypes) from the first step of selection in the two complementary populations presented in Figure 6 are susceptible (HH) to the herbicide. The S1 genotypes selected for resistance to CMD or high DMC are then crossed with the source of tolerance to herbicide. The resulting crosses will be heterozygous for the monogenic traits and intermediate for DMC. Self-pollination of the resulting crosses will allow the recovery of S1 genotypes that are homozygous for CMD and for tolerance to the herbicide (left side of Figure 6) or have improved levels of DMC combined with tolerance to the herbicide (right side of Figure 6). Crossing among the second-step products generates progenies that are 100% resistant to CMD (Cc), tolerant to the herbicide (hh), and have excellent levels of DMC. The key principle here is that the gametes produced by the selected S1 genotypes should carry a higher frequency of desirable alleles. This is clearly the case for the traits these genotypes had been selected for (e.g., resistance to CMD). For other traits the frequency of desirable alleles at worst should be (on average), similar in the S1 genotypes and in the S0 progenitors from which they were derived. More likely, however, for other traits the frequency of desirable alleles should be higher because deleterious factors (e.g., albino plants) exposed in partially inbred genotypes would be eliminated. Crosses among the selected partially inbred lines, because of their enhanced breeding value, will generate (on average) better performing hybrids. A second and fundamental advantage of the proposed scheme is that it allows for the gradual, consistent, stepwise fixation of simply inherited traits in the partially inbred selected genotypes. Eventually, partially inbred lines from different heterotic groups (when identified or developed) would allow the implementation of conventional reciprocal recurrent selection schemes. There are several traits in cassava that have relatively simple inheritance and would be easy to fix through (partial) inbreeding. For root quality traits, carotenoids, and DMCs; amylose-free starch and small starch granules have been reported to have high heritabilities or to depend on single recessive genes. Resistance to pests and diseases (thrips and whiteflies, bacterial blight, super-elongation disease, CMD) and plant architecture traits (erect vs. branching types) have simple inheritance. Certainly another group of traits that would benefit from partial inbreeding are those arising from genetic transformation and gene editing (e.g., herbicide tolerance). Future advances in our knowledge of plant biology (particularly from Arabidopsis) will foster the need and intensity of trait introgression as they are identified in cassava. The reduced within-family variation in progenies derived from partially inbred parents could also contribute toward improvement in more complex traits such as FRY. Results from this study highlight some key features of cassava breeding. There is a need to shift the current system based on crossing elite germplasm hoping to identify even better progenies, into a system based on the improvement of progenitors with enhanced breeding values, through partial or full inbreeding. This will improve the efficiency of cassava breeding and increase the likelihood of sustained and predictable genetic gains. Author contributions HC implemented the changes in the breeding process that generated the phenotypic data analyzed in this article. He made the analyses and wrote the manuscript; JP was an associate to the breeding program and curated and stored data year after year; OJ conducted a 1-year internship at the program, retrieved the stored data and organized it for its analysis; JL conducted the trials at the sub-humid environment; NM coordinated the production of segregating progenies and the seedling nurseries from which the planting material for the SRT was generated; FC is a senior associate of the program that helped in the overall activities of the program; LP is an assistant in charge of data uploading and curation; CH is the coordinator of the program and also a senior cassava breeding. He reviewed and improved earlier versions of the 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. Supplementary material The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2016.01227 Supplementary Image Breeding values for each progenitor. Click here for additional data file. ==== Refs References Aiemnaka P. A. Wongkaew J. Chanthaworn S. K. Nagashima S. Boonma J. Authapun S. (2012 ). Molecular characterization of a spontaneous waxy starch mutation in cassava (Manihot esculenta Crantz) . Crop Sci. 52 , 2121 –2130 . 10.2135/cropsci2012.01.0058 Akano A. O. Dixon A. G. O. Mba C. Barrera E. Fregene M. (2002 ). Genetic mapping of a dominant gene conferring resistance to cassava mosaic disease . Theor. Appl. Genet. 105 , 521 –525 . 10.1007/s00122-002-0891-7 12582500 Álvarez E. Mejía J. F. Llano G. A. Loke J. B. Calari A. Duduk B. (2009 ). Characterization of a phytoplasma associated with frogskin disease in cassava . Plant Dis. 93 , 1139 –1145 . 10.1094/PDIS-93-11-1139 Álvarez E. Mejía J. F. Valle T. (2003 ). Molecular and pathogenicity characterization of Sphaceloma manihoticola isolates from south-central Brazil . Plant Dis. 87 , 1322 –1328 . 10.1094/PDIS.2003.87.11.1322 Bernardo R. (2002 ). Best linear unbiased prediction , in Breeding for Quantitative Traits in Plants , ed Bernardo R. (Woodbury, MN : Stemma Press ), 211 –238 . Burton G. W. (1992 ). Recurrent restricted phenotypic selection . Plant Breed. Rev. 9 , 101 –113 . 10.1002/9780470650363.ch5 Calvert L. A. Cuervo M. Lozano I. Villareal N. Arroyave J. (2008 ). Identification of three strains of a virus associated with cassava plants affected by Frogskin disease . J. Phytopathol. 156 , 647 –653 . 10.1111/j.1439-0434.2008.01412.x Ceballos H. Hershey C. (2016 ). Road map for cassava genetic improvement , in Proceedings of the World Congress on Root and Tuber Crops and Third Scientific Conference of the Global Cassava Partnership for the 21st Century (Nanning ). Available online at: http://www.gcp21.org/wcrtc/PS08.html Ceballos H. Hershey C. Becerra-López-Lavalle L. A. (2012 ). New approaches to cassava breeding . Plant Breed. Rev. 36 , 427 –504 . 10.1002/9781118358566.ch6 Ceballos H. Iglesias C. A. Pérez J. C. Dixon A. G. O. (2004 ). Cassava breeding: opportunities and challenges . Plant Mol. Biol. 56 , 503 –515 . 10.1007/s11103-004-5010-5 15630615 Ceballos H. Kawuki R. S. Gracen V. E. Yencho G. C. Hershey C. H. (2015 ). Conventional breeding, marker assisted selection, genomic selection and inbreeding in clonally propagated crops: a case study for cassava . Theor. Appl. Genet. 9 , 1647 –1667 . 10.1007/s00122-015-2555-4 26093610 Ceballos H. Morante N. Sánchez T. Ortiz D. Aragón I. Chávez A. L. (2013 ). Rapid cycling recurrent selection for increased carotenoids content in cassava roots . Crop Sci. 53 , 2342 –2351 . 10.2135/cropsci2013.02.0123 Centro Internacional de Agricultura Tropical (CIAT) (2007 ). A new future for cassava in asia: its use as food, feed and fuel to benefit the poor , in Proceedings of the Eighth Regional Workshop (Vientiane; Cali : LAO PDR ). de Freitas J. P. X. da Silva Santos V. de Oliveira E. J. (2016 ). Inbreeding depression in cassava for productive traits . Euphytica 209 , 137 –145 . 10.1007/s10681-016-1649-7 de Oliveira E. J. Vilela de Resende M. D. da Silva Santos V. Fortes Ferreira C. Alvarenga Fachardo Oliveira G. Suzarte da Silva M. (2012 ). Genome-wide selection in cassava . Euphytica 187 , 263 –276 . 10.1007/s10681-012-0722-0 Egesi C. N. Ogbe F. O. Akoroda M. Ilona P. Dixon A. (2006 ). Resistance profile of improved cassava germplasm to cassava mosaic disease in Nigeria . Euphytica 155 , 215 –224 . 10.1007/s10681-006-9323-0 Falconer D. S. (1981 ). Introduction to Quantitative Genetics, 2nd Edn. New York, NY : Longman Inc. Fregene M. Angel F. Gomez R. Rodríguez F. Chavarriaga P. Roca W. (1997 ). A molecular genetic map of cassava (Manihot esculenta Crantz) . Theor. Appl. Genet. 95 , 431 –441 . 10.1007/s001220050580 Fregene M. Bernal A. Duque M. Dixon A. Tohme J. (2000 ). AFLP analysis of African cassava (Manihot esculenta Crantz) germplasm resistant to the cassava mosaic disease (CMD) . Theor. Appl. Genet. 100 , 678 –685 . 10.1007/s001220051339 Fregene M. Matsumura H. Akano A. Dixon A. Terauchi R. (2004 ). Serial analysis of gene expression (SAGE) of host–plant resistance to the cassava mosaic disease (CMD) . Plant Mol. Biol. 56 , 563 –571 . 10.1007/s11103-004-3477-8 15630620 Fu Y.-B. Wangsomnuk P. P. Ruttawat B. (2014 ). Thai elite cassava genetic diversity was fortuitously conserved through farming with different sets of varieties . Conserv. Genet. 15 , 1463 –1478 . 10.1007/s10592-014-0631-y Grüneberg W. Mwanga R. Andrade M. Espinoza J. (2009 ). Breeding clonally propagated crops , in Plant Breeding and Farmer Participation , eds Ceccarelli S. Guimarães E. P. Weltzien E. (Rome : Food and Agriculture Organization ), 175 –322 . Heffner E. L. Sorrells M. E. Jannink J.-L. (2009 ). Genomic selection for crop improvement . Crop Sci. 49 , 1 –12 . 10.2135/cropsci2008.08.0512 Henderson C. R. (1975 ). Best linear unbiased estimation and prediction under a selection model . Biometrics 31 , 423 –477 . 10.2307/2529430 1174616 Hernández Pérez R. Bermúdez Pérez D. González Ramírez J. Pairol Martínez A. (1999 ). Detección del RNA viral en muestras de yuca del clon CMC40 . Centro. Agrícola. 26 , 61 –66 . Heslot N. Yang H.-P. Sorrells M. E. Jannink J.-L. (2012 ). Genomic selection in plant breeding: a comparison of models . Crop Sci. 52 , 146 –160 . 10.2135/cropsci2011.06.0297 Hong Y. Stanley J. (1995 ). Regulation of African cassava mosaic virus complementary-sense gene expression by N-terminal sequences of the replication-associated protein AC1 . J. Gen. Virol. 76 , 2415 –2422 . 10.1099/0022-1317-76-10-2415 7595345 Kawano K. (2003 ). Thirty years of cassava breeding for productivity – biological and social factors for success . Crop Sci. 43 , 1325 –1335 . 10.2135/cropsci2003.1325 Kawano K. Cock J. H. (2005 ). Breeding cassava for underprivileged: institutional, socio-economic and biological factors for success . J. Crop Improv. 14 , 197 –219 . 10.1300/J411v14n01_09 Kawano K. Narintaraporn K. Narintaraporn P. Sarakarn S. Limsila A. Limsila J. (1998 ). Yield improvement in a multistage breeding program for cassava . Crop Sci. 38 , 325 –332 . 10.2135/cropsci1998.0011183X003800020007x Kaweesi T. Kyaligonza V. Baguma Y. Kawuki R. Ferguson M. (2016 ). Inbreeding enhances field resistance to cassava brown streak viruses . J. Plant Breed. Crop. Sci. 8 , 138 –149 . 10.5897/JPBCS2015.0555 Kawuki R. S. Nuwamanya E. Labuschagne M. T. Herselman L. Ferguson M. E. (2011 ). Segregation of selected agronomic traits in six S1 cassava families . J. Plant Breed. Crop Sci. 3 , 154 –160 . Kemp B. P. Beeching J. R. Cooper R. M. (2005 ). cDNA-AFLP reveals genes differentially expressed during the hypersensitive response of cassava . Mol. Plant Pathol. 6 , 113 –123 . 10.1111/j.1364-3703.2005.00268.x 20565643 Kemp B. P. Horne J. Bryant A. Cooper R. M. (2004 ). Xanthomonas axonopodis gumD gene is essential for EPS production and pathogenicity and enhances epiphytic survival on cassava (Manihot esculenta) . Physiol. Mol. Plant Pathol. 64 , 209 –218 . 10.1016/j.pmpp.2004.08.007 Lebot V. (2010 ). Sweet potato , in Root and Tuber Crops , ed Bradshaw J. (New York, NY : Springer Publishers ), 97 –125 . Legg J. P. Jeremiah S. C. Obiero H. M. Maruthi M. N. Ndyetabula I. Okao-Okuja G. . (2011 ). Comparing the regional epidemiology of the cassava mosaic and cassava brown streak virus pandemics in Africa . Virus Res. 159 , 161 –170 . 10.1016/j.virusres.2011.04.018 21549776 Losada Valle T. (2015 ). Mesa redonda: avanços para o melhoramento da mandioca , in XVI Congresso Brasileiro de Mandioca/I Congresso Latino-Americano e Caribenho de Mandioca (Foz de Iguaçu ). Maziya-Dixon B. B. Dixon A. G. O. (2015 ). Carotenoids content of yellow-fleshed cassava genotypes grown in four agroecological zones in Nigeria and their Retinol Activity Equivalents (RAE) . J. Food Agric. Environ. 13 , 63 –69 . Monger W. A. Seal S. Cotton S. Foster G. D. (2001a ). Identification of different isolates of Cassava brown streak virus and development of a diagnostic test . Plant Pathol. 50 , 527 –534 . 10.1046/j.1365-3059.2001.00589.x Monger W. A. Seal S. Isaac A. M. Foster G. D. (2001b ). Molecular characterization of the Cassava brown streak virus coat protein . Plant Pathol. 50 , 527 –534 . 10.1046/j.1365-3059.2001.00589.x Nakaya A. Isobe S. N. (2012 ). Will genomic selection be a practical method for plant breeding? Ann. Bot. 110 , 1303 –1316 . 10.1093/aob/mcs109 22645117 Okogbenin E. Porto M. C. M. Egesi C. Espinosa E. Santos L. G. Ospina C. (2007 ). Marker-assisted introgression of resistance to cassava mosaic disease into Latin American germplasm for the genetic improvement of cassava in Africa . Crop Sci. 47 , 1895 –1904 . 10.2135/cropsci2006.10.0688 Pander D. M. Allen F. L. (1995 ). Using best linear unbiased prediction to enhance breeding for yield in Soybean: I . choosing parents. Crop Sci. 35 , 397 –404 . 10.2135/cropsci1995.0011183X003500020020x Perera P. I. P. Ordoñez C. A. Becerra López-Lavalle L. A. Dedicova B. (2013 ). A milestone in the doubled haploid pathway of cassava . Protoplasma 251 , 233 –246 . 10.1007/s00709-013-0543-6 24026343 Piepho H. P. Möhring J. Melchinger A. E. Büchse A. (2008 ). BLUP for phenotypic selection in plant breeding and variety testing . Euphytica 161 , 209 –228 . 10.1007/s10681-007-9449-8 Quero-García J. Ivancic A. Lebot V. (2010 ). Taro and cocoyam , in Root and Tuber Crops , ed Bradshaw J. (New York, NY : Springer Publishers ), 149 –172 . Rabbi I. Y. Hamblin M. T. Lava Kumar P. Gedil M. A. Ikpan A. S. Jannink J. L. (2014 ). High-resolution mapping of resistance to cassava mosaic geminiviruses in cassava using genotyping-by-sequencing and its implications for breeding . Virus Res. 186 , 87 –96 . 10.1016/j.virusres.2013.12.028 24389096 Reilly K. Bernal D. Cortes D. F. Gomez-Vasquez R. Tohme J. Beeching J. R. (2007 ). Towards identifying the full set of genes expressed during cassava post-harvest physiological deterioration . Plant Mol. Biol. 64 , 187 –203 . 10.1007/s11103-007-9144-0 17318318 Restrepo S. Verdier V. (1997 ). Geographical differentiation of the population of Xanthomonas axonopodis pv. manihotis in Colombia . Appl. Environ. Microb. 63 , 4427 –4434 . 16535731 Rojas C. M. Pérez J. C. Ceballos H. Baena D. Morante N. Calle F. (2009 ). Introduction of inbreeding and analysis of inbreeding depression in eight S1 cassava families . Crop Sci. 49 , 543 –548 . 10.2135/cropsci2008.07.0419 SAS (2008 ). SAS/STAT 9.1 User's Guide . Cary, NC : SAS Institute Inc . Steel R. G. D. Torrie J. H. (1988 ). Bioestadística: Principios y Procedimientos, 2nd Edn. First in Spanish . Mexico : McGraw-Hill Naucalpan de Juárez . Wolfe M. D. Egesi C. Kawuki R. Kulakow P. Rabbi I. Y. Jannink J.-L. (2016 ). Next generation cassava breeding: progress implementing genomic selection , in Proceedings of the World Congress on Root and Tuber Crops and Third Scientific Conference of the Global Cassava Partnership for the 21st Century (Nanning ). Available online at: http://www.gcp21.org/wcrtc/PS08.html

PMC005xxxxxx/PMC5003058.txt

==== Front Front Plant SciFront Plant SciFront. Plant Sci.Frontiers in Plant Science1664-462XFrontiers Media S.A. 10.3389/fpls.2016.01292Plant ScienceOriginal ResearchTwo Novel Flavin-Containing Monooxygenases Involved in Biosynthesis of Aliphatic Glucosinolates Kong Wenwen 1†Li Jing 1†Yu Qingyue 1Cang Wei 1Xu Rui 1Wang Yang 2Ji Wei 1*1Department of Plant Biotechnology, College of Life Science, Northeast Agricultural UniversityHarbin, China2Department of Cell Biology, College of Life Science, Northeast Forestry UniversityHarbin, ChinaEdited by: Basil J. Nikolau, Iowa State University, USA Reviewed by: Masami Yokota Hirai, RIKEN Center for Sustainable Resource Science, Japan; Rita Maria Zrenner, Leibniz Institute of Vegetable and Ornamental Crops (LG), Germany *Correspondence: Wei Ji, iwei_ji@neau.edu.cn†These authors have contributed equally to this work. This article was submitted to Plant Metabolism and Chemodiversity, a section of the journal Frontiers in Plant Science 29 8 2016 2016 7 129229 4 2016 12 8 2016 Copyright © 2016 Kong, Li, Yu, Cang, Xu, Wang and Ji.2016Kong, Li, Yu, Cang, Xu, Wang and JiThis is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.Glucosinolates, a class of secondary metabolites from cruciferous plants, are derived from amino acids and have diverse biological activities, such as in biotic defense, depending on their side chain modification. The first structural modification step in the synthesis of aliphatic (methionine-derived) glucosinolates—S-oxygenation of methylthioalkyl glucosinolates to methylsulfinylalkyl glucosinolates—was found to be catalyzed by five flavin-containing monooxygenases (FMOs), FMOGS-OX1-5. Here, we report two additional FMOGS-OX enzymes, FMOGS-OX6, and FMOGS-OX7, encoded by At1g12130 and At1g12160, respectively. The overexpression of both FMOGS-OX6 and FMOGS-OX7 decreased the ratio of methylthioalkyl glucosinolates to the sum of methylthioalkyl and methylsulfinylalkyl glucosinolates, suggesting that the introduction of the two genes converted methylthioalkyl glucosinolates into methylsulfinylalkyl glucosinolates. Analysis of expression pattern revealed that the spatial expression of the two genes is quite similar and partially overlapped with the other FMOGS-OX genes, which are primarily expressed in vascular tissue. We further analyzed the responsive expression pattern of all the seven FMOGS-OX genes to exogenous treatment with abscisic acid, 1-aminocyclopropane-1-carboxylic acid (ACC), jasmonic acid (JA), salicylic acid, indole-3-acetic acid (IAA), and low and high temperatures. Although these genes showed same tendency toward the changing stimulus, the sensitivity of each gene was quite different. The variety in spatial expression among the FMOGS-OX genes while responding to environmental stimulus indicated a complex and finely tuned regulation of glucosinolates modifications. Identification of these two novel FMOGS-OX enzymes will enhance the understanding of glucosinolates modifications and the importance of evolution of these duplicated genes. Arabidopsis thalianaflavin-containing monooxygenase (FMO)glucosinolatesS-oxygenationexpression patternNational Natural Science Foundation of China10.13039/50110000180931370334, 31170368 ==== Body Introduction Glucosinolates (GSLs) are amino acid-derived natural products primarily present in plants belonging to the Brassicaceae family, such as cabbage, broccoli, and the model plant Arabidopsis thaliana. Upon wounding or mastication, GSLs are hydrolyzed by myrosinases (thioglucosidases) and release hydrolysis products, primarily isothiocyanates and nitriles (Halkier and Gershenzon, 2006; Bones and Rossiter, 2006; Zhang et al., 2006). These breakdown products exert diverse biological effects such as induction of direct toxic effects or other defense responses against pathogens and generalist herbivores (Bednarek et al., 2009; Clay et al., 2009; Hopkins et al., 2009; Laluk et al., 2012). In humans, the isothiocyanates derived from some aliphatic (methionine-derived) GSLs are considered to have numerous health benefits including potent anti-cancer property (Fahey et al., 1997, 2002; Rose et al., 2000; Zalcmann and Talalay, 2001). The most well-studied isothiocyanate is sulforaphane, which is derived from 4-methylsulfinylbutyl GSL; it can decrease the risk of various cancers such as breast cancer, prostate cancer, gastric cancer, and skin cancer (Fahey et al., 2002; Rose et al., 2005; Talalay et al., 2007; Atwell et al., 2015; Chang et al., 2015; Fofaria et al., 2015). Aliphatic GSLs are synthesized in three steps—elongation of methionine chain, formation of GSL core structure, and modification of side chain. The diverse biological activities of GSLs are largely dependent on the chemical modifications in their side chain. The basic aliphatic glucosinolate molecules, methylthioalkyl (MT) GSLs, which contain only the core structure, may undergo a variety of secondary modifications such as oxidation, hydroxylation, or desaturation. The first step in side-chain modification is S-oxygenation, which converts MT GSLs to methylsulfinylalkyl (MS) GSLs. Thus, S-oxygenation is of biological as well as biochemical interest because it influences not only the biosynthesis of MS GSLs, but also the further modification and the resulting activity of their hydrolysis products. Five flavin-containing monooxygenases (FMOs), FMOGS-OX1-5, have been identified to catalyze S-oxygenation during the conversion of MT GSLs to MS GSLs (Hansen et al., 2007; Li et al., 2008). In phylogenetic analysis of FMOs containing the conserved FMOGS-OX domains in Arabidopsis, At1g12130 and At1g12160 clustered to the same subclade and, together with FMOGS-OX1-5, formed a seven-protein group (Figure 1A). In a previous study, a phylogenetic tree of plant-derived FMOs in rice, Arabidopsis, and poplar was analyzed, and this seven-protein group was involved in the predicted S-oxygenation clade and was present only in the GSL-producing Arabidopsis plant. Thus, these genes were predicted to be involved in the S-oxygenation of GSL (Figure 1B). Further, another FMO-coding gene, At1g12200, was suggested as a candidate FMOGS-OX due to its strong co-expression with other GSL biosynthetic genes in response to Sclerotinia sclerotiorum (Stotz et al., 2011) infection. Therefore, in our study, we investigated At1g12130, At1g12160, and At1g12200 as FMOGS-OX candidate genes. It was discovered that both At1g12130 (named FMOGS-OX6) and At1g12160 (named FMOGS-OX7) are involved in S-oxygenation of aliphatic GSLs, while At1g12200 does not possess the expected activity. FIGURE 1 (A) Phylogenetic tree of Arabidopsis flavin-containing monooxygenases (FMOs) possessing the conserved GS-OX domains. (B) FMOGS-OX6 and FMOGS-OX7 are predicted to catalyze S-oxygenation of methylthioalkyl (MT) Glucosinolates (GSLs) to methylsulfinylalkyl (MS) GSLs. To better understand the coordination of these FMOGS-OXs and the upstream aliphatic GSL biosynthetic genes, responsive expression pattern of these genes toward several hormone treatments and environmental conditions are detected. Integrating the knowledge of catalysis, distribution and responsive pattern of these FMOGS-OXs, we speculate that on one hand, these duplicated genes are redundant and might functionally compensate and increase genetic robustness, whereas on the other hand, they show delicate functional variation and possibly contribute to the complex and finely tuned regulation of GSLs modification. Materials and Methods Generation of Phylogenetic Tree All the FMO protein sequences with conserved GS-OX domains in Arabidopsis were obtained from NCBI (Marchler-Bauer et al., 2015) and the full-length amino acid sequences of these proteins were aligned using ClustalX with default parameters. A phylogenetic tree was constructed using the neighbor-joining method with Mega software (version 5.0). Bootstrap support for the topology was estimated from 1000 replicates, and nodes occurring in less than 50% of the replicates were collapsed. Plant Material and Growth Conditions Wild-type (Col-0) and mutants of Arabidopsis seeds were obtained from Arabidopsis Biological Resource Center. Genotype of FMOGS-OX6 T-DNA insertion mutant (SALK_209008) was analyzed by PCR using the following three primers, 5′-CCGCCTAATTCAATCAGCTC-3′, 5′-GAGGCTCAGTGAGATGACC-3′, and the T-DNA-specific primer LBa1 5′-TGGTTCACGTAGTGGGCCATCG-3′. Genotype of FMOGS-OX7 T-DNA insertion mutant (SALK_062129) was analyzed by PCR using the following three primers, 5′-AAACCGTGAAGATTGCAATTG-3′, 5′-GACGAGGCTCAGTGAAGTGAC-3′, and the T-DNA-specific primer LBa1 5′-TGGTTCACGTAGTGGGCCATCG-3′. Genotype of At1g12200 T-DNA insertion mutant (SALK_136282) was analyzed by PCR using the following three primers, 5′-CCGCCTAATTCAATCAGCTC-3′, 5′-GAGGCTCAGTGAAGTGACC-3′, and the T-DNA-specific primer LBa1 5′-TGGTTCACGTAGTGGGCCATCG-3′. Plants were grown in a growth chamber with 16 h light/8 h dark photoperiod at a photosynthetic photon flux density of 100 μmolm-2 s-1 at 20 and 70% relative humidity, respectively. Plasmid Construction and Plant Transformation To prepare the overexpression constructs 35S::FMOGS-OX6, 35S::FMOGS-OX7, and 35S::At1g12200, coding sequences (CDS) of FMOGS-OX6, FMOGS-OX7 and At1g12200 were amplified through reverse transcription-PCR (RT-PCR) method. The total RNA from plant was isolated with TRIzol reagent according to the manufacturer’s instructions. First-strand cDNA was synthesized using the SupermoIII RT kit (Bioteke). The CDS of FMOGS-OX6, FMOGS-OX7 and At1g12200 were amplified with Pfu Turbo Cx Hotstart DNA polymerase (Stratagene) from the first-strand products using the following primers, respectively: 5′-GGCTTAAUATGACACCACCGCCTAATTC-3′, 5′-GGTTTAAUTTATGTAGACCAAACTTTGCTCG-3′ and 5′-GGCTTAAUATGACCAGCGTAATCACCTC-3′, 5′-GGTTTAAUCTAAGAAGAACAATCTTGGTTCG-3′, and 5′-ATGGCAACGAGTCATCCTGA-3′, 5′-TTAGGTTTTGAGCATCGGCAAAAG-3′. The PCR products were then cloned into the expression vector pCAMBIA230035Su using the USER Cloning method as described by Nour-Eldin et al. (2006). To identify the spatial expression pattern of FMOGS-OX6 and FMOGS-OX7, two constructs FMOGS-OX6Pro::GUS (GUS was driven by the promoter of FMOGS-OX6) and FMOGS-OX7Pro::GUS (GUS was driven by the promoter of FMOGS-OX7), respectively, were prepared. For FMOGS-OX6Pro::GUS, a 999bp DNA fragment containing the FMOGS-OX6 promoter was amplified with Pfu Turbo Cx Hotstart DNA polymerase (Stratagene) from genomic DNA using the PCR primers 5′-GGCTTAAUGAGTGGTTAATGTGCAACATCAGC-3′ and 5′-GGTTTAAUAGTATCAGTCAAAGTATTTGTTTCCTCG-3′. For FMOGS-OX7Pro::GUS, a 417 bp DNA fragment containing the FMOGS-OX7 promoter was amplified with Pfu Turbo Cx Hotstart DNA polymerase (Stratagene) from genomic DNA using the PCR primers 5′-GGCTTAAUAACGGGATTTTTGATTGGTT-3′ and 5′-GGTTTAAUTGTCAAGTCAATTAATGTCAATCTATCAG-3′. The PCR products were then cloned into the expression vector pCAMBIA3300 NLS-GUS by using the USER method as described by Nour-Eldin et al. (2006). pCAMBIA3300 NLS-GUS is generated from the original vector pCAMBIA3300 by adding the NLS (nuclear localization signal) and CDS of the GUS with User cloning sites (Nour-Eldin et al., 2006). Using Agrobacterium-mediated transformation, all the constructs were transformed into Arabidopsis using the floral-dip method (Clough and Bent, 1998). T0 transgenic plants were selected on 1/2 Murashige and Skoog (MS) medium containing 50 μg⋅mL-1 kanamycin. For each construct, three independent confirmed T2 transgenic lines were used in the following analysis. GSL Extraction and Analysis 35S::FMOGS-OX6, 35S::FMOGS-OX7, wild-type, FMOGS-OX6 knockout mutant, and FMOGS-OX7 knockout mutant plants were grown simultaneously for 24 days. Individual leaves from each plant were harvested for GSL measurement. 50 to 100 mg leaves were used for the extraction. GSL extraction was performed as previously described (Hansen et al., 2007). High performance liquid chromatography (HPLC) analysis was performed by using the method described by Pang et al. (2009). Spatial Expression Analysis The histochemical detection of GUS expression was performed as previously described (Jefferson et al., 1987). Plant materials were cut and incubated in substrate solution at 37°C for 12 h followed by removal of chlorophyll by submerging the samples in 96% ethanol. The samples were examined and photographed using a stereo microscope (Nikon SMZ1270) or a research microscope (Olympus CX21). For analysis of the responsive expression pattern, seeds of FMOGS-OX6Pro::GUS, FMOGS-OX7Pro::GUS, and wild-type plants were vernalized at 4°C for 72 h and allowed to germinate in 1/2 MS medium for 5 days. The plants were then transferred to 1/2 MS medium containing 50 μM methyl jasmonate (MeJA), 200 μM salicylic acid (SA), and 10 μM abscisic acid (ABA), respectively, and were treated for 48 h. The whole seedling was used for GUS detection using the abovementioned method. Responsive Gene Expression Analysis Wild-type seeds were vernalized at 4°C for 72 h and allowed to germinate in 1/2 MS medium for 5 days. The seedlings were then transferred to 1/2 MS medium containing 100 μM indole acetic acid (IAA), 200 μM SA, 50 μM MeJA, 20 mM 1-aminocyclopropane-1-carboxylic acid (ACC), and 10 μM ABA, respectively. For the cold stress and heat stress treatments, seedlings were transferred to 4 and 30°C growing chamber, respectively. All the treatments were performed for 24 h. Total RNA of the treated and control plants were isolated using TRIzol reagent (Invitrogen) and first-strand cDNA was synthesized using the SupermoIII RT kit (Bioteke). The primer pairs used in qRT-PCR for the detection of genes are listed in Supplementary Table S1. qRT-PCR was performed using SYBR Green Master Mix on an ABI 7500 sequence detection system. Relative transcript levels were normalized using ACTIN II as a control. Results Phylogenetic Analysis of FMOGS-OX Genes in Arabidopsis Flavin-containing monooxygenases protein sequences containing the conserved FMOGS-OX domains were obtained from TAIR1 and were aligned using ClustalX. It was found that At1g12130 and At1g12160 were located in the same subclade, and together with the annotated FMOGS-OX1-5 formed a seven-protein group (Figure 1A), while At1g12200 was located in a subclade adjacent to the seven-protein group (Figure 1A). Because both the overexpressor and knockout mutant of At1g12200 did not show the expected phenotype (Supplementary Tables S2 and S3), the possibility of At1g12200 to be a FMOGS-OX could be excluded. Thus, in the following study, we focused on At1g12130 (named FMOGS-OX6) and At1g12160 (named FMOGS-OX7) (Figure 1B). Overexpression of FMOGS-OX6 and FMOGS-OX7 Altered MT GSLs and MS GSLs Profile To confirm whether FMOGS-OX6 and FMOGS-OX7 possess S-oxygenation activity, transgenic plants 35S::FMOGS-OX6 and 35S::FMOGS-OX7 were generated and three independent lines of each FMOGS-OX6 and FMOGS-OX7 overexpressors were, respectively, used to analysis GSLs profile. Data of the line with the most significant phenotype was shown in Tables 1 and 2. These transgenic plants were confirmed to have much higher expression level than that in wild-type plants by RT-PCR analysis (Supplementary Figure S1). For each plant, the content of GSLs was detected in leaves and seeds of segregating progeny obtained from a heterozygous transgenic parent, thus the possible maternal effects could be minimized. Since the conversion of MT GSLs to MS GSLs rather than the absolute content of MS GSLs could better represent the S-oxygenation activity, MT: (MT+MS) was calculated from the HPLC data. Aliphatic GSLs with side chains of different lengths, including C3 (propyl), C4 (butyl), C5 (pentyl), C6 (hexyl), C7 (heptyl), and C8 (octyl), in the leaf and seed tissues of 35S::FMOGS-OX6, 35S::FMOGS-OX7, and wild-type are shown in Tables 1 and 2. In the seed, MT:(MT+MS) of all the detectable GSLs except propyl, were significantly lower in both 35S::FMOGS-OX6 and 35S::FMOGS-OX7 than in the wild-type. In the leaf, MT:(MT+MS) was lower in 35S::FMOGS-OX7 than in the wild-type, while it was not significantly changed in 35S::FMOGS-OX6. The significantly decreased MT:(MT+MS) in the seed tissue of the two overexpressors suggested that the FMOGS-OX6 and FMOGS-OX7 possess S-oxygenation activity for both short-chain and long-chain GSLs. The weak phenotype of overexpressors in the leaf tissue had been previously observed for other FMOGS-OX enzymes (Hansen et al., 2007; Li et al., 2008); this is possibly because of the low content of the substrate (MT GSLs) and the high content of the product (MS GSL) in the leaf, which showed impaired MT converting reaction. Table 1 Glucosinolates profile in 35S::FMOGS-OX6. MT:(MS+MT) Leaf tissue Seed tissue WT 35S:: FMOGS-OX6 P-value WT 35S:: FMOGS-OX6 P-value Propyl GSL (C3) 0.11 ± 0.009 0.10 ± 0.005 NS 0.02 ± 0.001 0.01 ± 0.000 NS Butyl GSL (C4) 0.28 ± 0.021 0.27 ± 0.022 NS 0.90 ± 0.010 0.57 ± 0.021 <0.05 Pentyl GSL (C5) 0.33 ± 0.013 0.31 ± 0.011 NS 0.75 ± 0.039 0.41 ± 0.024 <0.05 Hexyl GSL (C6) ND ND ND ND Heptyl GSL (C7) 0.30 ± 0.012 0.31 ± 0.012 NS 0.77 ± 0.025 0.40 ± 0.019 <0.05 Octyl GSL (C8) 0.17 ± 0.011 0.17 ± 0.012 NS 0.31 ± 0.013 0.14 ± 0.011 <0.05 Data presented are mean values of MT:(MS+MT) ± Standard Error for at least three replicates per sample. P-value for MT:(MS+MT) differences between the two genotypes were determined by Student’s t-test. ND means given GSL was not detectable; therefore, no statistical analyses were conducted. NS means non-significant P-value (P > 0.05).Table 2 Glucosinolates profile in 35S::FMOGS-OX7. MT:(MS+MT) Leaf tissue Seed tissue WT 35S:: FMOGS-OX7 P-value WT 35S:: FMOGS-OX7 P-value Propyl GSL (C3) 0.09 ± 0.004 0.06 ± 0.003 <0.05 0.03 ± 0.007 0.04 ± 0.010 NS Butyl GSL (C4) 0.39 ± 0.014 0.11 ± 0.003 <0.05 0.83 ± 0.024 0.59 ± 0.011 <0.05 Pentyl GSL (C5) 0.43 ± 0.009 0.30 ± 0.010 <0.05 0.74 ± 0.021 0.38 ± 0.012 <0.05 Hexyl GSL (C6) ND ND ND ND Heptyl GSL (C7) 0.35 ± 0.011 0.20 ± 0.009 <0.05 0.66 ± 0.012 0.31 ± 0.012 <0.05 Octyl GSL (C8) 0.13 ± 0.008 0.06 ± 0.005 <0.05 0.32 ± 0.014 0.13 ± 0.008 <0.05 Data presented are mean values of MT:(MS+MT) ± Standard Error for at least three replicates per sample. P-value for MT:(MS+MT) differences between the two genotypes were determined by Student’s t-test. ND means given GSL was not detectable; therefore, no statistical analyses were conducted. NS means non-significant P-value (P > 0.05).T-DNA knockout mutants FMOGS-OX6 and FMOGS-OX7 were, respectively, analyzed but no significant changes in the GSL profiles were detected (Supplementary Tables S4 and S5). Considering that there are five other identified FMOGS-OX enzymes with the same catalytic activity (Hansen et al., 2007; Li et al., 2008), the absence of observable phenotypic effects in the T-DNA knockout mutants could be attributable to the redundancy in gene function. Expression Pattern of FMOGS-OX6 and FMOGS-OX7 To investigate the spatial expression patterns of FMOGS-OX6 and FMOGS-OX7, we generated the transgenic plants FMOGS-OX6Pro::GUS and FMOGS-OX7Pro::GUS, respectively, expressing the GUS under the control of FMOGS-OX6 and FMOGS-OX7 promoters. Promoter activities were detected by GUS staining, a histochemical technique. As shown in Figure 2, the two genes presented similar expression patterns. In seedlings, the GUS signal was detected in the vascular tissue throughout the radicle, hypocotyls, cotyledon, foliage leaf, and root in both FMOGS-OX6Pro::GUS and FMOGS-OX7Pro::GUS transgenic plants (Figures 2A–D,A’–D’). This expression pattern partially overlaps with that of other FMOGS-OX (Li et al., 2011). Similarly, FMOGS-OX5, FMOGS-OX6, and FMOGS-OX7 were expressed in the vascular tissue of the foliage, including the mid-vein and almost all side-veins, while the other four FMOGS-OX genes were expressed either only in the mid-vein and some major big veins or only in the side-veins (Li et al., 2011). Both the genes were expressed in the vascular tissue throughout the whole root (Figures 2D,D’), which is again similar to the expression pattern of FMOGS-OX5 and different from that of the other FMOGS-OX genes, which are expressed only in the connections of the taproot and lateral roots (Li et al., 2011). FIGURE 2 Expression patterns of FMOGS-OX6 and FMOGS-OX7 detected by promoter-driven GUS expression. (A,B,A’,B’) Cotyledons; (C,C’) foliage leaf; (D,D’) root and hypocotyl; (E,E’) flower; (F,F’) pistil; (G,G’) silique wall; (H,H’) Pseudoseptum and seeds; Bar = 0.3 mm. For both FMOGS-OX6Pro::GUS and FMOGS-OX7Pro::GUS, in flowers, GUS signals were detected in the vascular bundles of calyces, carpels, and stamen filaments, but very little signal was detected in petals (Figures 2E–G,E’–G’). Dense GUS signal was clearly observed at the top and bottom of the ovary and seed funicles (Figures 2F,H,F’,H’). This expression pattern is quite similar to that of other FMOGS-OX genes (Li et al., 2011). In general, the spatial expression of FMOGS-OX6 and FMOGS-OX7 overlapped with each other to a large extent and showed a similar pattern in the vascular tissue to that of other GSL biosynthetic genes (Mikkelsen et al., 2000; Reintanz et al., 2001; Tantikanjana et al., 2001; Grubb et al., 2004; Levy et al., 2005; Skirycz et al., 2006; Li et al., 2011). We further detected the expression of the two genes in response to exogenous ABA, MeJA, and SA. Both FMOGS-OX6 and FMOGS-OX7 were strongly inhibited by MeJA and induced by SA, whereas ABA promoted the expression of FMOGS-OX6 significantly but did not affect FMOGS-OX7 considerably (Figure 3). FIGURE 3 Expression of FMOGS-OX6 and FMOGS-OX7 in response to exogenous abscisic acid (ABA), methyl jasmonate (MeJA), and salicylic acid (SA). Responsive Expression of FMOGS-OX Genes to Different Stimuli The S-oxygenation of MT GSLs to MS GSLs is highly important for plant defense because of the bioactivities of MS GSLs against both biotic and abiotic stresses (Stotz et al., 2011; Martínez-Ballesta et al., 2013, 2015). To date, including FMOGS-OX6 and FMOGS-OX7, seven FMOGS-OX enzymes with redundant functions and overlapping spatial distribution have been identified. Therefore, we investigated whether the response of these seven FMOGS-OX genes to environmental stimuli was co-ordinated or varied. The promoter sequences of the seven FMOGS-OX genes were obtained from NCBI2 and analyzed using the web tool Plantcare 3. The elements in each of the promoter are listed in Supplementary Table S6. Some common elements were found frequently among these promoters and the function of these elements is responsiveness to MeJA, ABA, auxin, ethylene, SA, cold, and heat stresses. To investigate the responsive expression of FMOGS-OX and other aliphatic GSL biosynthetic genes toward these hormone and environmental signals, 5-day old seedlings were treated with ABA, ACC, MeJA, SA, IAA, low temperature (4°C), and high temperature (30°C). Further, the expression level of the seven FMOGS-OX genes, together with that of MYB28 (encodes the main transcription factor that regulates aliphatic GSL biosynthetic genes) and CYP83A1 (an aliphatic GSL biosynthetic gene), were determined by qRT-PCR. Except FMOGS-OX4, which maintained relatively stable expression under different conditions, all the genes responded to the treatments (Figure 4). In general, the FMOGS-OX genes showed responsive expression pattern similar to that of the upstream genes (MYB28 and CYP83A1) in the aliphatic GSL biosynthesis pathway. For example, almost all genes were upregulated by ABA and ACC. However, to the same stimulus, different genes presented different level of sensitivity. FMOGS-OX1, FMOGS-OX2, FMOGS-OX6, and MYB28 showed upregulation (fivefolds), i.e., high sensitivity, toward ABA; FMOGS-OX5, FMOGS-OX6, FMOGS-OX7, and MYB28, toward ACC; FMOGS-OX1 and MYB28, toward MeJA; FMOGS-OX7 and MYB28, toward SA; and FMOGS-OX2, FMOGS-OX5, and MYB28, toward IAA. None of the aliphatic biosynthetic genes were very sensitive to cold and heat stresses. FIGURE 4 Responsive expression patterns of FMOGS-OX genes to different exogenous stimuli. The mRNA abundance of nine genes was analyzed by qRT-PCR. Samples for qRT-PCR were run in three biological replicates with three technical replicates and the data were represented as the Mean ± SD (n = 3). The relative gene expression was calculated using the ΔΔCt algorithm. The expression data were normalized using the invariant expression of ACTIN II. The leaves from mock control seedlings were used as reference sample, which was set to 1. Significant differences according to Student’s t-test are indicated: ∗P < 0.05, ∗∗P < 0.01. Discussion Aliphatic GSLs display high structural diversity due to variation in chain length and secondary modification, and the bioactivity of these compounds is largely dependent on the structure of the side chain. The chemical modification catalyzed by the different FMOGS-OX enzymes is highly important because MT GSLs and MS GSLs account for a large proportion of total GSLs, and production of MS GSLs is the basis of further modification of aliphatic GSLs. The five annotated FMOGS-OX enzymes were identified by Hansen (Hansen et al., 2007) and Li (Li et al., 2008). In the study conducted by Hansen, a phylogenetic tree of plant FMOs from rice, Arabidopsis, and poplar was analyzed. Within the clade considered to be involved in S-oxygenation, the presence of a subclade of seven proteins, including the annotated FMOGS-OX1-5, FMOGS-OX6, and FMOGS-OX7, only in the GSL-producing Arabidopsis plant suggests that these genes may be responsible for S-oxygenation of GSLs. Our result verified Hansen’s prediction. Compared with the known FMOGS-OX enzymes, FMOGS-OX6 and FMOGS-OX7 was quite similar to FMOGS-OX1-4, which contribute to the conversion of both short and long-chain MT GSLs, and was different from FMOGS-OX5, which specifically recognizes long-chain substrates. FMOGS-OX6 and FMOGS-OX7 showed very similar spatial expression patterns and it overlapped with that of most other GSL biosynthetic genes, which are frequently found in the vascular bundle. However, some small variations can be observed between these FMOGS-OX genes. Li et al. (2011) reported that in the leaf, FMOGS-OX1-4 were distributed partially in the veins and FMOGS-OX5 was expressed in the mid-vein and all the branched veins, whereas in the root, FMOGS-OX1-4 was limited to the connections between the taproot and lateral roots and FMOGS-OX5 was present throughout the vascular tissue in the whole root. Interestingly, FMOGS-OX6 and FMOGS-OX7, like FMOGS-OX1-4, catalyze S-oxygenation of GSLs without substrate specificity and exhibit less specific spatial expression like FMOGS-OX5. Based on the phylogenetic analysis of the FMOGS-OX genes in Arabidopsis (Figure 1A), our results suggest an FMOGS-OX ancestor with a broad range of non-specific substrates for catalysis and broad spatial expression. The gene encoding this ancestor could probably have been duplicated into two groups: one (comprising FMOGS-OX1-5) might have become more specific either in spatial expression (FMOGS-OX1-4) or in substrate recognition (FMOGS-OX5) and the other group (comprising FMOGS-OX6 and FMOGS-OX7) remained non-specific in both aspects. Some common elements were detected in the promoter sequences of these FMOGS-OX genes. Consistently, the FMOGS-OX genes presented similar responsive expression tendency under several hormone treatments and temperature stresses. However, the sensitivity of each gene differed in response to the same treatment. Integrating the knowledge of the various FMOGS-OX genes, we observed that they are quite redundant in their catalytic activity, distribution pattern, and response to exogenous stimuli. It is considered that duplicate genes play a more important role for the compensation of secondary metabolites than that of primary metabolites because there are fewer alternative pathways in secondary metabolite synthesis (Hanada et al., 2011). This redundancy of FMOGS-OX genes might play a significant role in functional compensation and increasing genetic robustness. Additionally, the variations found among these FMOGS-OX genes could possibly result in delicate functional difference and contribute to the finely tuned modification of aliphatic GSLs. Aliphatic GSLs are primarily considered to contribute to plant resistance toward pests (Beekwilder et al., 2008). Recent studies discovered multiple functions of these compounds; they are involved in toxicity against pathogens and have been suggested to contribute to maintaining water balance under salt stress (Tierens et al., 2001; Stotz et al., 2011; Martínez-Ballesta et al., 2015). The versatile bioactivity of aliphatic GSLs is largely dependent on the length and modification of their side chain (Stotz et al., 2011). However, the contribution of specific metabolites to the biotic or abiotic stress defenses is still poorly understood. Identification of new FMOGS-OX enzymes provides a way to investigate the specific bioactivity of aliphatic GSLs with different side chain structures. As an application for humans, these FMOGS-OX genes can potentially be used in breeding Brassica vegetables with improved anti-cancer properties conferred by the MS GSLs. Author Contributions JL and WK carried out experiments and analyzed experiment results. QY, WC, and RX carried out partial of the experiments. JL and WK wrote the manuscript. WJ designed the experiments. 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 National Natural Science Foundation of China (NSFC) 31370334, NSFC 31170368, and Science and Technology Research Project of Education Department of Heilongjiang Province, 2531003. 1 https://www.arabidopsis.org/ 2 http://www.ncbi.nlm.nih.gov/ 3 http://bioinformatics.psb.ugent.be/webtools/plantcare/html/ Supplementary Material The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2016.01292 Click here for additional data file. Click here for additional data file. Click here for additional data file. Click here for additional data file. Click here for additional data file. Click here for additional data file. FIGURE S1 Expression levels of FMOGS-OX6 and FMOGS-OX7 in the overexpression mutants. Click here for additional data file. Click here for additional data file. ==== Refs References Atwell L. L. Beaver L. M. Shannon J. Williams D. E. Dashwood R. H. Ho E. (2015 ). Epigenetic regulation by sulforaphane: opportunities for breast and prostate cancer chemoprevention. Curr. Pharmacol. Rep. 2 102 –111 . 10.1007/s40495-014-0002-x 26042194 Bednarek P. Pislewska-Bednarek M. Svatos A. Schneider B. Doubsky J. Mansurova M. (2009 ). A glucosinolate metabolism pathway in living plant cells mediates broad-spectrum antifungal defense. Science 323 101 –106 . 10.1126/science.1163732 19095900 Beekwilder J. Leeuwen W. V. Dam N. M. V. Bertossi M. Grandi V. Mizzi L. (2008 ). The impact of the absence of aliphatic glucosinolates on insect herbivory in Arabidopsis. PLoS ONE 3 :e2068 10.1371/journal.pone.0002068 Bones A. M. Rossiter J. T. (2006 ). The enzymic and chemically induced decomposition of glucosinolates. Phytochemistry 67 1053 –1067 . 10.1016/j.phytochem.2006.02.024 16624350 Chang Y. W. Jang J. Y. Kim Y. H. Kim J. W. Shim J. J. (2015 ). The effects of broccoli sprout extract containing sulforaphane on lipid peroxidation and Helicobacter pyloriinfection in the gastric mucosa. Gut Liver. 9 486 –493 . 10.5009/gnl14040 25287166 Clay N. K. Adio A. M. Denoux C. Jander G. Ausubel F. M. (2009 ). Glucosinolate metabolites required for an Arabidopsis innate immune response. Science 323 95 –101 . 10.1126/science.1164627 19095898 Clough S. J. Bent A. F. (1998 ). Floral dip: a simplified method for Agrobacterium-mediated transformation of Arabidopsis thaliana. Plant J. 16 735 –743 . 10.1046/j.1365-313x.1998.00343.x 10069079 Fahey J. W. Haristoy X. Dolan P. M. Kensler T. W. Scholtus I. Stephenson K. K. (2002 ). Sulforaphane inhibits extracellular, intracellular, and antibiotic-resistant strains of Helicobacter pylori and prevents benzo[α] pyreneinduced stomach tumors. Proc. Natl. Acad. Sci. U.S.A. 99 7610 –7615 . 10.1073/pnas.112203099 12032331 Fahey J. W. Zhang Y. Talalay P. (1997 ). Broccoli sprouts: an exceptionally rich source of inducers of enzymes that protect against?chemical?carcinogens. Proc. Natl. Acad. Sci. U.S.A. 94 10367 –10372 . 10.1073/pnas.94.19.10367 9294217 Fofaria N. M. Ranjan A. Kim S. H. Srivastava S. K. (2015 ). Mechanisms of the anticancer effects of isothiocyanates. Enzymes 37 111 –137 . 10.1016/bs.enz.2015.06.001 26298458 Grubb C. D. Zipp B. J. Ludwig-Muller J. Masuno M. N. Molinski T. F. Abel S. (2004 ). Arabidopsis glucosyltransferase UGT74B1 functions in glucosinolate biosynthesis and auxin homeostasis. Plant J. 40 893 –908 . 10.1111/j.1365-313X.2004.02261.x 15584955 Halkier B. A. Gershenzon J. (2006 ). Biology and biochemistry of glucosinolates. Annu. Rev. Plant Biol. 57 303 –333 . 10.1146/annurev.arplant.57.032905.105228 16669764 Hanada K. Sawada Y. Kuromori T. Klausnitzer R. Saito K. Toyoda T. (2011 ). Functional compensation of primary and secondary metabolites by duplicate genes in Arabidopsis thaliana. Mol. Biol. Evol. 1 377 –382 . 10.1093/molbev/msq204 20736450 Hansen B. G. Kliebenstein D. J. Halkier B. A. (2007 ). Identification of a flavinmonooxygenaseas the S-oxygenating enzyme in aliphatic glucosinolate biosynthesis in Arabidopsis. Plant J. 50 902 –910 . 10.1111/j.1365-313X.2007.03101.x 17461789 Hopkins R. J. Dam N. M. V. Loon J. J. A. V. (2009 ). Role of glucosinolates in insect-plant relationships and multitrophic interaction. Annu. Rev. Entomol. 54 57 –83 . 10.1146/annurev.ento.54.110807.090623 18811249 Jefferson R. Kavanagh T. A. Bevan M. W. (1987 ). GUS fusions: ß-glucuronidase as a sensitive and versatile gene fusion marker inhigher plants. EMBO J. 16 3901 –3907 .3327686 Laluk K. Prasad K. V. Savchenko T. Celesnik H. Dehesh K. Levy M. (2012 ). The calmodulin-binding transcription factor SIGNAL RESPONSIVE1 is a novel regulator of glucosinolate metabolism and herbivory tolerance in Arabidopsis. Plant Cell Physiol. 53 2008 –2015 . 10.1093/pcp/pcs143 23072934 Levy M. Wang Q. Kaspi R. Parrella M. P. Abel S. (2005 ). Arabidopsis IQD1 a novel calmodulin-binding nuclear protein, stimulates glucosinolate accumulation and plant defense. Plant J. 43 79 –96 . 10.1111/j.1365-313X.2005.02435.x 15960618 Li J. Hansen B. G. Ober J. A. Kliebenstein D. J. Halkier B. A. (2008 ). Subclade of flavinmonooxygenasesinvolved in aliphatic glucosinolate biosynthesis. Plant Physiol. 148 1721 –1733 . 10.1104/pp.108.125757 18799661 Li J. Kristiansen K. A. Hansen B. G. Halkier B. A. (2011 ). Cellular andsubcellular localization of flavin-monooxygenases involved in glucosinolatebiosynthesis. J. Exp. Bot. 62 1337 –1346 . 10.1093/jxb/erq369 21078824 Marchler-Bauer A. Derbyshire M. K. Gonzales N. R. Lu S. Chitsaz F. Geer L. Y. (2015 ). CDD: NCBI’s conserved domain database. Nucleic Acids Res. 43 222 –226 . 10.1093/nar/gku1221 Martínez-Ballesta M. C. Moreno-Fernández D. A. Carvajal M. (2013 ). The Physiological importance of glucosinolates on plant response to abiotic stress in Brassica. Int. J. Mol. Sci. 14 11607 –11625 . 10.3390/ijms140611607 23722664 Martínez-Ballesta M. C. Moreno-Fernández D. A. Castejón D. Ochando C. Morandini P. A. Carvajal M. (2015 ). The impact of the absence of aliphatic glucosinolates on water transport under salt stress in Arabidopsis thaliana. Front. Plant Sci. 6 :524 10.3389/fpls.2015.00524 Mikkelsen M. D. Hansen C. H. Wittstock U. Halkier B. A. (2000 ). Cytochrome P450 CYP79B2 from Arabidopsis catalyzes the conversion of tryptophan to indole-3-acetaldoxime, a precursor of indole glucosinolates and indole-3-acetic acid. J. Biol. Chem. 275 33712 –33717 . 10.1074/jbc.M001667200 10922360 Nour-Eldin H. H. Hansen B. G. Nørholm M. H. Jensen J. K. Halkier B. A. (2006 ). Advancing uracil-excision based cloning towards an ideal techniquefor cloning PCR fragments. Nucleic Acids Res. 34 :e122 10.1093/nar/gkl635 Pang Q. Chen S. Li L. Yan X. (2009 ). Characterizationof glucosinolate-myrosinase system in developing salt cress Thellungiella halophila. Physiol. Plant 136 1 –9 . 10.1111/j.1399-3054.2009.01211.x 19508363 Reintanz B. Lehnen M. Reichelt M. Gershenzon J. Kowalczyk M. Sandberg G. (2001 ). Bus, a bushy Arabidopsis CYP79F1 knockout mutant with abolished synthesis of short-chain aliphatic glucosinolates. Plant Cell 13 351 –367 . 10.2307/3871281 11226190 Rose P. Faulkner K. Williamson G. Mithen R. (2000 ). 7-Methylsulfinylheptyl and 8-methylsulfinyloctyl isothiocyanates from watercress are potent inducers of phase II enzymes. Carcinogenesis 21 1983 –1988 . 10.1093/carcin/21.11.1983 11062158 Rose P. Huang Q. Ong C. N. Whiteman M. (2005 ). Broccoli and watercress suppress matrix metalloproteinase-9 activity and invasiveness of human MDA-MB-231 breast cancer cells. Toxicol. Appl. Pharmacol. 209 105 –113 . 10.1016/j.taap.2005.04.010 15953625 Skirycz A. Reichelt M. Burow M. Birkemeyer C. Rolcik J. Kopka J. (2006 ). DOF transcription factor AtDof1.1 (OBP2) is part of a regulatory network controlling glucosinolate biosynthesis in Arabidopsis. Plant J. 47 10 –24 . 10.1111/j.1365-313X.2006.02767.x 16740150 Stotz H. U. Sawada Y. Shimada Y. Hirai M. Y. Sasaki E. Krischke M. (2011 ). Role of camalexin, indole glucosinolates, and side chain modification of glucosinolate-derived isothiocyanates in defense of Arabidopsis against Sclerotinia sclerotiorum. Plant J. 67 81 –93 . 10.1111/j.1365-313X.2011.04578.x 21418358 Talalay P. Fahey J. W. Healy Z. R. Wehage S. L. Benedict A. L. Min C. (2007 ). Sulforaphane mobilizes cellular defenses that protect skin against damage by UV radiation. Proc. Natl. Acad. Sci. U.S.A. 104 17500 –17505 . 10.1073/pnas.0708710104 17956979 Tantikanjana T. Yong J. W. Letham D. S. Griffith M. Hussain M. Ljung K. (2001 ). Control of axillary bud initiation and shoot architecture in Arabidopsis through the SUPERSHOOT gene. Gene Dev. 15 1577 –1588 . 10.1101/gad.887301 11410537 Tierens K. J. Thomma B. Brower M. Schmidt J. Kistner K. Porzel A. (2001 ). Study of the role of antimicrobial glucosinolate-derived isothiocyanates in resistance of Arabidopsis to microbial pathogens. Plant Physiol. 125 1688 –1699 . 10.1104/pp.125.4.1688 11299350 Zalcmann A. T. Talalay P. (2001 ). The chemical diversity and distribution of glucosinolates and isothiocyanates among plants. Phytochemistry 56 5 –51 .11198818 Zhang Z. Ober J. A. Kliebenstein D. J. (2006 ). The gene controlling thequantitative trait locus EPITHIOSPECIFIER MODIFIER1 alters glucosinolatehydrolysis and insect resistance in Arabidopsis. Plant Cell 18 1524 –1536 . 10.1105/tpc.105.039602 16679459

PMC005xxxxxx/PMC5003063.txt

==== Front Front Public HealthFront Public HealthFront. Public HealthFrontiers in Public Health2296-2565Frontiers Media S.A. 10.3389/fpubh.2016.00186Public HealthGeneral CommentaryCommentary: Implementing Pro-Poor Universal Health Coverage Jakovljevic Mihajlo 12*1The Faculty of Medical Sciences, University of Kragujevac, Kragujevac, Serbia2Hosei University Tokyo, Tokyo, JapanEdited by: Sandra C. Buttigieg, University of Malta, Malta Reviewed by: Krzysztof Kaczmarek, Medical University of Silesia, Poland; Kyriakos Souliotis, University of Peloponnese, Greece *Correspondence: Mihajlo Jakovljevic, sidartagothama@gmail.com, jakovljevicm@medf.kg.ac.rsSpecialty section: This article was submitted to Health Economics, a section of the journal Frontiers in Public Health 29 8 2016 2016 4 18616 7 2016 17 8 2016 Copyright © 2016 Jakovljevic.2016JakovljevicThis is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.A commentary on Implementing Pro-Poor Universal Health Coverage by Bump J, Cashin C, Chalkidou K, Evans D, González-Pier E, Guo Y, et al. Lancet Glob Health (2015) 4:e14–6. doi: 10.1016/S2214-109X(15)00274-0universal health coveragelow middle income countriesrisk sharingfinancialBRICSglobal health ==== Body Recently published extraordinary article entitled: “implementing pro-poor universal health coverage” depicts an issue of truly global outreach (1). Modern day health system establishments had their historical roots back in the early industrial era of late nineteenth century Europe (2). Risk sharing through introducing the first health insurance funds was initially targeted to protect industrial laborers as an important segment of the society of the time (3). Health coverage of citizens beneath poverty line therefore began to slowly expand to the other vulnerable groups. During the first half of twentieth century, such practice spreads to North America (4) and Japan (5). It is less known that the first nationwide success in achieving universal health coverage (UHC) is attributable to the early Soviet Union back in 1930s and its famous Semashko system (6). Disintegration of colonial system worldwide after the end of WWII and rise of the non-aligned movement gave significant impetus to the health system developments among the Third World nations (7). After the end of Cold War Era, accelerated pace of globalization saw the uneven growth of welfare in these countries (8). Although attractive as a policy goal, health coverage for massive rural populations remained a distant dream for many world regions (9). The aforementioned paper by Bump et al. pointed out to the core global UHC developments in a comprehensive manner. Their call to national governments to commit to the established milestones of UHC evolution is clear and might indeed serve the purpose. Nevertheless, few crucial facts were omitted, which might significantly narrow the horizon of perception on global evolution of UHC with regard to the role of BRICS nations (10). Due to overall increase in welfare, UHC for the poor rapidly expanded around the world (11). Therefore, it seems that we might be deceived by perception that all of these world regions contributed evenly or at least to the comparable extent (12). The reality is rather different: there is a very narrow circle of top emerging economies to which we own most of this progress. Lion share of the growth in UHC, the world owns to the BRICS nations (13). Accounting for roughly two-fifths of world’s population, over the past two decades these national governments lifted from poverty hundreds of millions of the world’s poorest citizens (14). Quite efficient government policies dedicated to reducing poverty took place in these economies since late 1990s with few notable examples led by Chinese overachievement (15–17). Such an increase in social welfare of poorest citizens was attributable to industrial enterprise and direct foreign investment (18, 19). Their health reforms were bold and successful to the large extent leading to the notable gains toward achieving UHC (20). Distinctive role of these economies in global health arena led WHO Bulletin to establish a specialty issue committed to BRICS back in 2014 (21). Some of the exposed weaknesses alongside this ambitious process were India’s inability to expand health expenditure in terms of GDP percentage (22). Socioeconomic inequalities in health care expanded in some members of the group driven by exploding prevalence of prosperity diseases (23). Despite the fact of these obstacles accelerated expansion of UHC remains clearly visible in Russia, Brazil, India, and China (24). One of the surprising developments is the strong and continuing upward trend of their national abilities to increase investment in health care and expand insurance coverage of the population below poverty line (25). Long-term commitment of BRICS governments ultimately resulted in significantly improved health outcomes, including nationwide longevity (26). A global landscape of UHC evolution implies that orchestrated international efforts should regard these nations as one of the pillars of any responsible policy aimed to protect the world’s poor from health-related risks. Author Contributions MJ has designed drafted and finalized the manuscript. Conflict of Interest Statement The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ==== Refs References 1 Bump J Cashin C Chalkidou K Evans D González-Pier E Guo Y Implementing pro-poor universal health coverage . Lancet Glob Health (2015 ) 4 :e14 –6 .10.1016/S2214-109X(15)00274-0 26700794 2 Rosen G From Medical Police to Social Medicine: Essays on the History of Health Care . New York : Science History Publications, Neale Watson Academic Publications, Inc. (1974 ). 3 Sigerist HE From Bismarck to Beveridge: developments and trends in social security legislation . Bull Hist Med (1943 ) 13 :365 –88 . 4 Getzen TE . Health care is an individual necessity and a national luxury: applying multilevel decision models to the analysis of health care expenditures . J Health Econ (2000 ) 19 (2 ):259 –70 .10.1016/S0167-6296(99)00032-6 10947579 5 Ogura S Jakovljevic M Health financing constrained by population aging – an opportunity to learn from Japanese experience . Ser J Exp Clin Res (2014 ) 15 (4 ):175 –81 .10.2478/SJECR 6 Semashko NA Health Protection in the USSR . London : Victor Gollancz (1934 ). p. 39 –40 . 7 Morphet S Multilateralism and the non-aligned movement: what is the global South doing and where is it going? Glob Govern (2004 ) 10 (4 ):517 –37 . 8 Jakovljevic MB . Resource allocation strategies in Southeastern European health policy . Eur J Health Econ (2013 ) 14 (2 ):153 –9 .10.1007/s10198-012-0439-y 23143312 9 Jakovljevic M Groot W Souliotis K Health care financing and affordability in the emerging global markets . Front Public Health (2016 ) 4 :2 10.3389/fpubh.2016.00002 26835444 10 O’Neill J Building Better Global Economic BRICs. Global Economics Paper No: 66 ©2001 . Goldman, Sachs & Co (2001 ). 11 Jakovljevic M Getzen T Growth of global health spending share in low and middle income countries . Front Pharmacol (2016 ) 2016 (7 ):21 10.3389/fphar.2016.00021 26903867 12 McIntyre D Thiede M Dahlgren G Whitehead M . What are the economic consequences for households of illness and of paying for health care in low-and middle-income country contexts? Soc Sci Med (2006 ) 62 (4 ):858 –65 .10.1016/j.socscimed.2005.07.001 16099574 13 Marten R McIntyre D Travassos C Shishkin S Longde W Reddy S An assessment of progress towards universal health coverage in Brazil, Russia, India, China, and South Africa (BRICS) . Lancet (2014 ) 384 (9960 ):2164 –71 .10.1016/S0140-6736(14)60075-1 24793339 14 Jakovljevic M The key role of leading emerging BRIC markets for the future of global health care, SJECR, UDK: 614.2(100) . Ser J Exp Clin Res (2014 ) 15 (3 ):139 –43 .10.5937/sjecr1403139J 15 Hemson D Owusu-Ampomah K A Better Life for All? Service Delivery and Poverty Alleviation . South Africa : State of the Nation (2004 ). p. 511 –36 . 16 Lemus E Abreu A The persistent poverty level in the United States vs. the BRICs emerging economies market . J Asia Pac Stud (2015 ) 4 (2 ):169 –76 . 17 Kickbusch I BRICS? Contributions to the global health agenda . Bull World Health Organ (2014 ) 92 (6 ):463 –4 .10.2471/BLT.13.127944 24940025 18 Yu M Industrial Structural Upgrading and Poverty Reduction in China, the Untold Story: Structural Change for Poverty Reduction: The Case of the BRICS . Beijing : Peking University (2012 ). 19 United Nations Industrial Development Organization . Structural Change, Poverty Reduction and Industrial Policy in the BRICS . Vienna : United Nations University (2012 ). Available from: http://www19.iadb.org/intal/intalcdi/PE/2013/10846.pdf 20 Rao KD Petrosyan V Araujo EC McIntyre D . Progress towards universal health coverage in BRICS: translating economic growth into better health . Bull World Health Organ (2014 ) 92 (6 ):429 –35 .10.2471/BLT.13.127951 24940017 21 McKee M Marten R Balabanova D Watt N Huang Y Finch AP BRICS? Role in global health and the promotion of universal health coverage: the debate continues . Bull World Health Organ (2014 ) 92 (6 ):452 –3 .10.2471/BLT.13.132563 24940020 22 Barik D Thorat A Issues of unequal access to public health in India . Front Public Health (2015 ) 2015 (3 ):245 10.3389/fpubh.2015.00245 26579507 23 Jakovljevic M Milovanovic O Growing burden of non-communicable diseases in the emerging health markets: the case of BRICS; Research topic: health care financing and affordability in the emerging global markets . Front Public Health (2015 ) 3 :65 10.3389/fpubh.2015.00065 25954740 24 Jakovljevic MB BRIC’s growing share of global health spending and their diverging pathways . Front Public Health (2015 ) 3 :135 10.3389/fpubh.2015.00135 26000273 25 Jakovljevic MM Comparison of historical medical spending patterns among the BRICS and G7 . J Med Econ (2016 ) 19 (1 ):70 –6 .10.3111/13696998.2015.1093493 26366470 26 Jakovljevic M Vukovic M Fontanesi J Life expectancy and health expenditure evolution in Eastern Europe – DiD and DEA analysis . Expert Rev Pharmacoecon Outcomes Res (2015 ) 17 :1 –10 .10.1586/14737167.2016.1125293 26606654

PMC005xxxxxx/PMC5003080.txt

==== Front Front PsycholFront PsycholFront. Psychol.Frontiers in Psychology1664-1078Frontiers Media S.A. 10.3389/fpsyg.2016.01166PsychologyOriginal ResearchOn the Intentionality of Cultural Products: Representations of Black History As Psychological Affordances Salter Phia S. 1*Adams Glenn 21Psychology and Africana Studies, Texas A&M UniversityCollege Station, TX, USA2Psychology, University of KansasLawrence, KS, USAEdited by: Keiko Ishii, Kobe University, Japan Reviewed by: Chris Sinha, Hunan University, China; Sawa Senzaki, University of Wisconsin–Green Bay, USA *Correspondence: Phia S. Salter psalter@tamu.eduThis article was submitted to Cultural Psychology, a section of the journal Frontiers in Psychology 29 8 2016 2016 7 116627 3 2016 20 7 2016 Copyright © 2016 Salter and Adams.2016Salter and AdamsThis is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.A cultural-psychological analysis emphasizes the intentionality of everyday worlds: the idea that material products not only bear psychological traces of culturally constituted beliefs and desires, but also subsequently afford and promote culturally consistent understandings and actions. We applied this conceptual framework of mutual constitution in a research project using quantitative and qualitative approaches to understand the dynamic resonance between sociocultural variance in Black History Month (BHM) representations and the reproduction of racial inequality in the U.S. In studies 1 and 2, we considered whether mainstream BHM artifacts reflect the preferences and understandings of White Americans (i.e., psychological constitution of cultural worlds). Consistent with the psychological constitution hypothesis, White American participants reported more positive affect, better recognition, and greater liking for BHM representations from the schools where White Americans were the majority than BHM representations from the schools where Black students and other students of color were the majority. Moreover, as an indication of the identity relevance of BHM representations, White identification was more positively associated with judgments of positive affect and preference in response to BHM representations from White schools than BHM representations from the schools where Black students were in the majority. In studies 3 and 4, we considered whether BHM representations from different settings differentially afford support or opposition to anti-racism policies (i.e., cultural constitution of psychological experience). In support of the cultural constitution hypothesis, BHM representations typical of schools where Black students were in the majority were more effective at promoting support for anti-racism policies compared to BHM representations typical of predominately White schools and a control condition. This effect was mediated by the effect of (different) BHM representations on perception of racism. Together, these studies suggest that representations of Black History constitute cultural affordances that, depending on their source, can promote (or impede) perception of racism and anti-racism efforts. This research contributes to an emerging body of work examining the bidirectional, psychological importance of cultural products. We discuss implications for theorizing collective manifestations of mind. Black History Monthcollective memoryintentional worldscultureracial inequality ==== Body Introduction In 1926, Carter G. Woodson proposed a tradition of “Negro History Week” to commemorate and legitimize Black Americans' contributions to mainstream American society. The explicit motivation for this tradition was to shape not only pride within Black American communities, but also anti-racist consciousness throughout American society (Dagbovie, 2004). The tradition has subsequently become an increasingly mainstream event, especially after the United States Congress recognized February as Black History Month (BHM) in 1986. As mainstream American society has increasingly appropriated BHM, this cultural tool has evolved to serve a variety of different purposes (Pitre and Ray, 2002). Indeed, scholars debate whether current BHM commemorations serve Woodson's liberatory goals or are instead primarily a means for corporations to market goods to the Black community (e.g., posters from Budweiser with the slogan “this chapter of history brought to you by the king of beer”; Persinger, 2011; see also Franklin, 1997; Dagbovie, 2005). These mainstream appropriations make clear that commemoration of BHM is not simply about disinterested documentation of Black American histories. Instead, the variety of forms and purposes suggest that representations of BHM may function as cultural affordances (Kitayama and Markus, 1999; Kitayama et al., 2006; see Gibson, 1977): that is, cultural tools that make possible particular beliefs, motivations, and actions. The present research applies a cultural psychology analysis to consider how displays for BHM reflect and promote identity-relevant action. Although the research involves a comparison of cultural products from Black American and White American spaces, what makes this a cultural psychology project is a focus on the mediation of human experience via cultural tools (Vygotsky, 1978; Wertsch, 2002; Rogoff, 2003). Cultural psychology: the study of intentional worlds In his field-defining work, Shweder (1990) famously defined cultural psychology as the study of mutual constitution: how “psyche and culture …make each other up” (p. 1). Subsequent work in cultural psychology has tended to emphasize one direction of this mutual constitution dynamic: the cultural constitution of psyche. This direction emphasizes that species-typical patterns of mind are not just natural, but instead emerge as people engage with structures of mind-in-context that provide the necessary ecological scaffolding for human experience (Markus and Kitayama, 1991; Kim and Markus, 1999; Heine and Lehman, 2004; Adams et al., 2010). Subsequent work has tended to neglect the other direction of the mutual constitution dynamic: the psychological constitution of cultural worlds. This direction emphasizes that the everyday ecology of the human organism is likewise not just natural, but instead emerges through everyday actions as people realize (i.e., make real; Berger and Luckmann, 1966/1992; Moscovici, 1984) their understandings and desires (e.g., Kim and Markus, 1999; Markus et al., 2006; Adams et al., 2010). The relative neglect of the psychological constitution process is striking given that, a few paragraphs later, Shweder (1990) defined cultural psychology as “the study of intentional worlds” (p. 3), that is, “human artifactual worlds populated with products of our own design” (p. 2). He proposed that “a sociocultural environment is an intentional world …as long as there exists a community of persons whose beliefs, desires, emotions, purposes, and other mental representations are directed at it and are thereby influenced by it (Shweder, 1990; p. 2; emphasis added). To refer to cultural worlds as directed (or constituted) is to say that everyday realities are not neutral; instead, people infuse them with the charge of their beliefs and desires. To refer to cultural worlds as directive (or constituting) is to say that everyday realities are not psychologically inert; instead, they afford some forms of experience, constrain other forms, and direct action toward particular ends. History narratives as directed cultural products In this research, we apply ideas about the intentionality of everyday worlds to a particular class of cultural-psychological products: historical narratives. Historical narratives are not often firsthand experiences of seminal past events (Adamczyk, 2002); instead, they are dynamically constructed, collective memories. When sources present historical narratives as “just the facts” or “objective” accounts of the past, they obscure how reconstructions of the past must necessarily omit some details (Loewen, 1995; Trouillot, 1995). Among the most important influences on production of historical memory are social identity processes. Social identification affects (re)production of historical narratives by influencing people to deny, minimize, silence, or otherwise de-select information about collective misdeeds or other unflattering events. People experience motivations to reconstruct the past in an identity-favorable light to avoid negative feelings and emotions associated with threats to collective identity (e.g., Branscombe and Miron, 2004; Wohl et al., 2006; Morton and Sonnenberg, 2011). These motivations can lead people who are high in collective identification (either as a stable individual difference or temporary situational affordance) to minimize the negative consequences of collective misdeeds (Doosje et al., 1998) or to use strategies of psychological distancing to minimize their relevance for present events (Pennebaker and Banasik, 1997; Kurtiş et al., 2010). Besides these relatively “hot” cognitions or motivated forms of minimization and denial, dominant constructions of social identity often have celebratory connotations that lead people to recall more positive (or fewer negative) features of the collective past, regardless of their individual motivations (Sahdra and Ross, 2007). The influence of social identity on production of historical memory has implications for the psychological constitution of cultural reality. When social identity processes lead people to (re)produce some historical narratives rather than others, the result is not merely the endpoint of an action sequence. Instead, identity-relevant action exerts selective pressure on the reproduction of cultural reality. The resulting narratives carry the psychological charge of particular beliefs and desires (and not others), and this charge circulates with these narratives beyond the immediate time and place of their production. The important implication is that the presence and absence of particular history narratives in everyday cultural ecologies are not the product of happenstance (Baumeister and Hastings, 1997; Pennebaker and Gonzales, 2009); instead, prominent collective narratives bear the influence of identity-relevant concerns regarding what is true and desirable. History narratives as a directive cultural force Representations of history provide the scaffolding for conceptions of nationhood and other collective identities (Reicher and Hopkins, 2001; Eyerman, 2004; Liu and Hilton, 2005; Hammack, 2010). They constitute tools that provide distinctiveness in social comparisons with other nations; confer legitimacy to norms, moral codes, and laws; and facilitate the assimilation (or rejection) of marginalized groups via official policies of monoculturalism or multiculturalism (Liu and Hilton, 2005). History representations provide access to a collective story around which individuals can locate their personal and collective selves. For example, the trauma of slavery is a collective memory that has served as a basis for African American identity (Eyerman, 2004). Engagement with representations associated with collective memory of American slavery—not the actual experience of slavery—influences how Americans of African descent see themselves in relation to other groups (e.g., collectively identifying as African American). Beyond collective identity construction, history narratives also impact identity-relevant perception and action. Of particular relevance to the present work is research that shows how differences in knowledge about past racism can mediate divergent perceptions of racism in contemporary societies (Nelson et al., 2013). That is, persistent identity-relevant differences in perception of racism can reflect identity-relevant differences in historical knowledge. Other research demonstrates that the salience of historical narratives can impact support for reparative action. For example, low-identified Dutch participants advocated more aid to an Indonesian cause when aspects of Dutch history (including colonial occupation of Indonesia) were salient than when these aspects of history were not salient (Doosje et al., 1998). In summary, historical knowledge can be a directive force that impacts construction of current events and subsequent responses (e.g., reparative action, apologies, or compensation). Power and racial privilege in cultural production An important implication of an intentional worlds analysis is that identity-relevant products prominent in a particular cultural setting are likely to reflect and promote identity concerns of dominant groups. Applied to the present study, historical narratives in mainstream American settings are likely to reflect and promote beliefs and desires of the dominant, White American majority. Moreover, the relationship between identity concerns and historical representation may be amplified in cases, like institutionally sanctioned commemoration of BHM, where the events under consideration—enslavement, systematic rape and torture, terroristic violence, apartheid/segregation, and other crimes against humanity—disrupt celebratory or glorifying narratives of nation and are therefore potentially threatening to dominant-group constructions of identity. Given social pressures to commemorate events associated with BHM, the impact of dominant-group identity concerns on cultural reproduction may lead to the evolution of relatively “sanitized” representations of history that silence or minimize discussions of oppression (Brown and Brown, 2010; Kurtiş et al., 2010). One strategy for production of sanitized representations of history is to frame discussions about racism in terms of multicultural tolerance, diversity, or other ways consistent with the “prejudice problematic” (e.g., Wetherell, 2011) rather than focus on the ongoing legacy of systemic expropriation, exploitation, and violent oppression (Adams et al., 2008; Wright and Lubensky, 2009; Hammack, 2011). Another strategy for production of sanitized representations of history is to highlight individual achievement (despite un-named barriers) rather than collective struggle (to eliminate barriers; Banks, 1999; Pitre and Ray, 2002). Besides deflection of attention away from structural barriers, the emphasis on heroic individual achievement has the added advantage of resonating with dominant group ideologies—specifically, colorblindness, meritocracy, protestant work ethic—that reinforce the atomistic construction of racism in terms of the prejudice problematic (e.g., Levin et al., 1998; Jost et al., 2004). To the extent that mainstream historical narratives promote denial of racism in contemporary society by omitting narratives of oppression and other objectionable events (e.g., genocide, slavery; Loewen, 1995; Schick and St. Denis, 2005), they constitute intentional worlds of racism that further perpetuate racial domination. Overview of the present research We conducted four studies that explore how representations of BHM function as cultural-psychological tools for identity-relevant perception and action. The design of the research relies heavily on the idea of cultural realities as psychologically constituted, intentional worlds that both reflect and promote identity-relevant, community-specific beliefs and desires. This design emphasizes two ideas regarding the “intentional” character of Black History representations. The first idea is that representations of Black History are not “natural” or “objective” accounts of the past. Instead, these cultural products bear the psychological imprint of community-specific beliefs and desires. We refer to this idea as the psychological constitution hypothesis. The second idea is that representations of Black History are not neutral or inert. Instead, these representations carry a charge that affords identity-relevant action in the service of community-specific goals. We refer to this idea as the cultural constitution hypothesis. The hypotheses are complementary and reflect the bi-directional, mutually-constituted character of psyche and culture. Psychological constitution of cultural worlds To investigate the psychological constitution hypothesis, we first conducted a field study (Study 1) in which we compared commemoration of BHM in effectively segregated high schools: one set of “White-Majority” schools where European American students were a near-exclusive majority, and another set of “Black-Majority” schools where African American and Latina/o students (i.e., racial/ethnic minorities across the U.S. in general) were in the numerical majority1. Consistent with discovery phases of the scientific method, the goal was to explore the everyday material affordances for BHM commemoration in different school communities. In Study 2, we exposed White American university students to photographs of BHM displays that we documented in Study 1. An intentional worlds framework suggests that BHM representations from different communities are material expressions of different understandings and desires: that is, identity-consistent traces of beliefs about what is right and good deposited in the structure of everyday ecologies. To the extent that cultural products in White American settings carry understandings that resonate with White Americans beliefs and desires, the psychological constitution hypothesis proposes that (1) White American undergraduates will express greater familiarity and preference for BHM displays from White-Majority schools than Black-Majority schools and (2) this pattern of differential recognition and preference will vary as a function of collective identification. Cultural constitution of psychological experience To investigate the cultural constitution hypothesis, we conducted a pair of between-participants experiments in which we exposed White American undergraduates to one of two treatment conditions or a third, control condition. In the treatment conditions, participants either engaged with photographs of BHM displays (Study 3) or facts reflecting BHM themes (Study 4) that we observed in either White-Majority schools or Black-Majority schools. We then examined the consequences of exposure (or not) to these treatments for two outcome variables related to Woodson's original motivations for proposing Negro History Week: (1) perceptions of racism in US society and (2) support for action on behalf of racial justice. Resonating with the idea that preferred representations systematically promote “preferred” identity-relevant outcomes, the cultural constitution hypothesis proposes that BHM displays or themes from Black-Majority schools will be more effective than BHM displays or themes from White-Majority schools (and control conditions) at increasing perception of racism and support for remediation of racial injustice. Implicit in our theoretical framework is the idea that BHM representations from Black-Majority schools afford support for anti-racism policies because they alert viewers to the ongoing significance of racism in contemporary U.S. society. This implies a mediation hypothesis whereby the effect of BHM representations on perception of racism exerts an indirect, facilitative effect of BHM representations on support for remedial social justice policy. The design of Studies 3 and 4 permits a test of this hypothesis. Study 1 Study 1 was a field study in which we used qualitative methods to investigate the ways in which local high schools commemorated BHM. [The first author] visited research sites to interview school personnel, to conduct naturalistic observations of material artifacts and display practices, and to collect photographic documentation of BHM displays. The epistemological goal was to sketch the cultural ecology of BHM commemoration in keeping with initial phases of the scientific method that emphasize discovery and naturalistic description of everyday worlds (see Barker, 1968). An institutional review board evaluated and approved all studies. Methods [The first author] contacted 16 schools in a large Midwestern U.S. metropolitan area and requested permission to observe educational displays during the month of February. Of these 16 schools, 12 had centralized BHM displays. In 7 of these 12 high schools, White students were the majority (M = 86%; range = 84–92%); in the remaining 5 schools, White students were in the minority (M = 16%; range = 2–28%). [The first author] documented “official,” centralized displays located in libraries, cafeterias, classrooms, and hallways, but also photographed any posters, pictures, signs, or miscellaneous items (e.g., televised daily announcements) related to BHM commemoration throughout the schools. Current conventions for writing and publication in psychological science require that authors provide information on the identity background of participants; however, current conventions do not (yet) require the disclosure of author or researcher identity. Indeed, conventional practice implies that author or researcher identity is unremarkable in “normal” circumstances and only becomes worthy of remark in “abnormal” circumstances where either participants or researcher identity deviates from the putative White standard. In contrast to this racialized imagination of the research process, a cultural-psychology perspective prescribes greater attention to researcher identity and its impact on the resulting science (e.g., Adams and Stocks, 2008). Although applicable to all psychological science, this prescription of reflexive disclosure is especially relevant in field observations, interviews, or other research that requires extensive interaction between researcher and participant (Blauner and Wellman, 1998; Wertz et al., 2011; also see Collins, 2002 for discussion of positionality). In keeping with this prescription, we note that all field research visits and empirical observations that we report in Study 1 were the work of the first author, an American Black woman whose own educational background included experiences in a predominantly Black high school (2 years), a predominately White high school (2 years), a predominately White college (3.5 years), and study for a brief period at a historically Black college (1 semester). The second author contributed to the design, analysis, and reporting of the present research. He is a White American man whose educational background included experience in predominantly White institutions and extensive work (total of more than 6 years) in African and other “Majority-World” settings of the Global South. Results and discussion This field study yielded several noteworthy observations about BHM commemoration practices. Brief interviews with staff (N = 17) revealed that responsibility for creating BHM displays often fell to school librarians, but creators also included students, teachers, guidance or diversity counselors, and (in one case) “the only Black staff” person. Displays generally emphasized Black icons from the past (e.g., Martin Luther King, Jr.), but some also included contemporary figures (e.g., Oprah Winfrey and Bill Cosby). The most frequent color scheme was a pan-African motif (i.e., red, black, green, and gold), but Kente-cloth motifs and multicultural rainbow were also common. Besides these observations, four major themes emerged. Theme #1: commercialized commemoration One notable theme was the frequency of commercially available, “pre-packaged” BHM materials. A particularly noteworthy example of such materials is a poster from the department store chain, Target (see Figure 1), which [the first author] observed in 25% (i.e., 3 of 12) of the schools that she visited. The poster shows images of African American inventors and products that they invented. Accompanying the images are two textual statements. One statement in small print invites viewers to “Join Target in celebrating the innovative spirit of African Americans.” The other statement in large print adds, “They didn't wait for opportunity. They invented it.” Figure 1 Example of a commercially-produced commemoration of Black History Month Representation by Target, Inc., Study 1. The Target poster is typical of pre-packaged materials and illustrates well the celebratory emphasis on heroic achievements of educators, inventors, athletes, and entertainers, with corresponding silence about contextual circumstances or structural barriers that required heroic action. Pre-packaged materials sometimes had information about each actor's contributions, but often included only the person's name and picture. With the exception of one display that included a timeline, the kits exclusively showcased people rather than places or events (for example, see Figure 2). Figure 2 Example of a commercially-produced centralized display for Black History Month from a White-Majority school, Study 1. Theme #2: celebrating diversity Materials at some schools linked BHM to larger issues of cultural diversity rather than “Black history” specifically. Examples of this theme include the posters in Figure 3, which not only make explicit reference to diversity (e.g., “Diversity is the one true thing we all have in common”), but also provide a visual representation of the theme through use of the multicultural rainbow motif or depiction of human faces. At one school where White students were the majority, the librarian explained that the diversity theme purposefully connected students to the display: “We tried to pull books to reflect hopefully every student in the building. We had books in there as far as different races go, different religions …we have multiracial, gay and straight, any kind of student.” Implicit in such statements is a concern that the purpose of BHM commemoration is not self-evident in schools without a large Black student population, and creators of displays may emphasize broader themes of diversity rather than history to make the occasion relevant to all students. Figure 3 Examples of materials from a White-Majority school that linked Black History Month to larger issues of cultural diversity, Study 1. Theme #3: celebrating achievements Another notable theme was the extent to which BHM representations focused on the individual achievements of African Americans. In many cases, materials highlighted individuals who were famous “firsts” (e.g., Jackie Robinson, famous for breaking the racial barrier within American baseball). Such representations alluded to racial barriers without specifically referring to them. There was also an emphasis on Civil Rights activists (i.e., Martin Luther King, Jr.; Rosa Parks); though, the racist conditions precipitating the Civil Rights Movement minimally figured into commemoration materials. Interview respondents were sometimes explicit about this de-emphasis of racism themes. One Diversity counselor (and designer of a BHM display at a majority White school) expressed agreement with a co-worker who had described her ideal BHM display as one that focused on positive contributions: Our students learn so much about slavery and different ways that people have been oppressed, but for BHM, in particular, to me that's a time to celebrate. Rather than watching the 12 h of Roots and be sad and defeated and cry—which is important, but not necessarily when it comes time to celebrate—like I'd rather, if the kids are going to watch things, for instance, I'd like them to see Something the Lord Made with Mos Def …or Soul Food. Movies that really the students can connect to and maybe learn something about themselves or the kids that they have class with. From this perspective, BHM provides teachers with an opportunity to celebrate African American culture. However, efforts to go beyond themes of racism and oppression in America's history may reinforce somewhat stereotypical portrayals of contemporary African American communities. This may especially be the case when individuals are presented as the exception to the rule. Theme #4: acknowledging racism The theme of racism was typically absent from BHM displays aside from serving as background for celebrating achievements. Despite this general de-emphasis of racism, some schools did have additional representations around the school that made more explicit reference to racism in discussions of slavery, “Jim Crow” segregation, and events surrounding the Civil Rights Movement (e.g., freedom rides; see Figure 4). Furthermore, interview responses clarified that presentation of racism-relevant content was not always in the form of visual displays, and some interviewees talked about integrating “the struggle” in Black History into their lessons during the month. A librarian, who was also teaching reading enhancement at a predominately Black school, indicated that lessons are sometimes timed so that they occur during BHM: I always try to make things …sort of relevant to the kids—and so the next unit we're going to do is on New Orleans. And, of course [Hurricane] Katrina and all of that stuff that happened there that affected African Americans on a greater scale than it did other groups of people and so that's another way of bringing Black history—that is current—that kids can relate to. Figure 4 Examples of materials from a Black-Majority school that linked Black History Month to historical racism, Study 1. Within this conception, BHM is the time to discuss various manifestations of racial oppression in contemporary events that differentially impacted Black communities (i.e., Hurricane Katrina). Variation in themes across research settings Although there were many commonalities across the schools, a primary interest in the present research is the possibility of systematic differences in representations of BHM as a function of the community the school serves. A precise quantification of variation in themes was impractical for two reasons. First, the multiplicity of different and sometimes conflicting representations within a given school made it difficult to quantify precisely the extent to which a school's commemoration of BHM expressed relevant themes. Second, the small sample size precluded the use of statistical tests for differences as a function of community setting. Even so, the totality of qualitative evidence from interview responses, naturalistic observation, and photographic documentation affords more holistic conclusions about the distribution of themes as a function of community settings. A summary of these ethnographic observations appears in Table 1. Taken as a whole, this evidence suggests that tendencies to emphasize collective self-enhancement over collective self-knowledge—that is to focus on values of diversity vs. history and to prioritize individual achievements over historical barriers—were all stronger in White-Majority schools than in Black-Majority schools. Table 1 Black History Month themes in central displays and other materials by school site. Site Population (%) Centralized BHM display Themes White Black Title Location COM DIV ACH RAC 1 92 3 “Celebrate Diversity and our Common Heritage” Library book display ✓ 2 91 2 “Black History Month: celebrating a Rich Heritage” Hallway bulletin board ✓ ✓ 3 92 3 None ✓ ✓ ✓ 4 90 1 “African American Achievers” Library bulletin board ✓ ✓ ✓ 5 87 4 None ✓ ✓ 6 88 5 None ✓ 7 85 9 No title/slogan; featured historical figures Hallway bulletin board ✓ ✓ 8 87 6 None 9 84 7 No title/slogan; featured historical figures Library book display ✓ ✓ 10 84 6 “The price of freedom is visible here” Library book display 11 75 8 “Celebrate Black History Month” Classroom bulletin board ✓ ✓ 12 39 43 No title/slogan; featured writers Library book display ✓ 13 29 65+ “Black History Month” Hallway display case ✓ 14 27 29* “African American history” Hallway bulletin board ✓ ✓ ✓ No title/slogan; featured historical/present-day figures Cafeteria display case ✓ 15 15 76 None ✓ ✓ ✓ 16 10 60 “Remembering Rosa Parks” Hallway bulletin board: ✓ ✓ 17 2 96 “Black Inventors and their Inventions” Hallway bulletin board ✓ ✓ ✓ BHM, Black History Month. COM, Commercialized Commemoration; DIV, Celebrating Diversity; ACH, Celebrating Achievements; RAC, Acknowledging Racism. +Site 13 was added after initial data collection for Study 1, but was considered in Study 1 analyses and utilized in Study 3. * Site 14 is the school in which Latin@/Hispanic students were in the majority (40%). Although necessarily tentative, these conclusions are consistent with education research on history representations in the U.S. (Banks, 1999; Alridge, 2006; Journell, 2008). For example, an analysis of social studies standards in nine states revealed two primary themes—“Inclusion of Individuals” and “Oppression and Emancipation vs. Culture and Contribution”—(Journell, 2008) that resonate with observations of the present study. Similarly, a comparison of historical knowledge revealed differences as a function of school setting such that White students were less likely than Black students to cite the Civil Rights Movement, Civil War, and slavery as important historical events (Epstein, 2000). The objective of Study 1 was to provide a qualitative sense for everyday practices of BHM commemoration in different school communities (Marecek et al., 1997; Denzin and Lincoln, 2012). Although this objective is especially appropriate for the naturalistic observation and description stages of the scientific method, the exploratory character of this initial study precluded strong conclusions about the distribution of themes in BHM displays. We return to this issue with respect to the present materials in Study 3, but more definitive conclusions about the distribution of themes require additional analysis of materials from a larger sample of schools. This remains a direction for future research. Study 2 Study 2 considers the psychological constitution hypothesis. To investigate this idea, we exposed White American undergraduates to photographs of BHM displays from both White-Majority and Black-Majority schools. In response to each photograph, we asked participants to make ratings of recognition and liking without providing information about the source of each display (i.e., whether from a White-Majority or Black-Majority school). If cultural products from White-Majority schools resonate better with White American beliefs and desires than do cultural products from Black-Majority schools, then White American participants will express stronger recognition and liking of BHM displays from White-Majority schools than BHM displays from Black-Majority schools. Moreover, this pattern of responses will bear traces of relevant social identity concerns. Specifically, White American identity will be more positively (or less negatively) related to positive affect, liking, and recognition for ratings of White-Majority displays than for ratings of Black-Majority displays. Method Participants A total of 52 undergraduates from a predominately-White, Midwestern American university ranged in age from 18 to 30 years of age (M = 19.22, SD = 1.98). Of participants who indicated race or ethnicity, fewer than 10% indicated a non-White race or ethnicity (Asian/Asian American, 2; African American/Black, 2; Hispanic/Latino, 1). We report analyses using only data from the 47 participants (27 men and 18 women) who indicated White and American identities. Procedure We recruited participants from an introductory psychology subject pool. A White research assistant administered the study. Participants viewed a PowerPoint presentation containing photographs of the centralized BHM displays (for example, see Figure 4) from the seven high schools with majority White Student populations (i.e., White-Majority condition) and the five high schools where American students of color were in the majority (Black-Majority condition). We assigned participants at random to view these displays in one of two, alternating-order conditions (i.e., with either a White-Majority or Black-Majority display presented first). Participants completed ratings of affective responses, familiarity, and liking as they viewed each display. Participants were unaware that the displays came from schools with different racial/ethnic profiles. After completing the rating task, participants completed measures of identification and demographics. Materials Affective responses We measured affective responses by adapting items from the Positive and Negative Affect Scale (PANAS; Watson et al., 1988). Immediately after viewing each display (and before they viewed the next display), participants rated six positive emotions (e.g., proud; Cronbach's α = 0.96 and 0.95 for White-Majority and Black-Majority displays, respectively) and six negative emotions (e.g., guilty; Cronbach's α = 0.91 and 0.95 for White-Majority and Black-Majority displays, respectively) on a scale from 1 (not at all) to 5 (extremely) in response to the following prompt: “Indicate to what extent you feel this way, right now, that is, at the present moment.” Display ratings Participants responded to six evaluative questions using a 7-point scale ranging from 1(not at all) to 7 (very much). A principal components analysis using varimax rotation yielded two reliable factors. A first, “liking” factor consisted of 4 items—e.g., How much do you like this display and To what extent would you like to see this display in your former high school—that accounted for 73.15% of the variance (Cronbach's α = 0.89 and 0.90 for White-Majority and Black-Majority displays, respectively). A second, “recognition” factor consisted of 2 items—How familiar are you with the contents of this display and To what extent does the overall display present the material accurately—that accounted for 12.39% of the variance (Cronbach's α = 0.80 and 0.77 for White-Majority and Black-Majority displays, respectively). Identification We adapted the 4-item identity subscale of the Collective Self Esteem Scale (CSE; Luhtanen and Crocker, 1992) to assess racial/ethnic identity (Cronbach's α = 0.84)2. This subscale measures individual differences in importance of a social identity category for self-definition (e.g., The racial/ethnic group I belong to is an important reflection of who I am). Participants responded to these questions with a 7-point Likert scale ranging from 1 (not at all) to 7 (very much). Results and discussion To investigate whether representations of BHM from different school settings carry community-consistent understandings and preferences, we conducted within-subjects analyses of ratings about positive and negative emotions, liking, and recognition by White American participants in response to White-Majority and Black-Majority displays. To explore the identity-relevance hypothesis, we tested whether the relationship between identification and responses to displays differed as a function of community source. Affective responses We conducted a 2 (Display Source: White-Majority school, Black-Majority school) × 2 (Affect Valence: Positive, Negative) repeated measures ANOVA to investigate the effect of display source on affective responses. Results revealed main effects of display source, F(1, 42) = 19.88, p < 0.001, ηp2=0.32, and valence, F(1, 42) = 105.33, p < 0.001, ηp2 = 0.72, qualified by a Display × Valence interaction, F(1, 42) = 21.33, p < 0.001, ηp2 = 0.35. Follow-up analyses indicated a hypothesized effect of display source that was primarily true for positive emotion items. Participants reported more positive emotions while viewing White-Majority displays (M = 2.41, SD = 0.67) than while viewing Black-Majority displays (M = 2.05, SD = 0.67), t(43) = 5.035, p < 0.001, d = 0.83. In contrast, participants indicated little negative emotion while viewing either Black-Majority displays (M = 1.20, SD = 0.36) or White-Majority displays (M = 1.13, SD = 0.21), t(44) = −1.72, p = 0.092, d = 0.39. To summarize, something about viewing representations of Black History from White-Majority schools prompted White American undergraduates to feel better than they did after viewing representations of Black History from Black-Majority schools (i.e., where American students of color were in the majority)3. Display ratings We conducted a 2 (Display Source: White-Majority school, Black-Majority school) × 2 (Evaluation Type: Liking, Recognition) repeated measures ANOVA to investigate the effect of display source on evaluations of liking and recognition. Results revealed the hypothesized main effect of display source, F(1, 46) = 76.71, p < 0.001, ηp2 = 0.63, and a main effect of evaluation type, F(1, 46) = 38.57, p < 0.001, ηp2 = 0.46, qualified by a significant Display × Evaluation Type interaction, F(1, 46) = 49.31, p < 0.001, ηp2 = 0.52. Although the difference between ratings of White-Majority and Black-Majority displays was large in both cases, paired-samples t-tests indicated that it was greater on the liking dimension (respective Ms = 3.83 and 2.88, SDs = 0.77 and 0.82), t(46) = 10.25, p < 0.001, d = 1.68 than on the recognition dimension (respective Ms = 4.12 and 3.72, SDs = 0.86 and 0.86), t(46) = 4.98, p < 0.001, d = 0.78. Consistent with the psychological constitution hypothesis, these White American participants expressed relatively high recognition and liking of BHM material from White-Majority schools, but they expressed less recognition and profoundly less liking of BHM material from Black-Majority schools. Identity relevance What accounts for the pattern of observed differences in affect, liking, and recognition as a function of display source? Although there might be many factors that differ systematically across sets of BHM displays, an intentional worlds analysis suggests that ratings diverge, in part, because the displays resonate differently with beliefs and desires associated with White American identity concerns. To evaluate this feature of the psychological constitution hypothesis, we conducted one-tailed tests of within-subject differences in correlation coefficients. Results revealed the hypothesized pattern. We created a single index of affective response for each participant by subtracting negative affect scores from positive affect scores. The relationship between racial identification and this affective index was less positive (more precisely, more negative) in reaction to Black-Majority displays (r = −0.273, p = 0.03) than in response to White-Majority displays (r = −0.057, p = 0.36), t(43) = 1.76, p = 0.04, Cohen's q = 0.22 (Figure 5). Likewise, the relationship between racial identification and judgments of liking was more negative in reaction to Black-Majority displays (r = −0.134, p = 0.19) than in reaction to White-Majority displays (r = 0.089, p = 0.28), t(44) = 1.92, p = 0.03, Cohen's q = 0.22 (Figure 6). Finally, the relationship between racial identification and judgments of recognition was more negative in reaction to Black-Majority displays (r = −0.068, p = 0.33) than in reaction to White-Majority displays (r = 0.083, p = 0.29), t(44) = 1.61, p = 0.06, Cohen's q = 0.15 (Figure 7). Figure 5 Relationship between racial identity and affective responses to Black-Majority and White-Majority Black History displays, Study 2. Figure 6 Relationship between racial identity and display liking among Black-Majority and White-Majority Black History displays, Study 2. Figure 7 Relationship between racial identity and display recognition among Black-Majority and White-Majority Black History displays, Study 2. Summary Study 2 provides evidence of hypothesized preferences among White American college students for culturally consistent constructions of BHM. Although they were unaware of the source of displays, White American undergraduates exposed to photographs of BHM displays reported more positive affect, greater recognition, and greater liking for displays from White-Majority schools than displays from Black-Majority schools. An intentional worlds perspective suggests that these divergent reactions reflect the different affordances for identity present in displays from different settings. That is, displays from White-Majority schools are likely to rehearse content that affords White American students positive social identity, but displays from Black-Majority schools are likely to include themes that challenge White American identity. Although the exact features of BHM displays that afford divergent reactions remains an open question, evidence for the identity relevance of these features comes from hypothesized differences in the relationship between display ratings and White American identification. The relationship between White American identification and responses to Black History representations was significantly more negative in reaction to displays from Black-Majority schools than White-Majority schools. Study 3 Studies 1 and 2 provide evidence for one half of the intentional worlds framework, corresponding to the psychological constitution hypothesis. Specifically, results suggest that BHM representations in different communities are not neutral constructions of the past, but instead are directed products that realize and objectify community-specific beliefs and desires. Study 3 examines the other half of the intentional worlds framework, corresponding to the cultural constitution hypothesis. This idea proposes that realizations of BHM in different communities are not inert matter, but instead exert directive force that influences subsequent activity toward “desirable” or identity-serving ends. Method Participants A total of 136 undergraduates (74 women and 62 men) from a predominately-White, Midwestern U.S. university ranged in age from 18 to 27 (M = 19.15, SD = 1.59). We retained data from the 123 participants who identified as American. Of these participants, 73.5% indicated White or European race or ethnicity (n = 100), 21% indicated another race or ethnicity (Asian/Asian American, 10; African/African American, 7; Hispanic/Latino, 4; American Indian, 1), and 5.5% did not indicate a category. Procedure After obtaining informed consent, we randomly assigned participants to one of two BHM conditions (White-Majority or Black-Majority) or a third, control condition. In the Black-Majority condition, participants viewed photographs of six displays from Black-Majority schools4. In the White-Majority condition, we selected photographs of six displays at random from the original set of seven displays from White-Majority schools that participants rated in Study 2. To encourage engagement with displays, participants rated each display on dimensions of liking and recognition (as in Study 2). After engaging with the displays, participants completed a questionnaire. In the Control condition, participants completed the primary dependent measures (i.e., perceptions of racism, support for anti-racism policies) before viewing BHM photographs. Near the end of the survey, all participants responded to an open-ended prompt: Please think about the History display that you liked the most. Please describe it and tell us what you liked about the display. Last, participants completed the demographic questions. Dependent measures Perceptions of racism Four items (α = 0.73) assessed perception of racism in American society (adapted from Adams et al., 2006). Participants used a 7-point Likert scale (1, not at all due to racism; 7, certainly due to racism) to indicate the extent to which features of U.S. society were due to racism (e.g., the disproportionate number of African Americans and Latinos in the criminal justice system). Support for anti-racism policies Four items (α = 0.63) assessed endorsement of policies aimed to address racial inequalities in the U.S. (e.g., Public and private universities should make every effort to attract qualified minority students). Participants indicated the extent to which they agreed with the policies on a 7-point Likert scale (1, Strongly disagree; 7, Strongly agree). Results and discussion To provide focused tests of the hypothesized effect of the historical representation manipulation on outcomes of interest, we conducted orthogonal planned contrasts with codes of (−1/3, −1/3, 2/3,) and (−1/2, 1/2, 0) for control, White-Majority, and Black-Majority representation conditions, respectively. The first contrast tested the primary hypothesis that Black-Majority representations will produce greater perceptions of racism in ambiguous events and endorsement of anti-racism policy than the control and mainstream, White-Majority conditions. The second contrast tested a secondary hypothesis that perceptions of racism and anti-racism policy support will be greater among participants in the White-Majority condition than the control conditions. Although contrast analyses provide the primary tests of hypotheses, we follow convention and report results of omnibus ANOVA tests. Effects of the display manipulation Perceptions of racism The omnibus ANOVA revealed a significant effect of historical representation on racism perception, F(2, 120) = 4.47, p = 0.013, ηp2 = 0.069. The first contrast revealed the hypothesized effect of historical representation, t(120) = 2.52, p = 0.013, d = 0.46. Participants exposed to Black-Majority representations of Black History perceived greater racism in U.S. society (M = 4.13, SD = 1.11) than did participants exposed to White-Majority BHM representations (M = 3.84, SD = 1.11) and participants in the control condition (M = 3.38, SD = 1.21). The second contrast indicated that the difference between the latter two conditions was also significant, t(120) = 1.90, p = 0.060, d = 0.40. The pattern of means—greatest in the Black-Majority condition, followed by the White-Majority condition, and least in the control condition—suggests a linear trend in effectiveness of Black History displays in promoting perception of racism. Support for anti-racism policies The omnibus ANOVA revealed a marginal effect of historical representation on support for anti-racism policy, F(2, 120) = 2.64, p = 0.075, ηp2 = 0.04. The first contrast revealed the hypothesized effect of historical representation, t(120) = 2.10, p = 0.038, d = 0.37. Participants exposed to Black-Majority representations of BHM indicated greater support for anti-racism policy (M = 4.01, SD = 1.21) than did participants exposed to mainstream, White-Majority representations (M = 3.77, SD = 0.93) and participants in the control condition (M = 3.47, SD = 1.01). The second contrast indicated that the difference between the latter two conditions was not significant, t(120) = 1.359, p = 0.177, d = 0.0.31. This pattern of means suggests Black-Majority representations have greater effectiveness in promoting support for anti-racism policy than White-Majority representations and the control. Mediation analyses To test the hypothesis that the effect of BHM representations on perception of racism mediates an indirect effect of BHM representations on support for remedial social justice policy, we used a statistical procedure for mediation analysis with a multicategory independent variable (Hayes and Preacher, 2014). Specifically, we created two contrast-coded variables from the three-level independent variable in the manner that we did for ANOVAs. This permitted a consideration of direct and indirect effects of Black-Majority BHM representations relative to the other two (i.e., White-Majority BHM and control) conditions. Consistent with the contrast analysis, participants in the Black-Majority BHM condition perceived more racism [b = 0.53, SE = 0.10, t(120) = 2.28, p = 0.024] compared to other conditions. In turn, racism perception significantly predicted support for anti-racist policy, [b = 0.52, SE = 0.07, t(120) = 7.51, p < 0.001]. When accounting for the contrasts (independent predictors) and racism perception (proposed mediator) as simultaneous predictors of anti-racism policy, the overall model was significant, R2 = 0.35, F(3, 119) = 21.39, p < 0.001. A bootstrapping analysis with 5000 iterations indicated that the relative indirect effect of the Black-Majority BHM representations (Contrast 1) through racism perception on anti-racism policy endorsement was significant (i.e., the 95% bias corrected confidence interval of 0.0490 to 0.5177 excludes zero), Figure 8. The relative direct effect of Black-Majority BHM representations was non-significant, b = 0.11, SE = 0.18, t(120) < 1. This pattern of results is consistent with the interpretation that Black-Majority BHM representations promote support for anti-racism policy because they afford perception of racism. Figure 8 Multicategorical mediation model of the effect of Black History Month displays on policy support via racism perception, Study 3. Indirect effect 95% Confidence Interval [0.0490; 0.5177]. C1, contrast 1; C2, contrast 2; *p < 0.05, +p = 0.06. Open-ended responses To explore the “active ingredients” of displays that produce differences in liking and recognition (in Study 2) or promote perception of racism and support for anti-racist policy (Study 3), we asked participants to indicate their favorite display and to describe what they liked about it. To analyze responses to this open-ended item, we developed a coding scheme based on a priori research questions and field observations from Study 1. Specifically, we created two categories to capture dynamics of critical, self-knowledge motivations vs. celebratory, self-enhancement motivations in BHM representations (e.g., Dagbovie, 2004; Journell, 2008; Chapman-Hilliard and Adams-Bass, 2015). Two independent raters used this scheme to code two thirds of the essays each: Rater A coded essays 1-45, Rater B coded essays 90-135, and both raters coded essays 45–90. Table 2 describes the binary coding categories (i.e., for judgments of absence/presence) and reports interrater reliability for each category based on the jointly coded one-third of essays. Discrepancies were resolved by averaging the coders' responses. We then summed across instances of the coding categories struggle and historical barriers, historical racism, specific historical events, specific historical people, and educational value to provide a total score for each participant on the dimension of critical self-knowledge. We summed across instances of cultural diversity, inclusion/exclusion, aesthetic appeal, and neutral design to provide a total score for each participant on the dimension of celebratory self-enhancement. The identity-relevance hypothesis suggests that celebratory self-enhancement themes will be more evident, but critical self-knowledge themes less evident, in responses to White-Majority displays than responses to Black-Majority displays. Table 2 Descriptions of essay coding categories and inter-rater reliability statistics, Study 3. Category Theme Item scale Kappa Mentions educational value or informativeness of display CSK 0, no; 1, yes 0.526, p < 0.001 Mentions specific historical events or time periods CSK 0, no; 1, yes 0.291, p = 0.037 Mentions specific and individual historical or contemporary people CSK 0, no; 1, yes 0.928, p < 0.001 Mentions an instance of historical racism, discrimination, civil rights CSK 0, no; 1, yes 0.000, p = 0.500* Mentions themes of struggle, hardship, challenge or historical barriers African Americans face CSK 0, no; 1, yes 0.023, p = 0.877* Mentions aesthetic values or dimensions CSE 0, no; 1, yes 0.615, p < 0.001 Mentions diversity values CSE 0, no; 1, yes 0.459, p = 0.001 Mentions inclusion CSE 0, no; 1, yes 0.692, p < 0.001 Mentions seemingly neutral, organizational, and structural aspects of the display CSE 0, no; 1, yes 0.259, p = 0.082 CSK, Critical Self Knowledge. CSE, Celebratory Self-Enhancement. * Of the 44 overlapping essays, the coders agreed that 42 of the essays did not mention historical racism or historical barriers. The Kappas reflect a divergence on 2 essays. For racism, coder A indicated that 2 essays reflected this category while coder B did not. For barriers, coder A indicated that 1 essay contained this theme and coder B indicated that a different essay reflected this category. Critical self-knowledge We operationalized critical themes as those that included consciousness of historical barriers and promoted a focus on Black History information and facts (e.g., specific events and individuals). Some participant responses reflected a focus on specific BHM content. …there seems to be a lot of information in the center of the display. I liked that there are many different pictures of Rosa Parks and various magazine covers, newspapers and books, which shows what an important part she had in ending segregation [#15, White, female]. Similarly, the response of a participant in the Mainstream display condition reflected a focus on historical barriers faced by the Tuskegee Airmen. “Remembering the Past …Shaping the future,” this quote next to the picture is important because we acknowledge what past struggles that have been made for African-Americans to give us the strength to continue to do better as a people [#107, Black, female]. Celebratory self-enhancement We operationalized celebratory representations of BHM as those that deemphasized content in favor of broader inclusion, feel-good rhetoric (Trawalter et al., 2015), and a focus on relatively superficial aesthetic interests. In some cases, participants specifically commended a focus on broader diversity themes because they avoided negative information and afforded good feelings. I liked the diversity theme. I feel that educating about others' cultures is best served on a more passive level rather than displays with the in your face approach. This display appeals to several minority groups, religious backgrounds, and achievements [#69, White, male]. I enjoyed the fact it did not push the thought of discrimination as much. The books shown show that it leaves the reader to find out about diversity and the African American movement. The display shows a door that can be opened to introduce African American information [#21, White, male]. Quantitative analyses of themes To investigate whether responses to displays from different communities differed in the presence of these themes, we conducted a 2 × 2 mixed-model ANOVA with coding category (critical and celebratory) as the within-participant factor and historical representation condition (White-Majority and Black-Majority) as the between-participants factor. There was no main effect of condition, F(1, 119) < 1.0, p = 0.60, but there was a main effect of coding category, F(1, 119) = 10.57, p = 0.001, ηp2 = 0.08. Generally, participants' open-ended responses referenced more celebratory themes (M = 1.60, SD = 0.90) than they did critical themes (M = 1.10, SD = 1.03), t(120) = 3.86, p < 0.001, d = 0.36. The interaction was also significant, F(1, 119) = 10.57, p = 0.001, ηp2 = 0.08. Analyses to probe the interaction revealed support for hypothesized differences in display content. BHM displays from White-Majority schools (M = 1.79, SD = 0.87) elicited more celebratory themes in participants' open-ended responses than did displays from Black-Majority schools (M = 1.31, SD = 0.87), t(119) = 2.94, p = 0.004, d = 0.55. In contrast, responses to displays from Black-Majority schools (M = 1.31, SD = 1.17) tended to include more critical themes than did responses to displays from White-Majority schools (M = 0.97, SD = 0.92), although this difference did not reach conventional levels of statistical significance, t(119) = −1.82, p = 0.071, d = 0.33. These differences in frequency across open-ended responses provide some evidence that displays from different school settings bear (or afforded perception of) systematically different themes. We propose that this difference between critical and celebratory representations of history might be the “active ingredient” that affords the differential effects of the displays that we observed in Studies 2 and 3.That is, displays from Black-Majority schools may have produced more dislike among White American students in Study 2 in part because they include more critical representations (that focus on historical barriers and racism) than do displays from White-Majority schools. Likewise, the relative emphasis on past racism in displays from Black-Majority schools may have alerted students in Study 3 to both the ongoing legacy of racism in the present and the need for energetic measures to overcome that legacy. Summary Consistent with the cultural constitution hypothesis, results of Study 2 provide some evidence that BHM displays from Black-Majority schools were more effective at promoting perception of racism and support for anti-racist policy than were displays from White-Majority schools and a no-display control condition. Results also indicate that displays from White-Majority schools were somewhat effective at producing perceptions of racism relative to the no-display control condition, but just not as effective as displays from Black-Majority schools. As existing cultural products, representations of BHM from White-Majority schools may have included a mix of content, some of which promoted perception of racism and support for anti-racist policy and some of which promoted denial of racism and opposition to anti-racist policy. Indeed, analysis of open-ended responses reveal that critical history themes were present in White-Majority displays, although not to the same extent as in Black-Majority displays. Study 4 A primary strength of the preceding studies is ecological validity and the use of existing cultural products to examine the hypothesis that representations of history reflect and promote goals of different communities. Consistent with the psychological constitution side of the intentional worlds framework, Study 2 provides evidence that these cultural products bear the beliefs, desires, and preferences of people in the communities from which they originate. Consistent with the cultural constitution side of the intentional worlds framework, Study 3 provides evidence that these cultural products differentially promote identity-relevant action. A corresponding weakness of this “existing-products” approach is that, because displays from different schools incorporate many overlapping themes and elements, it is difficult to isolate the conceptually important features that serve as active ingredients that produce observed effects5. In order to provide a more precise test of the cultural constitution hypothesis and the role of different representations of history in promoting identity-relevant perception and action, Study 4 includes an experimental manipulation of celebratory and critical constructions of BHM. Method Participants A total of 37 white American undergraduates (20 women and 17 men) from a predominately-White, Midwestern U.S. university ranged in age from 18 to 25 years of age (M = 19.54, SD = 1.52). Procedure After obtaining informed consent, we assigned participants at random to one of three conditions associated with different constructions of American history. After exposure to one of the different sets of historical information, each participant completed measures of racism perception, policy endorsement, and demographics (similar to Study 3). Materials Historical fact manipulation The control condition modeled a “standard” American history approach and presented participants with 12 facts from which Black Americans and other minority groups were largely absent (e.g., Benjamin Franklin, one of the most distinguished scientific and literary Americans of his era, was the first American diplomat). The historical achievements condition replaced 5 of the 12 facts from the control condition with information about celebratory achievements of Black Americans to model a BHM approach found primarily in White-Majority schools and to a lesser extent in Black-Majority schools (e.g., As a mission specialist aboard the Shuttle Endeavour, Mae Jemison was the first African American woman to enter outer space). The historical barriers condition replaced the same 5 facts with critical historical information about racial barriers in American history to model a BHM approach primarily found in Black-Majority schools (e.g., Dred Scott, a slave, sued for his freedom in 1847. The Supreme Court ruled that he was property and could not sue in federal court). To encourage engagement with the facts, participants rated each fact on dimensions of importance and familiarity. After reading the facts, participants completed a questionnaire. Perceptions of racism We added six items to the racism perception measure from Study 3 (10 items; α = 0.82; 1, not at all due to racism; 7, certainly due to racism). Support for anti-racism policies The same four items in Study 3 assessed endorsement of policies aimed to address racial inequalities in the U.S. (α = 0.65; 1, Strongly disagree; 7, Strongly agree). Results and discussion Again, we conducted ANOVAs with orthogonal planned contrasts using codes of (−1/3, −1/3, 2/3,) and (−1/2, 1/2, 0) for control, celebratory historical achievement, and critical historical barrier conditions, respectively. The first contrast tested the primary hypothesis that the relatively critical, barriers condition produced greater perceptions of racism and endorsement of anti-racism policy than the relatively sanitized, celebratory achievements and control conditions. The second contrast tested whether perceptions of racism and policy support differed for participants in the latter two conditions. Although these orthogonal planned contrasts provided the primary test of hypotheses, we again follow convention and report results of the omnibus ANOVA tests. Perceptions of racism The omnibus ANOVA revealed a marginal effect of historical representation on racism perception, F(2, 36) = 2.57, p = 0.092, ηp2 = 0.13. A more precise test of the hypothesis comes from planned contrasts. The first contrast indicated that participants in the historical barriers condition (M = 4.74, SD = 0.79) perceived more racism than did participants in the historical achievements (M = 4.01, SD = 1.26) and control conditions (M = 3.91, SD = 0.80), t(34) = 2.44, p = 0.02, d = 0.80. The second contrast was not significant, t(34) < 1. Support for anti-racism policies The omnibus ANOVA revealed an effect of historical representation on anti-racism policy support, F(2, 34) = 3.66, p = 0.036, ηp2 = 0.18 The first contrast revealed the hypothesized effect such that endorsement of anti-racism policies was greater among participants in the historical barriers condition (M = 4.29, SD = 0.84) than in the historical achievements (M = 3.21, SD = 1.19) and control (M = 3.54, SD = 0.97) conditions, t(34) = 2.72, p = 0.010, d = 0.91. The second contrast was not significant, t(34) < 1. Mediation analyses We used the same procedure for mediation analysis with a multicategory independent variable (Hayes and Preacher, 2014) to test the hypothesis that the effect of information about historical barriers on perception of racism mediates an indirect effect of information about historical barriers on support for anti-racism policies. Specifically, we created two contrast-coded variables from the three-level independent variable to permit evaluation of direct and indirect effects of the historical barriers condition relative to the other two (i.e., historical achievements and control) conditions. Consistent with the contrast analysis, exposure to information about historical barriers exerted a positive effect (compared to other conditions) on perception of racism [b = 0.78, SE = 0.35, t(34) = 2.26, p = 0.031]. In addition, racism perception significantly predicted support for anti-racist policy, [b = 0.44, SE = 0.16, t(34) = 2.69, p = 0.011]. When accounting for the contrasts (independent variables) and racism perception (proposed mediator) as simultaneous predictors of anti-racism policy, the overall model was significant, R2 = 0.34, F(3, 33) = 5.30, p = 0.004. Bootstrapping tests with 5000 iterations indicated that the relative indirect effect of the critical historical barrier representations (Contrast 1) through racism perception on anti-racism policy endorsement was significant (i.e., the 95% confidence interval of 0.0307 to 0.9485 excludes zero), Figure 9. The relative direct effect of critical historical barrier representations non-significant, b = 0.57, SE = 0.35, t(34) = 1.63, p = 0.11. This pattern of results is consistent with the interpretation that critical BHM representations depicting historical barriers increase support for anti-racism policy because they increase perceptions of racism. Figure 9 Multicategorical mediation model of the effect of Black History Month facts on policy support via racism perception, Study 4. Indirect effect 95% Confidence Interval [0.0307; 0.9485]. C1, contrast 1; C2, contrast 2; *p < 0.05. Summary Similar to Study 3, participants in the historical barriers condition (modeled after BHM displays from Black-Majority schools) perceived more racism in ambiguous events and endorsed anti-racism policy to a greater extent than did participants in both the historical achievements (modeled after BHM displays from White-Majority schools) and control conditions. Perhaps reflecting more precise control over the content of representations, these effects were stronger in Study 4 than Study 3. General discussion This research draws upon a diverse methodological toolkit—including qualitative field research, quantitative analyses, and experimental design—to investigate sociocultural variation in BHM representations and their consequences for perceptions of racial inequality in the U.S. The intellectual foundation of the project lies in the theoretical perspective of cultural psychology and its emphasis on the intentional character of everyday worlds. In one direction, (the psychological constitution hypothesis), the idea of intentional worlds emphasizes the directed character of everyday cultural ecologies. The relative prominence of different BHM representations in a given environment does not emerge by accident. Instead, the prominence or absence of particular ideas in material reality is the residue of purposeful activity: that is, the identity-consistent deposit of what seems good and true into local context. In the other direction (the cultural constitution hypothesis), the idea of intentional worlds emphasizes the directive character of everyday cultural ecologies. The BHM representations prominent in different ecologies are not the inert end-product of previous activity. Instead, they systematically afford courses of action consistent with identity concerns of different communities. Support for the psychological constitution hypothesis came from field research in public schools (Study 1) and participants' open-ended reactions to photographs (Study 3). Celebratory BHM representations were most commonly found in material from White-Majority schools, and BHM representations highlighting historical racism were most commonly found in Black-Majority schools. Study 2 provides additional support for the psychological constitution hypothesis. White American participants expressed greater positive affect, familiarity, and liking for White-Majority BHM displays than Black-Majority BHM displays. Moreover, the relationship between White American identification and responses to BHM representations was significantly more negative in response to displays from Black-Majority schools than in response to displays from White-Majority schools. Together, these results suggest the extent to which BHM displays are not identity-neutral. Instead, displays from White-Majority settings reflect and objectify preferences and understandings that resonate with dominant constructions of White Americans identity. Displays from Black-Majority settings reflect and objectify preferences and understandings that conflict with dominant constructions of White American identity. Studies 3 and 4 provide support for the cultural constitution hypothesis. In Study 3 we exposed participants to existing BHM displays from different communities to investigate the extent to which these cultural tools afford different tendencies of racism perception and policy support. Results indicated that exposure to the BHM displays from Black-Majority schools (i.e., the same representations for which White Americans expressed dislike in Study 2) led subsequent participants both to perceive a greater role of racism in American society and to express greater support for anti-racism policy than exposure to both BHM displays from White-Majority schools (i.e., the same ones for which White Americans expressed a preference in Study 2) and a no-display control condition. In Study 4, we experimentally manipulated exposure to themes that we distilled from existing representations. Results again indicated that exposure to critical history representations that emphasized historical barriers led subsequent participants to perceive greater influence of racism in American society and to express greater support for anti-racism policy than did exposure to both celebratory BHM representations that emphasized individual achievements (i.e., a theme that was more typical of displays from Black-Majority schools than White-Majority schools) and representations that made no mention of Black contributions to American history. Together, these results suggest the extent to which BHM displays are not psychologically inert objects. Instead, displays from different communities bear beliefs and desires of their producers that systematically direct perception and action toward different ends. White Americans preferences (in Study 2) for BHM displays from majority White American schools are not accidental. Instead, whether or not they are conscious of the source of their preferences, White American undergraduates may prefer these displays precisely because these displays afford denial of racism, opposition to anti-racism policy, and preservation of the system of racial domination from which they benefit. As students and teachers act on these preferences, choose to emphasize diversity or individual accomplishments, and omit information about racist barriers, they selectively reproduce cultural worlds of sanitized history representations that in turn afford denial of racism and weak support for anti-racism policy. Limitations and future directions Although results are consistent with hypotheses, the research is not without limitations. We mention three that constitute important directions for future research. One limitation is the basis in a relatively narrow sample of BHM displays from schools in a single U.S. metropolitan area. Accordingly, we make no claims that results reflect the broader distribution of BHM themes across a wider set of cultural products and cultural ecologies. Indeed, we are doubtful of any claims about characteristic tendencies that apply uniformly across African American or European American settings. Variation in BHM constructions within and across communities (and over time) remains a fruitful direction for research. Another limitation concerns the narrow sample of participants (e.g., along dimensions of race-ethnicity, age, and educational background). Although the focus on White American spaces is appropriate for addressing our theoretical interest in historical representations as tools of domination, an important task for future research is to consider identity-relevant influences on reproduction of historical representations among people from historically oppressed groups. For example, an interesting question for future research is the extent to which dynamics of BHM commemoration among Black Americans vary as a function of the “double consciousness” (DuBois, 1989/1903, p. 5) associated with American-ness and Black-ness. Black American students vary in preferences for Black History as a function of their prior racial socialization experiences (Thornhill, 2014), and their preferences may change across the life course if they move from predominantly Black American to predominantly White American spaces (Starck et al., unpublished manuscript). Another reason to consider a broader diversity of participants concerns consequences of engagement with different BHM constructions. The present research demonstrates that an emphasis on critical history narratives promotes greater awareness of racism in contemporary U.S. society. Although we have interpreted this outcome in a relatively positive light (as a precursor to anti-racist action among White Americans), previous research has emphasized that racism perception can put people with marginalized identities at risk for negative psychological outcomes (e.g., Pinel, 1999; Mendoza-Denton et al., 2002; cf. Tatum, 1997). Extension of the present research to participants with marginalized identities would permit investigation of this important dilemma that parents and educators face regarding socialization about racism (Tatum, 1987; Hughes et al., 2006). Finally, our investigation focused on the implications of historical representation for perception of racism and support for anti-racist policy. We did not consider racist attitudes or other individual difference measures that have figured more prominently in social psychological research. Our focus on outcomes of racism perception and support for anti-racist policy is consistent with emerging perspectives that express misgivings about constructions of racism as a matter of individual prejudice (i.e., “prejudice problematic”; Wetherell and Potter, 1992; see also Leach, 2002; Paluck, 2009; Wright and Lubensky, 2009; Hammack, 2011; Dixon and Levine, 2012). However, it remains unclear whether the patterns we observed in the present research would extend to measures of individual prejudice. Conclusion: the intentionality of everyday worlds Without downplaying limitations, the primary contribution of the present project is to provide a dynamic account of the mutual constitution of culture and psyche at the “ecological” level of local affordances and everyday activity. In the process, the project illuminates a cultural psychological approach to the collective character of psychological experience. Cultural dynamics Social-psychological research in the field of cultural psychology has typically emphasized comparison across settings of psychological tendencies (and occasionally, cultural products; Snibbe and Markus, 2005; Markus et al., 2006; Miyamoto et al., 2006; see Morling and Lamoreaux, 2008). Such work has been indispensable for illuminating both sociocultural variation in psychological processes and the particular constructions of reality that underlie conventional scientific wisdom. Critics of this work have noted that it often presents an overly static and essentialist understanding of culture associated with Orientalism and other kinds of cultural imperialism (e.g., Gjerde, 2004; Okazaki et al., 2008). Against this background, the present work contributes to a more dynamic account of cultural variation that emphasizes the active, mutual constitution of cultural context, and psychological experience (Shweder, 1990; Markus and Hamedani, 2007). Regarding the cultural constitution hypothesis, our investigation focuses less on broad, national spaces, and instead emphasizes a more local form of cultural-ecological variation: the cultural practices and material products of BHM displays in schools serving different communities. This emphasis resonates with ecological perspectives that consider the mutual constitution of organism and habitat at a more “micro” level of analysis than is typical in canonical forms of cultural psychology (Barker, 1968; Gibson, 1977; with respect to cultural psychology, see also Uskul et al., 2008; Oishi and Graham, 2010). It likewise makes contact with sociocultural-historical perspectives in cultural psychology (e.g., Cole, 1995) that emphasize cultural constitution of psychological experience via mediation of cultural tools. The result is a dynamic account of cultural psychology that avoids broad, essentialist generalizations about reified, monolithic groups. Regarding the psychological constitution hypothesis, our investigation focuses on the active reproduction of everyday worlds through repeated acts of preferential selection. Although previous work has emphasized the role of social cognitive processes or personal motivations on preferential selection of cultural units (e.g., Zajonc, 1980; McIntyre et al., 2004; Savani et al., 2008), the present project extends consideration to the role of social identification and associated motivations in preferential reproduction of cultural reality (e.g., “To what extent would you like to see this display in your former high school”). By considering ways in which people who occupy positions of dominance act on identity considerations to reproduce identity-charged realities, the present research lays the foundation for a cultural psychology of power, privilege, and oppression. More generally, this focus on preferential selection and active maintenance contributes to a dynamic account of cultural psychology that retains an appreciation for human agency (Gjerde, 2004). Collective self-regulation An important implication of this account is that even when everyday realities appear to be static and unchanging across generations, this apparent stability masks dynamic activity as people repeatedly select (or de-select) features that resonate with their understandings and desires. Moreover, the results of these acts of preferential selection (or de-selection) are not merely end-products of a behavioral cycle; instead, they construct everyday realities that constitute the “facilitating conditions” (Kroeber and Kluckhohn, 1952) for future action. With respect to the present work, identity-relevant acts of preferential selection produce everyday worlds that are intentional in two important senses. First, these worlds are “intentional” because they are directed; that is, people in the course of everyday life constantly re-fashion worlds that bear their identity-enhancing beliefs and desires. Second, these worlds are “intentional” because they are directive; that is, they nudge subsequent action toward identity-enhancing ends. The intentionality of these behavioral products is an underappreciated feature of psychological experience with important implications for theorizing collective manifestations of mind. Consider the topic of memory. The bidirectional relationship between memory and identity has been an enduring theme of psychological research. Yet, most discussions have considered how individuals reconstruct autobiographical memories in ways that serve present identity concerns or how different autobiographical memories have implications for experience of personal identity (Wilson and Ross, 2003). Recent work has begun to explore this relationship at the level of collective self, investigating how different understandings of history both reflect and impact present experience of social identity (Wertsch, 2002; Liu and Hilton, 2005; Roccas et al., 2006; Sahdra and Ross, 2007; Blatz and Ross, 2009; Licata and Klein, 2010; Figueiredo et al., 2011; Gunn and Wilson, 2011). In many of these discussions, collective refers to individual experience of identity or memory when people imagine themselves in terms of a social identity category. A cultural psychology emphasis on the intentionality of everyday worlds provides a broader understanding of collective as a process that is evident not only when people imagine themselves in terms of a social identity category, but any time that they appropriate cultural tools and other culturally evolved affordances (Rogoff, 2003). Thus, the collective character of memory is not merely evident when categorizing self in terms of social identities, but more generally any time a person appropriates psychologically constituted, cultural affordances for understanding the past. This distinction becomes especially important for understanding the concept of collective self-regulation in referencing the present work. We propose that the present work is an investigation of collective self-regulation not in the sense of regulation of the collective self, but instead in the double sense of collective regulation of collective self. To illustrate, consider a teacher who commemorates BHM with mass-marketed, mainstream artifacts that—by focusing on individual achievements of black heroes without mentioning the racism that required heroic resistance—reflect and promote racism denial. Even if the teacher displays these artifacts without personal motivations to deny the extent of racism, the products she deposits nevertheless bear identity-defensive beliefs and desires of the people who produced them (see Study 2). Similarly, even if the teacher personally intends to promote awareness of racism and support for reparative policy, her actions nevertheless reconstitute ecologies filled with cultural tools that promote denial of racism and opposition to reparative policy (as in Studies 3 and 4). In other words, the motivations and intentions associated with the teacher's action are not reducible to her personal motivation to deny racism or personal intention to oppose antiracist policy; instead, the relevant motivations and intentions reside in the cultural tools for memory and identity on which the teacher draws. We refer to this process as collective self-regulation not only because the memory, motivation, and emotion refer to “collective self,” but also because the ways in which people control and direct actions of others is both (a) collaborative or distributed across multiple actors and (b) embedded within the cultural tools. The directive force of these everyday affordances need not work through the individual intentions of an individual actor; instead, regardless of one's personal motivation or intention, engagement with material traces of psychologically constituted, cultural affordances direct one's action toward identity-relevant ends. As such, the present work highlights a process of collective self-regulation that operates not as individual regulation of the collectively-identified self, but instead as intentional worlds that are effective regulatory tools regardless of whether any given individual shares the motivations, intentions, or even identification patterns of the original actors. This idea of collective self-regulation is central to both Woodson's vision for the cultural practice of BHM and mainstream appropriation of this practice. Social justice advocates propose that “recovering historical memory” (p. 31, Martín-Baró, 1994; see also Chapman-Hilliard and Adams-Bass, 2015) is essential for liberation and community healing. BHM commemorations that recognize (rather than repress) collective memories of historical injustice are cultural tools that alert people to occurrences of present injustice, provide alternative (less victim-blaming) explanations for present inequalities, motivate the desire for social change, and otherwise regulate action toward social justice ends. At the same time, the present work cautions that not all representations of BHM will contribute equally to this goal. Indeed, selective pressures in mainstream society shape reactionary development of BHM representations that serve as tools for maintenance and reproduction of racial domination. Author contributions This manuscript is based on a Ph.D. dissertation that PS submitted to the University of Kansas. PS conducted data collection and wrote the initial draft. PS and GA both contributed to study design, data analysis, and revision of subsequent drafts. 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 work benefited from feedback and support provided by dissertation committee members Christian Crandall, Randal Jelks, Ludwin Molina, and Jessica Vasquez. The work also benefited from discussions with colleagues in the Culture in Mind Research Collaboratory at Texas A&M University, the Cultural Psychology Research Group at the University of Kansas, and the Center for Advanced Study in the Behavioral Sciences at Stanford University, which provided fellowship support to GA. 1Throughout this manuscript, we refer to Black History Month displays from schools in which Black students are in the numerical majority as “Black-Majority” schools. This includes one school in which Hispanic students were in the majority, followed by Black students, and then by White students. 2For ease of communication, we focus here on the CSE identity subscale for White racial/ethnic belonging. However, we also included the private collective self-esteem subscale and a version of these CSE subscales that assessed connection to American national identity. Patterns for these subscales are similar to those that we report here. Details of analyses involving these additional dimensions of identification are available from the authors. 3The demographic questionnaire included questions about the racial/ethnic background of participants' educational experience in studies 2 and 3. We asked what % of the students at different schooling levels were White (e.g., high school; Sanders-Thompson, 1995). Supplementary analyses of these data are available from the authors. Results indicated that reported racial/ethnic background of high school was not correlated with the primary dependent variables in Study 2 (affective responses, display evaluations) or Study 3 (racism perceptions, policy support). This question was not included in Study 4. 4Five of these displays are materials that the first author obtained in Study 1 and that participants rated in Study 2. We added the sixth display (and did not use one of the seven displays from White-Majority schools in the White-Majority condition) in order to equate procedures across conditions. This sixth display came from a teacher at a predominately Black school in the same metropolitan area who sent pictures of the school's BHM display after we completed Study 1. 5As an example of the ways in which different artifacts can contain mixed messages, consider again the psychological consequences of the celebratory, achievement-oriented poster that Target department stores created and distributed for BHM (Figure 1). Presumably, the purpose of this representation is to highlight Black American achievements and thereby promote a sense of pride and collective self-efficacy. However, the poster also re-presents a sanitized version of American history that fails to emphasize racist barriers and anti-racist struggle. Moreover, the tagline—“They didn't wait for opportunity. They invented it”—carries the individualist implication that true heroes do not complain about structural barriers or rely on collective solutions; they overcome by force of individual will. Such celebratory, sanitizing narratives maintain the status quo and reproduce racial inequality by promoting strategies of individual mobility within the system at the expense of collective action to change the system (Pitre and Ray, 2002; Schick and St. Denis, 2005; Tajfel and Turner, 1986). ==== Refs References Adamczyk A. (2002 ). On Thanksgiving and collective memory: constructing the American tradition. J. Hist. Sociol. 15 , 343 –365 . 10.1111/1467-6443.00182 Adams G. Biernat M. Branscombe N. R. Crandall C. S. Wrightsman L. S. (2008 ). Beyond prejudice: toward a sociocultural psychology of racism and oppression , in Commemorating Brown: The Social Psychology of Racism and Discrimination , eds Adams G. Biernat M. Branscombe N. R. Crandall C. S. Wrightsman L. S. (Washington, DC : American Psychological Association ), 215 –246 . Adams G. Salter P. S. Pickett K. M. Kurtiş T. Phillips N. L. (2010 ). Behavior as mind-in-context: a cultural psychology analysis of “paranoid” suspicion , in The Mind in Context , eds Barrett L. F. Mesquita B. Smith E. (New York, NY : Guilford ), 277 –306 . Adams G. Stocks E. (2008 ). A cultural analysis of the experiment and an experimental analysis of culture. Soc. Psychol. Pers. Compass 2 , 1895 –1912 . 10.1111/j.1751-9004.2008.00137.x Adams G. Thomas Tormala T. O'Brien L. T. (2006 ). The effect of self-affirmation on perception of racism . J. Exp. Soc. Psychol. 42 , 616 –626 . 10.1016/j.jesp.2005.11.001 Alridge D. (2006 ). The limits of master narratives in history textbooks: an analysis of representations of Martin Luther King, Jr . Teach. Coll. Rec. 108 , 662 –686 . 10.1111/j.1467-9620.2006.00664.x Banks J. A. (1999 ). An Introduction to Multicultural Education, 2nd Edn . Boston, MA : Allyn and Bacon . Barker R. G. (1968 ). Ecological Psychology: Concepts and Methods for Studying the Environment of Human Behavior . Stanford, CA : Stanford University Press . Baumeister R. F. Hastings S. (1997 ). Distortions of collective memory: how groups flatter and deceive themselves , in Collective Memory of Political Events: Social Psychological Perspectives , eds Pennebaker J. W. Paez D. Rimé B. (Mahwah, NJ : Erlbaum ), 277 –293 . Berger P. Luckmann T. (1966/1992 ). The Social Construction of Reality. New York, NY : Anchor Books. Blatz C. Ross M. (2009 ). Historical memories , in Memory in Mind and Culture , eds Boyer P. Wertsch J. W. (New York, NY : Cambridge University Press ), 223 –237 . Blauner R. Wellman D. (1998 ). Toward the decolonization of social research , in Death of White Sociology , ed Ladner J. (Baltimore, MD : Black Classic Press ), 310 –330 . Branscombe N. R. Miron A. M. (2004 ). Interpreting the ingroup's negative actions toward another group: emotional reactions to appraised harm , in The Social Life of Emotions , eds Tiedens L. Z. Leach C. W. (New York, NY : Cambridge University Press ), 314 –335 . Brown K. D. Brown A. L. (2010 ). Silenced memories: an examination of the sociocultural knowledge on race and racial violence in official school curriculum . Equity Excell. Educ. 43 , 139 –154 . 10.1080/10665681003719590 Chapman-Hilliard C. Adams-Bass V. (2015 ). A conceptual framework for utilizing black history knowledge as a path to psychological liberation for black youth . J. Black Psychol . [Epub ahead of print]. 10.1177/0095798415597840 Cole M. (1995 ). Socio-cultural-historical psychology: some general remarks and a proposal for a new kind of cultural-genetic methodology , in Sociocultural Studies of Mind , eds Wertsch J. V. Ri'0 P. D. Alvarez A. (Cambridge : Cambridge University Press ), 113 –144 . Collins P. H. (2002 ). Black Feminist Thought: Knowledge, Consciousness, and the Politics of Empowerment . New York, NY : Routledge . Dagbovie P. G. (2004 ). Making Black history practical and popular: Carter G. Wilson, the proto Black studies movement, and the struggle for Black liberation. West. J. Black Stud. 28 , 735 –751 . Dagbovie P. G. (2005 ). Of all our studies, history is best qualified to reward our research: black history's relevance to the Hip Hop generation . J. Afr. Am. Hist. 90 , 299 –323 . Available online at: http://www.jstor.org/stable/20064002 Denzin N. K. Lincoln Y. S. (2012 ). Introduction: the discipline and practice of qualitative research , in Strategies of Qualitative Research , eds Denzin N. Lincoln Y. (Thousand Oaks, CA : Sage Publications ), 1 –42 . Dixon J. Levine M. (eds.). (2012 ). Beyond Prejudice: Extending the Social Psychology of Conflict, Inequality and Social Change . New York, NY : Cambridge University Press . Doosje B. Branscombe N. R. Spears R. Manstead A. S. R. (1998 ). Guilty by association: when one's group has a negative history . J. Pers. Soc. Psychol. 75 , 872 –886 . 10.1037/0022-3514.75.4.872 DuBois W. E. B. (1989/1903 ). The Souls of Black Folk . NewYork, NY : Bantam Books . (Original work published 1903). Epstein T. (2000 ). Adolescents' perspectives on racial diversity in U.S. history: case studies from an urban classroom . Am. Educ. Res. J. 37 , 185 –214 10.3102/00028312037001185 Eyerman R. (2004 ). The past in the present: culture and the transmission of memory. Acta Sociol. 47 , 159 –169 . 10.1177/0001699304043853 Figueiredo A. Valentim J. Doosje B. (2011 ). A shared past and a common future: the Portuguese colonial war and the dynamics of group-based guilt . Span. J. Psychol. 14 , 163 –171 . 10.5209/rev_SJOP.2011.v14.n1.14 21568174 Franklin J. H. (1997 ). Black history month: serious truth telling or a triumph in tokenism? J. Blacks High. Educ. 18 , 88 10.2307/2998774 Gibson J. J. (1977 ). The theory of affordances , in Perceiving, Acting, and Knowing: Toward an Ecological Psychology , eds Shaw R. Bransford J. (Hillsdale, NJ : Erlbaum ), 67 –82 . Gjerde P. F. (2004 ). Culture, power, and experience: toward a person-centered cultural psychology . Hum. Dev. 47 , 138 –157 . 10.1159/000077987 Gunn G. R. Wilson A. E. (2011 ). Acknowledging the skeletons in our closet: the effect of group affirmation on collective guilt, collective shame, and reparatory attitudes. Pers. Soc. Psychol. Bull. 37 , 1474 –1487 . 10.1177/0146167211413607 21734164 Hammack P. L. (2010 ). Identity as burden or benefit? Youth, historical narrative, and the legacy of political conflict . Hum. Dev. 53 , 173 –201 . 10.1159/000320045 Hammack P. L. (2011 ). Narrative and the Politics of Identity: The Cultural Psychology of Israeli and Palestinian Youth . New York, NY : Oxford University Press . Hayes A. F. Preacher K. J. (2014 ). Statistical mediation analysis with a multicategorical independent variable . Br. J. Math. Stat. Psychol. 67 , 451 –470 . 10.1111/bmsp.12028 24188158 Heine S. J. Lehman D. R. (2004 ). Move the body, change the self: acculturative effects on the self-concept , in Psychological Foundations of Culture , eds Schaller M. Crandall C. (Mahwah, NJ : Erlbaum ), 305 –331 . Hughes D. Rodriguez J. Smith E. P. Johnson D. J. Stevenson H. C. Spicer P. (2006 ). Parents' ethnic-racial socialization practices: a review of research and directions for future study . Dev. Psychol. 42 :747 . 10.1037/0012-1649.42.5.747 16953684 Jost J. T. Banaji M. R. Nosek B. A. (2004 ). A decade of system justification theory: accumulated evidence of conscious and unconscious bolstering of the status quo . Polit. Psychol. 25 , 881 –919 . 10.1111/j.1467-9221.2004.00402.x Journell W. (2008 ). When oppression and liberation are the only choices: the representation of African Americans within state social studies standards . J. Soc. Stud. Res. 32 , 40 . Kim H. Markus H. R. (1999 ). Deviance or uniqueness, harmony or conformity? A cultural analysis . J. Pers. Soc. Psychol. 77 :785 . 10.1037/0022-3514.77.4.785 24113521 Kitayama S. Markus H. R. (1999 ). Yin and yang of the Japanese self: the cultural psychology of personality coherence , in The Coherence of Personality: Social Cognitive Bases of Personality Consistency, Variability, and Organization , eds Cervone D. Shoda Y. (New York, NY : Guilford Press ), 242 –302 . Kitayama S. Mesquita B. Karasawa M. (2006 ). Cultural affordances and emotional experience: socially engaging and disengaging emotions in Japan and the United States . J. Pers. Soc. Psychol. 91 :890 . 10.1037/0022-3514.91.5.890 17059308 Kroeber A. L. Kluckhohn C. K. (1952 ). Culture: A Critical Review of Concepts and Definitions . New York, NY : Random House . Kurtiş T. Adams G. Yellowbird M. (2010 ). Generosity or genocide? identity implications of silence in American thanksgiving commemorations . Memory 18 , 208 –224 . 10.1080/09658210903176478 19693724 Leach C. W. (2002 ). The social psychology of racism reconsidered . Feminism Psychol. 12 , 439 –444 . 10.1177/0959353502012004005 Levin S. Sidanius J. Rabinowitz J. L. Federico C. (1998 ). Ethnic identity, legitimizing ideologies, and social status: a matter of ideological asymmetry. Polit. Psychol. 19 , 373 –404 . 10.1111/0162-895X.00109 Licata L. Klein O. (2010 ). Holocaust or benevolent paternalism? Intergenerational comparisons on collective memories and emotions about Belgium's colonial past . Int. J. Conf. Violence 4 , 46 . Retrieved from: http://www.ijcv.org/index.php/ijcv/article/view/60/186 27011787 Liu J. H. Hilton D. J. (2005 ). How the past weighs on the present: social representations of history and their role in identity politics . Br. J. Soc. Psychol. 44 , 537 –556 . 10.1348/014466605X27162 16368018 Loewen J. W. (1995 ). Lies My Teacher Told Me: Everything Your American History Textbook Got Wrong . New York, NY : Touchstone . Luhtanen R. Crocker J. (1992 ). A collective self-esteem scale: self-evaluation of one's social identity . Pers. Soc. Psychol. Bull. 18 , 302 –318 . 10.1177/0146167292183006 Marecek J. Fine M. Kidder L. (1997 ). Working between worlds: qualitative methods and social psychology. J. Soc. Issues 53 , 631 –644 . 10.1111/j.1540-4560.1997.tb02452.x Markus H. R. Hamedani M. G. (2007 ). Sociocultural psychology: The dynamic interdependence among self-systems and social systems , in Handbook of Cultural Psychology , eds Kitayama S. Cohen D. (New York, NY : Guilford ), 3 –46 . Markus H. R. Kitayama S. (1991 ). Culture and the self: implications for cognition, emotion, and motivation . Psychol. Rev. 98 :224 10.1037/0033-295X.98.2.224 Markus H. R. Uchida Y. Omoregie H. Townsend S. S. Kitayama S. (2006 ). Going for the gold models of agency in Japanese and American contexts . Psychol. Sci. 17 , 103 –112 . 10.1111/j.1467-9280.2006.01672.x 16466417 Martín-Baró I. (1994 ). Writings for a Liberation Psychology . Cambridge, MA : Harvard University Press . McIntyre A. Lyons A. Clark A. Kashima Y. (2004 ). The microgenesis of culture: serial reproduction as an experimental simulation of cultural dynamics , in The Psychological Foundations of Culture , eds Schaller M. Crandall C. S. (Mahwah, NJ : Lawrence Erlbaum Associates ), 227 –258 . Mendoza-Denton R. Downey G. Purdie V. J. Davis A. Pietrzak J. (2002 ). Sensitivity to status-based rejection: implications for African American students' college experience. J. Pers. Soc. Psychol. 83 , 896 –918 . 10.1037/0022-3514.83.4.896 12374443 Miyamoto Y. Nisbett R. E. Masuda T. (2006 ). Culture and the physical environment: holistic versus analytic perceptual affordances. Psychol. Sci. 17 , 113 –119 . 10.1111/j.1467-9280.2006.01673.x 16466418 Morling B. Lamoreaux M. (2008 ). Measuring culture outside the head: a meta-analysis of individualism-collectivism in cultural products . Pers. Soc. Psychol. Bull. 12 , 199 –221 . 10.1177/1088868308318260 18544712 Morton T. A. Sonnenberg S. J. (2011 ). When history constrains identity: expressing the self to others against the backdrop of a problematic past. Eur. J. Soc. Psychol. 41 , 232 10.1002/ejsp.770 Moscovici S. (1984 ). The phenomena of social representations , in Social Representations , eds Farr R. M. Moscovici S. (Cambridge : Cambridge University Press ), 3 –69 . Nelson J. C. Adams G. Salter P. S. (2013 ). The Marley hypothesis: racism denial reflects ignorance of history. Psychol. Sci. 24 , 213 –218 . 10.1177/0956797612451466 23232861 Oishi S. Graham J. (2010 ). Social ecology lost and found in psychological science . Perspect. Psychol. Sci. 5 , 356 –377 . 10.1177/1745691610374588 26162183 Okazaki S. David E. J. R. Abelmann N. (2008 ). Colonialism and psychology of culture . Soc. Personal. Psychol. Compass 2 , 90 –106 . 10.1111/j.1751-9004.2007.00046.x Paluck E. L. (2009 ). Reducing intergroup prejudice and conflict using the media: a field experiment in Rwanda . J. Pers. Soc. Psychol. 96 , 574 –587 . 10.1037/a0011989 19254104 Pennebaker J. W. Banasik B. L. (1997 ). On the creation and maintenance of collective memories: history as social psychology , in Collective Memory of Political Events: Social Psychological Perspectives , eds Pennebaker J. W. Paez D. Rime B. (Mahwah, NJ : Erlbaum ), 3 –19 . Pennebaker J. W. Gonzales A. L. (2009 ). Making history: social and psychological processes underlying collective memory , in Memory in Mind and Culture , eds Boyer P. Wertsch J. W. (New York, NY : Cambridge University Press ), 171 –193 . Persinger H. (2011 ). ‘Black History’ Beer Poster Makes a Stir: Budweiser Sponsored the Commemoration of Kings and Queens of Africa . San Diego Union-Tribune [Electronic Version]. Available online at: http://www.utsandiego.com/news/2011/feb/16/black-history-beer-poster-makes-a-stir (Accessed February 15, 2013). Pinel E. C. (1999 ). Stigma consciousness: the psychological legacy of social stereotypes . J. Pers. Soc. Psychol. 76 , 114 –128 . 10.1037/0022-3514.76.1.114 9972557 Pitre A. Ray R. (2002 ). The controversy around Black history . West. J. Black Stud. 26 , 149 –154 . Reicher S. Hopkins N. (2001 ). Self and Nation . London : Sage . Roccas S. Klar Y. Liviatan I. (2006 ). The paradox of group-based guilt: modes of national identification, conflict vehemence, and reactions to the in-group's moral violations. J. Pers. Soc. Psychol. 91 , 698 –711 . 10.1037/0022-3514.91.4.698 17014294 Rogoff B. (2003 ). The Cultural Nature of Human Development . New York, NY : Oxford University Press . Sahdra B. Ross M. (2007 ). Group identification and historical memory . Pers. Soc. Psychol. Bull. 33 , 384 –395 . 10.1177/0146167206296103 17312319 Sanders-Thompson V. L. (1995 ). Sociocultural influences on African-American racial identification. J. Appl. Soc. Psychol. 25 , 1411 –1429 . 10.1111/j.1559-1816.1995.tb02624.x Savani K. Markus H. R. Conner A. L. (2008 ). Let your preference be your guide? Preferences and choices are more tightly linked for North Americans than for Indians . J. Pers. Soc. Psychol. 95 , 861 . 10.1037/a0011618 18808264 Schick C. St. Denis V. (2005 ). Troubling national discourses in anti-racist curricular planning . Can. J. Educ. 28 , 295 –317 . 10.2307/4126472 Shweder R. A. (1990 ). Cultural psychology: what is it? , in Cultural Psychology: Essays on Comparative Human Development , eds Stigler J. Shweder R. Herdt G. (Cambridge, UK : Cambridge University Press ), 1 –46 . Snibbe A. C. Markus H. R. (2005 ). You can't always get what you want: educational attainment, agency, and choice . J. Pers. Soc. Psychol. 88 , 703 –720 . 10.1037/0022-3514.88.4.703 15796669 Tajfel H. Turner J. C. (1986 ). The social identity theory of intergroup behaviour , in Psychology of Intergroup Relations , eds Worchel S. Austin W. G. (Chicago, IL : Nelson-Hall ), 7 –24 . Tatum B. D. (1987 ). Assimilation Blues: Black Families in a White Community . Westport, CT : Greenwood Press . Tatum B. D. (1997 ). “Why are All the Black Kids Sitting Together in the Cafeteria?” and Other Conversations about Race . New York, NY : Basic Books . Thornhill T. E. (2014 ). Resistance and assent how racial socialization shapes black students' experience learning African American history in high school . Urban Educ . [Epub ahead of print]. 10.1177/0042085914566094 Trawalter S. Driskell S. Davidson M. N. (2015 ). What is good isn't always fair: on the Unintended effects of framing diversity as good . Anal. Soc. Issues Public Policy . [Epub ahead of print]. 10.1111/asap.12103 Trouillot M. R. (1995 ). Silencing the Past: Power and the Production of History . Boston, MA : Beacon Press . Uskul A. K. Kitayama S. Nisbett R. E. (2008 ). Ecocultural basis of cognition: farmers and fishermen are more holistic than herders . Proc. Natl. Acad. Sci. U.S.A. 105 , 8552 –8556 . 10.1073/pnas.0803874105 18552175 Vygotsky L. S. (1978 ). Mind in Society: The Development of Higher Psychological Processes . Cambridge, MA : Harvard University Press . Watson D. Clark L. A. Tellegen A. (1988 ). Development and validation of brief measures of positive and negative affect: the PANAS scales. J. Pers. Soc. Psychol. 54 , 1063 –1070 . 10.1037/0022-3514.54.6.1063 3397865 Wertsch J. V. (2002 ). Voices of Collective Remembering . Cambridge : Cambridge University Press . Wertz F. J. Charmaz K. McMullen L. Josselson R. Anderson R. McSpadden E. (2011 ). Five Ways of Doing Qualitative Analysis: Phenomenological Psychology, Grounded Theory, Discourse Analysis, Narrative Research, and Intuitive Inquiry . New York, NY : Guilford Press . Wetherell M. (2011 ). The prejudice problematic , in Beyond Prejudice: Extending the Social Psychology of Conflict, Inequality and Social Change , eds Dixon J. Levine M. (Cambridge : Cambridge University Press ), 158 –178 . Wetherell M. Potter J. (1992 ). Mapping the Language of Racism: Discourse and the Legitimation of Exploitation . New York, NY : Columbia University Press . Wilson A. E. Ross M. (2003 ). The identity function of autobiographical memory: time is on our side . Memory 11 , 137 –149 . 10.1080/741938210 12820827 Wohl M. J. A. Branscombe N. R. Klar Y. (2006 ). Collective guilt: justice-based emotional reactions when one's group has done wrong or been wronged . Eur. Rev. Soc. Psychol. 17 , 1 –37 . 10.1080/10463280600574815 24993860 Wright S. C. Lubensky M. (2009 ). The struggle for social equality: collective action versus prejudice reduction , in Intergroup Misunderstandings: Impact of Divergent Social Realities , eds Demoulin S. Leyens J. P. Dovidio J. F. (New York, NY : Psychology Press ), 291 –310 . Zajonc R. B. (1980 ). Feeling and thinking: preferences need no inferences . Am. Psychol. 35 , 151 –175 . 10.1037/0003-066X.35.2.151

PMC005xxxxxx/PMC5003090.txt

==== Front J Vis Art PractJ Vis Art PractRJVPrjvp20Journal of Visual Art Practice1470-20291758-9185Routledge 104174310.1080/14702029.2015.1041743ArticlesTalking about a Christine Borland sculpture: effective empathy in contemporary anatomy art (and an emerging counterpart in medical training?) Richardson Craig a * Borland Christine b a The College of Art, Design and the Built Environment, Nottingham Trent University, Burton Street, NottinghamNG1 4BU, UKb Department of Arts, The University of Northumbria, BALTIC 39, 31–39 High Bridge, Newcastle upon TyneNE1 1EW, UK* Corresponding author. Current address: University of Huddersfield, Huddersfield, UK. Email: C.Richardson2@hud.ac.uk4 5 2015 24 7 2015 14 2 146 161 © 2015 The Author(s). Published by Taylor & Francis.2015The Author(s)This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.This Introduction and interview discusses the poetical and empathic insights that are a key to the effectiveness of contemporary artist Christine Borland's practice and its relevance to the medical humanities, visual art research and medical students’ training. It takes place in a context of intensive interest in reciprocity and conversation as well as expert exchange between the fields of Medicine and Contemporary Arts. The interview develops an understanding of medical research and the application of its historical resources and contemporary practice-based research in contemporary art gallery exhibitions. Artists tend not to follow prescriptive programmes towards new historical knowledge, however, a desire to form productive relationships between history and contemporary art practice does reveal practical advantages. Borland's research also includes investigations in anatomy, medical practices and conservation. Keywords anatomymedicine and contemporary artsculpturemuseumWellcome Trust10.13039/100004440093571/Z/10/ZThis work was supported by the Wellcome Trust [grant number 093571/Z/10/Z]. ==== Body Introduction: Craig Richardson Christine Borland's response to a mid-Victorian plaster cast, to which she devoted her attention between late 2010 and early 2012, can be contextualised through an incident in W.G. Sebald's meditation on transience The Rings of Saturn (1995). Sebald speculated the lesson represented underway in Rembrandt's Anatomy Lesson (1632) (Figure 1) had a punitive dimension and he highlighted how the inversion of the anatomical – one hand is pictured the wrong way round – at the painting's centre was intended to draw attention to the violence that has been doubly performed upon Kindt's executed body. It was:with him, the victim, and not the Guild [of surgeons] that gave Rembrandt his commission, that the painter identifies. He alone sees that greenish annihilated body, and he alone sees the shadow in the half-open mouth and over the dead man's eyes. (1995 [English translation, 1998], 17) Rembrandt's empathy is a reminder of the vital insights artists may bring to historical medical practices and their contemporary counterpart. Borland's encounter with the norms of specialised medical research often means a specifically empathic insight can be relocated into the space between their norms and then their historical or public presentation in galleries and Museums but also to feed back into contemporary clinical settings and training. In Borland's practice – to date – there is no direct engagement with patients in a clinical setting and the ethical frameworks for treatment would be prohibitive, however, her contingent dialogue with those in the medical and Museum professions means she often reaches an implicit point of convergence. Acting at one remove and as a generalist, her focus is on material culture and making a differently activated environment through a performative dimension; the culmination of her investigations (a more appropriate terms than research) involves new visualisations of old and new medical practices, while engaging differently and unconventionally with the technical apparatus and processes and/or overlooked ephemera found in these settings. In subsequent exhibitions the sometimes overlooked, incidental, ephemeral and technical elements are extracted from their context; such inversions and elisions enable another type of ‘knowing’. (Kohn 2011, 3–4). 1 Her ‘conversational expert dialogue’, an informal method deployed by Borland during her initial investigations in medical environments, sometimes entails emergent surprises and discoveries that are often re-evaluated and synthesised for exhibition purposes. Borland produces art-works in which medical methodologies are not necessarily included within their exhibition, although such extrapolations are sometimes disseminated in accompanying catalogues and educational publications. Figure 1. Rembrandt van Rijn, The Anatomy Lesson of Dr. Nicolaes Tulp, 1632, (detail). Collection of Royal Picture Gallery Mauritshuis, The Hague. The interview's focal points are the decisions Borland made during a Production Residency and Exhibition at Glasgow Sculpture Studios in 2010–2011. We note how Borland's programme of plaster cast research was complemented by her previous exhibition SimBodies & NoBodies (2009) (Figures 3 and 4) at The Ormeau Baths Gallery, Belfast, which included sculptures derived from mannequins normally deployed in medical teaching to assist demonstrations of the Heimlich manoeuvre or contemporary prosthetic mannequins such as ‘Choking Charlie’. Figure 2. Christine Borland, Film Still – NoBodies; Cast From Nature, 2011, 12minute 30 second loop, H.D. Video, projected. (Made possible by The Morton Award for lens based media, 2010) Considering both the 2011 plaster work and earlier mannequin projects we ask how an artist can raise questions and explore the ways in which the subject of scientific study can evoke empathy, through artistic interpretation of medical apparatus and anatomical representation. Such questions are discussed as creative and pedagogical issues. Her research for this stemmed from an initial encounter with a fibreglass cast permanently displayed in the Royal College of Surgeons (RCS) in Edinburgh. RCS Surgeons’ Hall Museum for the History of Surgery attributes the fibreglass as Cast From Nature and dated 1845, by [Sir] John Goodsir (1814–1867), and it is likely the fibreglass cast was taken from a plaster cast original in the nearby Anatomical Museum in Edinburgh University. There the body of a partially flayed man is posed after Michelangelo's Vatican Pièta (1499). The Glasgow Production Residency phase later culminated in the exhibition Cast From Nature at Glasgow Sculpture Studio a few days after this interview, and which restaged the recovery and representation of the plaster cast found in the basement of the Edinburgh Anatomical Museum. The installation was further synthesised at a second venue, Camden Arts Centre, May–July 2011, and elements have continued to appear in exhibitions at Edinburgh, 2 Orkney and Liverpool (Figure 2). Figure 3. Christine Borland, SimBodies & Nobodies, 2009, Detail, Medical Manikins. From the beginning of the Glasgow Sculpture Studios exhibition (late 2010) to the exhibition at Camden Arts Centre in May 2011 Borland seemed perplexed by her fascination with the plaster cast and sought to apply this in exhibition arrangements that emphasised its metaphysical value. Rather than providing anecdotal cover, the historical context is also fructifying, or at least the relationship between art and history is consolidated in the manner by which exhibition seemed to contain elements of conservation practice. Historical understanding has channelled Borland's empathic response but has not delimited its representation through, for instance, explanatory text panels. 3 The materiality of plaster was very much in evidence in both exhibitions. 4 Figure 4. Christine Borland, SimBodies & NoBodies, 2009, Detail Plaster & Silicon Moulds. Figure 5. Christine Borland, Cast From Nature, Glasgow Sculpture Studios, 2010, Casting in Progress. In Camden two plaster cast figures were separated by a long-draped plaster-soaked cloth diagonally traversing from opposite walls of the gallery that tended to emphasise horrific aspects of the two figures within a dramatic tableau (Figure 6). Once more the exhibition of these casts was an arena for subjective retrieval (a subjectivity Borland similarly describes in the progressive training of medical students studying anatomy in New Zealand). One of the casts, held in the classical Pieta recline but fixed upon a steel-rod rack, was positioned slightly higher than table height, coincidently Dr Tulp's pose when stationed in front of Rembrandt's famous group portrait. This steel-rod rack kept the cast figure's head in a painfully unsupported distortion with the throat pushed outwards, surely a choking position were the figure alive. Contradictions were evident. This Pieta derived plaster cast's eyes face the heavens, the pelvis clothed, the body given a semi-dignified appearance. Nevertheless the image is a body in an arc of pain. On the other side of the long brittle drape the same body cast was then displayed 180° facing downwards, the painful throat distension was now an unremarkable effect and the naturally dropping mouth was now seemingly alive and even responsive. The visage was placed at standing face height, the figure was no longer reclining, its flayed and drooping flesh now resembling a fast falling life rushing mercilessly downward towards oblivion. While this may have conjured the horror-fall of aerial victims the cast eyes no longer looked to the heavens but stared ahead with a languid but superior menace. Two figures, one classically reposed after death in contrast with the other, caught in the fast race towards death set in a room with an almost imperceptible even settling of fine chalk dust. Plaster is heavier than other dust motes. Figure 6. Christine Borland, Cast From Nature, 2011, Camden Arts Centre. (Photo: Andy Keate) Figure 7. Basement, Edinburgh Anatomical Museum Stores. Figure 8. Christine Borland, SimWoman, 2010, Film Still, 12minute 30 second loop, H.D. Video, projected. Figure 9. Christine Borland, NoBodies, (Detail) Choking Charlie, 2009, Film Still, 10 minute loop, monitor. Christine Borland interview with Craig Richardson, Glasgow Sculpture Studios, 19 November 2010 Borland: This plaster cast is from a fibreglass mould taken from an original plaster cast. The original was lost; the fibreglass started me on the quest to find the original, which I found eventually. I tracked it down from hearsay. It had been spotted in a museum store in the Anatomy building of Edinburgh Medical School with its extremities broken off. Looking at it now, I’m not going to put it back together permanently but to make the new cast I’ll make a partial repair, otherwise it's a significant conservation job. The Edinburgh University Anatomical Museum used to be over a number of floors of the Medical building. Two magnificent Asian elephant skeletons remain placed at the entrance. Through the years the Anatomical Museum was squeezed to make more space for teaching and by the 1980s many things were taken off display and hidden away in vaults, for example, there is a huge phrenology collection including skulls of indigenous peoples, some of which have since been repatriated. There is an attic space, which still has William Burke's (of Burke and Hare) skeleton and a life mask of William Hare. But the store is not a systematic academic presentation, there are random artefacts everywhere, crammed into a basement in a way that can’t help but resemble a gothic horror. It's certainly not set up as a teaching resource, unlike the Hunterian Museums in Glasgow or London.Richardson: With nothing catalogued and no proper documentation, how did you set about finding this lost plaster cast? B: Via the curator of the Anatomical Museum. Dr Gordon Findlater, Director of the Anatomy Department at Edinburgh, is a proper enthusiast and is completely distraught that all these damaged things often lack of any proper documentation. It was in a corridor alongside other neglected and damaged objects. R: Assuming this is then a relatively recent plaster cast does it have a named maker? Similarly had the more contemporary fibreglass, 5 or even the original plaster if the one you located is not the true original? B: Yes, and no. The fibreglass cast of the same figure in the centre of the RCS Edinburgh is labelled ‘a cast from nature, 1845, by John Goodsir’. Goodsir is credited with post-Burke and Hare restoration of anatomy's reputation in Edinburgh following the 1832 Anatomy Act. He was also renowned for his interest in art, music and for his humanity. But still there are no details on anything to do with an artist or at least there's no proper documentation as far as I know. Of the figure the Medical Historian Ruth Richardson has suggested that, although the dates don’t match exactly, that the figure could be a cast of a dissection mentioned in Goodsir's memoirs ‘an Edinburgh carter of intensely whisky habits, who in a drunken state fell from his cart and died on the spot [and] remained free from decomposition during thirty days.’ (2011, 2073–2074) R: Let's talk about the recovered cast and the bio-information you notice. B: When you consider the upper chest, neck – are these folds of skin or drapes of cloth? The cloth around the head could cover a major head injury I suppose, supporting Ruth Richardson's idea. It certainly seems to originate from a direct cast of a dissection. This is cloth draped over the groin area, I think by the way it is cracked that it is an actual piece of cloth, rather than something that has been moulded. And looking at the feet (it was easier to see when they were still broken off the main plaster section and detached) because of my little forensic knowledge I know that although they are now suspended in repose, that after death they were resting on a hard surface and that is commensurate with a conventional dissection with the back of the soles resting on the table. In other words evidence of resultant hypostasis and the indelible imprint of what they were resting on. Therefore the posing must have happened post-dissection, possibly in preparation for casting. Figure 10. Christine Borland, Cast From Nature, Glasgow Sculpture Studios, 2010. (Photo: Ruth Clark) R: Looking for further evidence of post-dissection intervention, perhaps towards the creation of an artwork, would the person posing the body have followed instructions? B: The usefulness of this cast as a teaching tool would always have been doubtful, in fact the detail in the dissected parts is not that good and frankly how useful is that type of modesty (pointing at the figure cast section of draped cloth) in a medical context? The pose is in a distinct classical tradition, Michelangelo's Pieta, so its sculptural qualities matter more than its medical education potential.As the posing follows classical traditions there is an obvious comparison with the recent project by Joan Smith, who teaches anatomical drawing at Edinburgh College of Art, who worked with an anthropologist Jeanne Cannizzo to study a decrepit plaster cast ‘écorché’ (trans: a flayed figure) in the Edinburgh College collection (and since badly stored in a cupboard) which is posed after the Dying Gaul (unknown artist). That cast's nickname ‘Smugglerius’, in mock Latin, noted the criminal past and the classical pose of the originating figure. Smith and Cannizzo found out that it was likely to be James Langar, a convicted highway robber, taken from London’s Newgate prison, hanged at Tyburn in 1776 and then dissected the next day by William Hunter (1718–1783), of the brothers John and William Hunter, both famous anatomists. William Hunter moved back to Glasgow, ergo the Hunterian Collection. Figure 11. Christine Borland, Cast From Nature, Glasgow Sculpture Studios, 2010, Casting in Progress. R: And have you tried to find out who made the original cast of this Edinburgh carter, or this plaster copy for that matter, or does that matter? B: I’ve tried to find out if the information is readily available, I’ve looked through available records and have asked conservators and curators, nobody in the RCS is able to tell me. The fibreglass cast of this, which is on permanent display, is credited to Goodsir and a particular date but then Goodsir's life predates fibreglass, so the plaque may refer to the plaster original. But then there's a little cardboard label saying this fibreglass cast of the object was made possible by a grant from the Museum of Scotland, so I went to them to ask if they had records of the grant and they had the name of a sculptor who had made the fibreglass. Originally I was ambitious enough to ask ‘do you have the moulds’ that must have been made to produce the fibreglass cast. R: Okay. So would this kind of casting have any medical purpose? B: Well that's interesting, what I was describing there was a fibreglass cast that was made in the 1986. Nobody could tell me why it was made. Perhaps the original plaster was in a poor condition. There must have been a plaster original that the fibreglass cast was made from but the one I’m using here isn’t it because when you make a fibreglass you always leave traces on the original, so this definitely isn’t the one that was used because of the lack of fibreglass on any of its surfaces. This is not the source of the 1986 fibreglass on display. Incidentally, with fibreglass casting the more casts you make from the original the worse quality it is. The figure in the RCS is shrunken in scale because every time you cast you take away some of the surface, really a large percentage from the overall area. The next generation of cast is 12% smaller. But as for any medical purpose, I can only imagine the making of the original plaster cast was for artistic reasons. I don’t know if an artist or technician would have made it, you would need some sort of historical knowledge of the material that I don’t have. Gelatine might have been used for the inner mould. When I make a cast of this the substance I will apply is a silicone rubber, a rubber mould captures all the detail, then it is wrapped in a jacket of plaster or some such material as a rigid case for the rubber that is doing the work, then the plaster will be poured in to the rubber mould. When the original cast was made of the dissected body, rubber was not available in 1845 so I can only speculate that plaster was directly applied and that's a horrible image to me because I know what you have to do to physically remove it. R: So the process then had a salacious aspect, even if used here for more elevated purposes? B: The overall pose suggests the decision was taken to not just to cast the dissected area as a record, there is no dissection of the legs and arms and relatively little of the face; the actual proportion of the body that is dissected is relatively small. R: Thinking of other instances of artists attending dissections and invited to make something visual, perhaps this posing and original cast is such an example? In the absence of relevant historical information, your gut instinct will provide us with the best speculative insight. So, speculatively speaking, the posing of the body to resemble a Pieta, is this an artistic intervention? B: No, not in the sense we mean it now. Even if an artist had made the cast I would say they would be acting under instruction of the anatomist, I very much doubt the artist would be making the decision about the pose, I would really feel the artist in this case – if indeed it was an artist or more likely a technician – if the artist was asked and commissioned to make the pose, they would have been asked to make use of many references, in this case referring to the Pieta. From other things I have looked at in the period the ‘voice’ of the artist or an independent aesthetic decision-making of an artist really is not discussed – the person who is taking those decisions is the anatomist, such as Hunter. It is then credited as the work of the anatomist, no artist mentioned there and any aesthetic decision-making is an expression of the medical professional. R: So accepting some of your points about an obscure purpose in an anatomical-training setting, why make the original cast? B: As a memento of the skill of the anatomist? Which is rather dangerous territory. Or to mark the anatomist's belief the elevation of the status of the body. R: Could the cast have consciously been made as an example of technical display rather than artistry? B: It seems unlikely, though that might have been a factor if there was a destination in mind for the cast – maybe as a centre piece for a museum, as it is today in RCS Surgeon's Hall. It could have been for teaching purposes, that is possible, but who were the students? Sculpture Students from the Art School? There certainly doesn’t seem to be any record of the artists/sculptors or technicians involved. At the same time, I just feel that, speaking with anatomists, there is little practical learning to be gained from this cast; it is too removed from a scientific context. This cast is a kind of elevation, away from the obvious ethics associated with this science, the immediate reference being more to a Pieta – like a Dead Christ figure. R: Would an anatomist at this time be engaged with something other than seeking to understand bodily functions? B: These contradictions are built into anatomical dissection, what was happening was actually quite horrific. Before 1832 the bodies came directly from the gallows and dissection was a public spectacle. In the earliest Anatomy School Fabricus, Vesalius (1514–1564) introduced elements to the anatomy theatre like engravings and inscriptions such as ‘The dead teach the living’. That kind of attempt to elevate through art includes the early anatomical texts accompanying the flayed dissected bodies, posed with various references to classical art, in Albinus, Vesalius, etc., but also building towards a representation of an interior mapping of the body. R: In other recent works you’ve been looking at the technological modern version of simulated patients. You consider these do not encourage an empathic response, and the medical students have tended to treat the simulated figures as caricatures. But there seems to be, in this earlier cast in front of us, an attempt by the profession to encourage an empathic response. Do you think that is actually more conducive to the encouragement of empathy than the how you’ve witnessed the application of modern technology? B: Yes, is the short answer. It's what I like to explore in discussions with contemporary medical and anatomy students and teachers. I suppose this cast directs me to wonder whether the actual physical process of dissection can be explored further through an emotional context, a context often completely absent from new teaching methods using contemporary simulated patients. As far as I can tell, in the world of anatomy education the students often experience extreme emotional reactions to dissection, which are not explored or even considered a valuable learning experience. Sometimes students are told to just put on a brave face – just get on with it! Teenagers confronting a dead body for the first time, those anonymous bodies donated to science. But by taking away this emotional engagement and replacing this with over-dependence on mannequins and simulated patients, a lot seems to be getting lost, ultimately a loss for future patients.Glasgow University still values anatomical dissection, their students work with both plastinated specimens (the procedure patented by Gunther von Hagens) and preserved cadavers. I’ve been in and out of the labs there for more than 20 years. A really gifted teacher and anatomist Dr John Shaw-Dunn, and Dr Quentin Fogg, have been exploring how the Humanities and Fine Art in particular can be brought into day-to-day teaching. Fogg put me on to a documentary made in a medical school in New Zealand, with students dissecting the same cadaver over a period of two years. The students will strip it down and empty it out over the course of the years. It is very, very intense. The medical students are followed on the journey and their emotions are explored, how they form a narrative of the person and their life. We also see them as they attend the memorial service for the donor and meet the families. A prior recording was made in which those who were planning to donate their bodies, the bodies the students have dissected, were interviewed about their motivation and their decision. Some were terminally ill patients or simply old and they nicely described what they thought of their own body, they said things like ‘I think my heart is really strong’, ‘my lungs will be shot with the smoking’. Obviously they had been prompted to think of themselves, as they will be shortly, as cadavers. After two years, at the conclusion of the longer-term dissection and having learned a great deal, with some forming their own narratives of the dissected person from their own personal perspective, the students were given an opportunity to view the recorded interview with the donors. Some of them took up the opportunity. R: In earlier works you’ve been through a similar process, having gone through the ephemera of a forensically recovered body, but declined to include the background, contact, relationships or biographical information. B: I made a decision to stop, it is enough to point out the context includes an historical context but not to bring it down to the nuts and bolts of an investigation. Viewers can fill it in for themselves. It's the opposite of what those Medical students did when they chose to watch the film interviews of the body donors. That level of detail isn’t helpful for an artwork. R: I suppose we are talking here about posthumous but visual genealogies again, the visual information gained from the dead, but then let's say this is my ancestor's cast here, I’d be keen to find out, from a genealogical fascination. B: you could do that, but that's not my way! I’m bringing this into the public realm, I’m taking a cast of someone which hasn’t been looked at for decades, who has been climbed all over, and broken, and now brought out of the basement. If anyone wants to do that type of genealogical research I would be really delighted, but I don’t want to do that. What I found interesting about the New Zealand medical students were not those who choose to hear the stories but those who choose not to hear the stories on the prior recording, they had found a means to come to terms with the experience of the donated body through their own engagement and by their own stories. That's what I found most interesting, the comparison of the students’ narratives with the real narratives. These students managed to find a way to come to terms with it in a really positive way by being given a forum to talk about feelings that didn’t ‘fit’ in a medical context. When they had to remove the heart, which in scientific terms is just another piece of the puzzle of the body, to many of them that had a symbolic significance and meaning, which in this medical context was difficult to talk about. R: Do they talk like that? Do they talk from that ethical position? B: Yes, absolutely. I don’t know if it's that's the purpose or a by-product of this discursive film-making process in New Zealand but that's what happens. The students value it and speak in terms of it having been the most important two years of their lives. The ones planning to become surgeons, who might end up spending a lot of time operating on people who come into Accident and Emergency wards, they’ll never meet them until they’re on the operating table. They talked about how this experience would help with that moment, having had such an intimate engagement and over those two years to build that ethical code for themselves from their own perspective. R: This all seems far removed from this idealistic Pieta cast in front of us now, this cast doesn’t seem then to invite opening up any exploratory biographical channel to the students. B: No, it has been completely removed from any sense of a day-to-day, nineteenth century existence. It might be doing that top-line job that you sort of introduced earlier on, ‘isn’t anatomy crossing the line, towards an artistic form’. Hunter and other anatomists think of themselves as artists and performers, ‘it's an artform’ that's said all the time. So it's a form of expression of that … R: do you think it's an artform? B: Emm, maybe unfair to ask me, I’m too involved at the moment. There's been lots of peripheral research that has helped me think about it and make decisions. The reference to Pieta … the only one I could have direct access to is one in the Edinburgh College of Art cast collection, only the dead ‘Christ’ now has the figure of Mary gone, excised. You are left with a figure that looks a lot like this cast here. There are still lots of folds of drapery, which we know from the original is Mary's skirt. If you look carefully, where this phony rock is, there is a hand but no Mary. So the support is obvious, if you know what you are looking for, under the voluminous drapery. But She's absent. But I had a feeling of that whenever I saw this cast, I thought about the suggested support right away, the physical bonding. It's been mocked up here, the roughly emulated rocks which hold him in position now. But someone, maybe many people, would have lifted this guy, placed him in position to cast. When I make the cast I won’t include such fake supports I’ll just have him free floating, and once I have the new cast I want to work with that, to make a new extended support structure, but no plinths or rocks. R: When you devise a new support as an extension of the sculpture, what do you think you will create as a support? B: There are many possibilities. One of the things I want to do with the new cast is use a technique resulting from an accidental part of the casting process – the films I made of the plaster cast heads shown in Belfast of simulated patients (the exhibition SimBodies & NoBodies). Once I made the plaster casts I took them out of their mould while they were steaming hot, I wasn’t sure where it was going at that point. When I put them under glass the condensation obscured them but over time it started to clear and their features became visible (Figure 9). I may do something like that; the cast will be put in a Museum case, it will create a big smokescreen obscuring the object, but the cast will gradually reappear. I’d like to try that.(We move round to the underneath of the cast where the neck and base meet) If you look inside here this is the hole which if you look inside with a torch you can see the fingerprints of the person who made the cast. This is the reverse side, all their finger marks of how they pushed the plaster into the mould is there. I was amazed when I saw the Edinburgh Pieta because you can get round to look at it from the back. It's propped up with a series of iron rods which you’d expect but when you look closely they are just branches, just stuck in. It has an incredible violence and traces of hand-marks particularly when you consider it's the space behind Mary's legs and that's the hand of the sculptor. When I do this cast what we have on the outside will be a replica of this sculpture but inside will record my hands and my touch. I’m not replicating the existing support so when you look up from underneath it will be hollow, you can clearly see this evidence of my hand. R: Returning to the contemporary mannequins, ‘Choking Charlie’ and so on, is there any linkage in thinking about the inside of this plaster cast? B: Only some of the mannequins are cast from life, transformed into flesh-tone latex and rubber, then bits of mechanics and computing technology are inserted. There is a rather lovely and widely used mannequin who is a living Japanese basketball player, he is a heart-monitor simulation figure, his figure lies on a table and is modelled only to the top of his legs. But most are composite figures, an artist must have modelled them at some point in time but the seven I chose to work with in my exhibition in Belfast (SimBodies & NoBodies) were a mixture of obviously lifelike starting point and then obvious composites. Of more importance in a training context are the internal organs; the way the models look, their face, has no real medical educational benefit (Figure 8). R: Not portraiture then, but ‘faces’. A lot of which appear in many of your previous works. The face in this cast has a highly emotional contemporary content, it is redolent of images associated with the twentieth century history, famine, war, Holocaust. B: That's important – the historical anchoring. But of course it just feels familiar from that kind of imagery, we don’t need to be explicit about it. R: In the Glasgow Exhibition there will be a separate viewing area with a live film feed, showing the cast as it is underway (Figure 10). In the same way you are fascinated by some of the techniques you see or hear about in anatomy, art processes and their demystification might fascinate others? B: I feel a certain relief as well and it's been built into the decision-making that I’m not actually physically representing the spectacle of the sculpture as part of the show, in its raw, broken and undignified form – there's plenty of that conveyed in this mediated image, providing me with a bit of relief that I’m not further exploiting the actual physical being. We talked about it before, he artist's rules, one's own ethical code, and this fitted with mine, not to have the original object physically present. R: On the theme of the exploitative, there has been a live dissection broadcast on television, which you might correlate to historical live dissection in an actual space as it were, perhaps this was regulated event which might have been more dignified, perhaps it's inevitable that the television image would bring into question the dignity of the event. This new space you’ve made, its low lighting and tiered seating, it brings a certain dignity to the event? B: There's a myriad of references but essentially it's an intimate experiential space which hopefully can be valuable on a one-to-one basis – and we’ve talked about the experience of seeing that kind of pose and that kind of expression [of the sculpture] mediated through news broadcast media – but by making this very particular space to experience this work, I hope I can enable other possible forms of engagement. Yes, you are viewing a projected image on a screen the same as all that news footage but the physical context is so very different, as a viewer you are allowed to have different thoughts and so on. R: The risk here would be that this is described as a sculpture being transformed into a single fixed image whereas this viewing room and the second projected image showing the place of construction is about the spatial experience. It is redolent of other types of viewing experience in history and the contemporary realm. Incidentally, thinking back to next door and looking at it onscreen now, your casting room looks explicitly medical or is that my imagination? B: These tables were borrowed from the casting workshop next door and all these pots are the real thing. When I showed the staff here in Glasgow Sculpture Studios images from Pathology laboratories they immediately noticed exactly how like this casting set-up it looked, full of plastic pots, cutting and scraping, measuring and stirring things. Notes 1. A fascinating advocacy for such performed interaction is expressed in Kohn's ‘Performing medicine: the role of theatre in medical education’:Theatre in particular can provide irreducible experiences emanating from the realms of flow knowing and whole knowing. Flow knowing is knowing in relation to others and to our environment. It is expressed through narrative and kinesthetics – our stories and bodies in spatiotemporal communication with our fellow creatures and the world. Whole knowing comes in a flash (epiphanically) through engagement with the arts, and panoramically, through engagement with synoptic disciplines such as history, philosophy and religious studies. The combination of flow and whole knowing, with empirical knowing, grounds every medical story within the context of a human story. It also moves our students to meld their ‘being’ (values, character and beliefs) with their ‘doing’ (technique) towards creation of a full and reflective practice life. Medical Humanities (37): 3–4. 2. Related works appeared in Niet Normaal: Difference on Display (Bluecoat Gallery, Liverpool 21st Jun-9 September 2012); also in the exhibition Cast Contemporaries: artists respond to the completion of the Cast Collection Project (Edinburgh College of Art, 2 August–2 September 2012, curated by Chris Dorsett and Margaret Stewart); see http://castcontemporaries.weebly.com 3. The cast presents an interesting avenue of research for students of the history of Scottish photography. The Scottish anatomist John Goodsir studied Arts at St Andrews University. In 1840 Goodsir went to Edinburgh and in 1841 he was appointed Conservator of the RCS Museum, and in 1842 he moved to the University becoming Curator of the University's Anatomical Museum. The original sculpture was photographed in 1847 two years after its creation and the archives at Surgeons Hall Museum include a unattributed ‘Copy of a wet plate photograph taken in 1847 which was in the possession of Dr William Guy. Presented by D. J. Menzies Campbell’. This is probably Dr William Guy (1810–1885). The linkage of surgery to photography at this time is found in Guy's contemporary Thomas Keith who, with his brother George Skene Keith, both were members of Sir James Young Simpson's team that pioneered the use of chloroform as an anaesthetic. The Keith's were early proponents of photography with George as a founder member of the Photographic Society of Scotland, and who first exhibited his work in 1854. The Keith's father Rev. Dr Alexander Keith sat for calotypes taken by Hill and Adamson around 1843. 4. For a fascinating response to the USA’ growing aversion to the once-hallowed plaster cast as imitation and how ‘dutiful study’ was supplanted by the ‘aesthetic uplift’ of temporary exhibitions within collections such as the Metropolitan Museum see Siegel, K. 2011. p. 140: ‘However plaster casting continues unabated in contemporary American art, George Segal, Jasper Johns, Allan McCollum, and the art of the reproducible and of a simulacra is one of the pillars of contemporary Western art’. Siegel continues her discussion of plaster casts on p. 147. 5. In correspondence Andrew Connell at the Surgeons Hall Museum posited an idea that Arthur Wyllie, conservator of the National Museums of Scotland, and his assistant Brian Melville, may have been responsible for the fibreglass model now on permanent display. Wyllie was a graduate of Edinburgh College of Art and a contemporary of the Scottish artist Joan Eardley. Disclosure statement No potential conflict of interest was reported by the authors. Notes on contributors Craig Richardson is Professor of Fine Art at Nottingham Trent University, and author of ‘Scottish Art since 1960’ (2011), Ashgate: Farnham. Christine Borland is BALTIC Professor at Northumbria University, she was nominated for the Turner Prize in 1997. ==== Refs References Kohn Martin. 2011 Performing Medicine: The Role of Theatre in Medical Education. Medical Humanities 37 3 4 10.1136/jmh.2011.007690 21593242 Richardson Ruth. 2011 Cast into the Light. The Lancet 377 9738 2073 2074 10.1016/S0140-6736(11)60906-9 Sebald Winfried Georg. 1995 The Rings of Saturn. English translation by Michael Hulse. 1998 London Harvill. Siegel Katy. 2011 Since ’45: America and the Making of Contemporary Art. London Reaktion Books.

PMC005xxxxxx/PMC5003127.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00105910.3205/zma001059Doc60urn:nbn:de:0183-zma0010590ArticlePeyton's 4-Steps-Approach in comparison: Medium-term effects on learning external chest compression – a pilot study Die 4-Schritt-Methode nach Peyton im Vergleich: Mittelfristiger Effekt auf das Erlernen der Herzdruckmassage – eine Pilotstudie Münster Tobias *1Stosch Christoph 1Hindrichs Nina 1Franklin Jeremy 2Matthes Jan 31 University of Cologne, Cologne interprofessional SkillsLab and Simulation Centre (KISS), Cologne, Germany2 University of Cologne, Institute of Medical Statistics, Informatics and Epidemiology (IMSIE), Cologne, Germany3 University of Cologne, Department of Pharmacology, Cologne, Germany*To whom correspondence should be addressed: Tobias Münster, University of Cologne, Cologne interprofessional SkillsLab and Simulation Centre (KISS), Cologne, Germany, E-mail: tobias.muenster@uni-koeln.de15 8 2016 2016 33 4 Clinical skillsDoc6023 11 2015 03 6 2016 19 2 2016 Copyright © 2016 Münster et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001059.shtmlIntroduction: The external chest compression is a very important skill required to maintain a minimum of circulation during cardiac arrest until further medical procedures can be taken. Peyton’s 4-Steps-Approach is one method of skill training, the four steps being: Demonstration, Deconstruction, Comprehension and Execution. Based on CPR skill training, this method is widely, allegedly predominantly used, although there are insufficient studies on Peyton’s 4-Steps-Approach for skill training in CPR in comparison with other methods of skill training. In our study, we compared the medium- term effects on learning external chest compression with a CPR training device in three different groups: PEY (Peyton’s 4-Steps-Approach), PMOD (Peyton’s 4-Steps-Approach without Step 3) and STDM, the standard model, according to the widely spread method “see one, do one” (this is equal to Peyton’s step 1 and 3). Material and Methods: This prospective and randomised pilot study took place during the summer semester of 2009 at the SkillsLab and Simulation Centre of the University of Cologne (Kölner interprofessionelles Skills Lab und Simulationszentrum - KISS). The subjects were medical students (2nd and 3rd semester). They volunteered for the study and were randomised in three parallel groups, each receiving one of the teaching methods mentioned above. One week and 5/6 months after the intervention, an objective, structured single assessment was taken. Compression rate, compression depth, correct compressions, and the sum of correct checklist items were recorded. Additionally, we compared cumulative percentages between the groups based on the correct implementation of the resuscitation guidelines during that time. Results: The examined sample consisted of 134 subjects (68% female; age 22±4; PEY: n=62; PMOD: n=31; STDM: n=41). There was no difference between the groups concerning age, gender, pre-existing experience in CPR or time of last CPR course. The only significant difference between the groups was the mean compression rate (bpm): Group 1 (PEY) with 99±17 bpm, Group 2 (PMOD) with 101±16 bpm and Group 3 (STDM) with 90±16 bpm (p=0,007 for Group 3 vs. Group 1 and Group 3 vs. Group 2, Mann-Whitney- U-Test). We observed no significant differences between the groups after the second assessment. Conclusion: Our study showed that there are no essential differences in external chest compression during CPR performed by medical students dependent on the teaching method (Peyton vs. “Non-Peyton”) implemented with regard to the medium-term effects. The absence of benefits could possibly be due to the simplicity of external chest compression. Zusammenfassung Einführung: Die extrakorporale Herzdruckmassage ist eine wichtige Fertigkeit, um ein Minimum der Organdurchblutung bei Patienten/-innen mit Herz-Kreislauf-Stillstand zu gewährleisten, bis weitere medizinische Hilfe geleistet werden kann. Eine Methode um diese Fertigkeit zu vermitteln ist der Ansatz nach Peyton. Dieser besteht aus 4 Schritten: Demonstration, Dekonstruktion, Verständnis und Durchführung. Bezogen auf die kardiopulmonale Reanimation wird eine Überlegenheit dieser Methode gegenüber Anderen angenommen, ist jedoch bisher durch Studien nicht ausreichend gesichert. In unserer Studie haben wir den mittelfristigen Lernerfolg durch die 4-Schritt-Methode nach Peyton („PEY“) mit dem der Modifikation „PMOD“ (4-Schritt-Methode ohne Schritt 3) und dem „STDM“ (Standardmodell, entsprechend der weit verbreiteten Methode „See one Do one“, entspricht der 4-Schritt-Methode ohne Schritt 1 und 3,) anhand von Parametern der Durchführung einer Herzdruckmassage am Modell verglichen. Material und Methoden: Die prospektiv, randomisierte Pilotstudie wurde im Sommersemester 2009 im Kölner Interprofessionellen SkillsLab und Simulationszentrum (KISS) an der Universität zu Köln durchgeführt. Die Probanden (Studierende der Humanmedizin des zweiten und dritten Fachsemesters) nahmen freiwillig an der Studie teil. Die Studierenden wurden in drei parallele Studiengruppen randomisiert, die nach jeweils einer der oben genannten Methoden unterrichtet wurden. Eine Woche sowie fünf bzw. sechs Monate nach Intervention wurden die Probanden in einer objektivierten strukturierten Einzel-Überprüfung bzgl. ihrer Reanimationsfertigkeiten geprüft. Die Ergebnisse der Gruppen wurden hinsichtlich der Parameter Kompressionsfrequenz, Kompressionstiefe, Anteil richtiger Kompressionen sowie Anzahl in einer Checkliste erreichter Items verglichen. Verglichen wurden außerdem Häufigkeitsverteilungen bezogen auf das Umsetzen oder Verfehlen damals geltender Leitlinienempfehlungen zur kardiopulmonalen Reanimation. Ergebnisse: Die untersuchte Stichprobe umfasste 134 Probanden (68% weiblich; Alter: 22±4 Jahre; PEY: n=62; PMOD: n=31; STDM: n=41). Es bestand kein Unterschied zwischen den Gruppen bzgl. Alter, Geschlechterverteilung, Vorerfahrung oder zeitlichem Abstand zu einem vorherigen (Reanimations-) Kurs. Nur bei einem Endpunkt fand sich ein statistisch signifikanter Unterschied: Die mittlere Kompressionsfrequenz in der ersten Prüfung lag in Gruppe 1 (PEY) bei 99±17/Minute, in Gruppe 2 (PMOD) bei 101±16/Minute und in Gruppe 3 (STDM) bei 90±16/Minute (p=0,007 für Gruppe 3 vs. Gruppe 1 sowie Gruppe 3 vs. Gruppe 2 im Mann-Whitney-U-Test). Keiner der untersuchten Parameter unterschied sich zwischen den drei untersuchten Gruppen 5 bzw. 6 Monate nach Durchführung des Kurses. Schlussfolgerung: Unsere Studie liefert keinen Hinweis auf wesentliche Unterschiede bei der Durchführung einer Herzdruckmassage durch Studierende im Rahmen der kardiopulmonalen Reanimation am Modell in Abhängigkeit von der Unterrichtsmethode (Peyton vs. „Non-Peyton“) bezogen auf den mittelfristigen Überprüfungszeitraum. Möglicherweise ist die Herzdruckmassage als Fertigkeit nicht komplex genug, um von der Vermittlung nach Peyton zu profitieren. Cardiopulmonary ResuscitationEducationPeyton's 4 steps approachBasic Life SupportBLSCPR2 stages4 stagesStudentsExternal chest compression ==== Body 1. Introduction The first human external chest compression was described in 1690 by Kouwenhoven et al. [1] and modified over time in terms of setting as well as compression rate, depth and point. It is an effective skill employed to maintain minimum circulation in order to provide organs intolerant against hypoxia with oxygen until the “Return of Spontaneous Circulation” (ROSC) occurs. In Germany, there are approximately 123 sudden cardiac arrests per 100.000 people per year, in only 50-80 cases of which bystanders commence “Basic Life Support” (BLS) [2]. 48% of cardiac arrests occur in public but only in 23% of these do people start Cardiopulmonary Resuscitation (CPR) [3]. Fast diagnoses of cardiac arrest, immediate emergency calls and administration of good quality CPR are the keys to survival in and out of hospital [4]. It should therefore be of particular importance to teach these skills effectively. Practical (clinical) medical skills occupy an increasingly important role in German medical studies. From the amendment of the regulations for medical licensure [Ärztliche Approbationsordnung (ÄApprO)] in 2005 [5] up to the new “National competence based learning goal catalogue” (NKLM), [Nationaler Kompetenzbasierter Lernzielkatalog Medizin (NKLM)], all of these emphasize the importance of practical clinical skills [6]. There is a need for simple teaching methods every teacher can adopt, which are accepted by the students and provide a sustainable outcome. A widespread methodical approach is Peyton’s 4-Steps-Approach [7], [8]. The approach by R. Peyton [9] to teaching practical clinical skills consists of four steps: Demonstration: The teacher performs the skill in real time without comment. This step is taken to provide a benchmark. Deconstruction: The teacher performs every step slowly with an added explanation. The skill should be divided into smaller subsections. Comprehension: The student describes every step of the skill whereupon the teacher performs on instruction. The description and execution do not occur simultaneously. Execution: The student simultaneously narrates and executes step by step. Peyton’s approach combines multiple aspects of learning theory. The learning in Steps 1 and 2 is based on a social-cognitive approach to learning theory, that of model-learning according to Banduras [10], whereas Step 4, the actual implementation and training of the procedure up to its successful application, is associated with the behaviourist learning theory. According to Jawhari et al. [11], the third step of Peyton’s approach is crucial: „The perceptually processed information (Step 1 & Step 2) must be actively manipulated in the working memory in Step 3 to be transferred into the long-term memory“. According to Krautter et al. [12], the description of the procedure without simultaneous administration produces a mental correlate of the procedural motions, which leads to more efficient motor learning and better reproduction. From a constructivist view of learning theory (as in the sense of constructivist pedagogy according to Reich 1997 [13]) Peyton’s approach can be described as the endeavour to create a constructive “place of furthest reaching own world invention” (ibid. 266). Consequently, a combination of all four steps would be necessary in order to achieve success in learning in the sense of a well-established self-construction. Thus, our study was based upon two hypotheses: The execution of all four steps according to Peyton is superior to modifications that omit one or more of the steps in terms of medium-term learning success. Step 3 of Peyton’s approach is crucial and the sole omission of this step diminishes the generated learning success. Based on this we compared the learning success of medical students taught a) according to Peyton (PEY1; Steps 1-4), b) under omission of Step 3 (PMOD1) or c) only with Steps 2 & 4 according to Peyton (STDM1), once a week later and again 5 to 6 months later. The learning success was determined through the practical test of an external chest compression in the context of cardiopulmonary resuscitation on a model. The 4-Step method according to Peyton as an approach to the instruction of practical skills has already been empirically researched. Some of the surveys conveyed that students taught with the help of this method profited from their instruction in comparison with other groups [11], [12], [14], [15], [16]. Other surveys found no advantages to the Peyton method [17], [18], [19]. One study compared the 4-Step method with the 2-Step standard method “See one, do one” in the instruction of cardiopulmonary resuscitation (CPR), just as we did for this very paper, that study however only conducted research into immediate short-term effects after course completion [20]. Despite the generally contradictory and, with respect to CPR, insufficient existence of surveys, Peyton’s 4-Step method has been propagated for the instruction of cardiopulmonary resuscitation by Bullock [21] since the beginning of the twenty-first century, and recommended by the European Resuscitation Council (ERC) until 2015 [22], [23], [24]. For these reasons (and regarding Best Evidence Medical Education, BEME [25], [26]), we have researched the instruction of cardiopulmonary resuscitation by means of the Peyton method compared to modified means of instruction for this particular study. 2. Material and methods 2.1. Study design This prospective and randomised pilot study with three parallel study groups took place during the summer semester of 2009 as part of the first aid course at the Cologne Interprofessional SkillsLab and Simulation Centre (KISS) of the University of Cologne. To this end, the at that point current guidelines of resuscitation of 2005 [27], [28] were taught and appropriately tested. Consequently, there are discrepancies from the now current guidelines of the ERC concerning the resuscitation data. 2.1.1. Test subjects The test subjects of this study were medical students in their second and third semester of graduate school. Participation in the study was voluntary and all subjects enrolled in the course were randomised. The contents and time frame of the course were identical to those of the rescue organisations, but our course required the successful completion of a practical test. The practical instruction in resuscitation alone took up 90 minutes. Tutors of the SkillsLab employed a variety of instructional methods in compliance with the study protocol. The course size was limited to a maximum of 16 participants (Median=13; at a maximum of 16 participants and a minimum of 9 participants). The evaluation only included participants present on the date of the intervention and on the day of the test a week later, as well as having filled out and submitted a questionnaire with epidemiologic data. Test subjects with prior experience2 were also excluded. Further criteria for exclusion are depicted in Figure 1 (Fig. 1). 2.1.2. Tutors/ examiners The course tutors were student employees of the SkillsLab in higher clinical semesters of graduate school, who were also experienced in resuscitation due to prior work in emergency medical services before or during their studies. Being tutors in other courses, they were also trained in the instruction of practical skills, as well as having been prepared for the study in a separate training seminar. The tutors were also implemented as examiners, but only for groups they hadn’t tutored beforehand. 2.1.3. Blinding, data privacy, ethics All subjects enrolled in the course agreed to pseudonymous collection of data as well as anonymous data processing and evaluation in writing3. The participants were given no information concerning the aim of the study, they were only informed about their participation in it. The tutors were not blinded. The emergency education in the course of medical studies at the University of Cologne prescribes annual emergency trainings. Each seminar begins with a revision of known skills learnt beforehand and continues with the instruction of new contents. This way, the students successively further their personal emergency competence with every year. The external chest compression is a firm part of every emergency course. By way of this longitudinal emergency training, we considered no disadvantages for participants of our study regarding the acquisition of personal emergency competence. Potential shortcomings can be compensated during the remainder of their studies. 2.2. Randomisation The subjects were divided into eight groups (A-H) through random numbers (generated by SPSS) with further randomised divisions in each of the eight groups into three instruction groups (Group 1 (PEY), Group 2 (PMOD) and Group 3 (STDM)). The first division clarified the period of instruction: groups A-D were tutored in the first four weeks of semester, groups E-H in weeks five to eight4. The second randomisation determined the instruction/ intervention group. The assignment of tutors to different instruction groups was also randomised. 2.2.1. Course of events In accordance with the fixed study protocol, the groups were tutored and assessed through an objective, structured evaluation regarding their resuscitation competence one week after intervention. First assessment (after 1 week) In this setting, the students were given one minute to read through the description of a situation containing the tasks and contextual conditions (one-helper method). Afterwards, the subjects were given five minutes to complete their tasks: Assessment of the necessity for resuscitation Begin and administration of resuscitation The students were evaluated by an examiner using a (binary) checklist and the relevant performance data of the resuscitation phantom was collected (Laerdal, Resusci Anne Advance Skill ReporterTM). The checklist contained 16 dichotomous items (see Figure 2 (Fig. 2)), all of which were derived from the resuscitation guidelines of 2005 [27]. The completion of the test occurred after more than five full resuscitation cycles or after two minutes of continuous external chest compression, if the task was the administration of “compression-only” resuscitation. Second assessment (after 5-6 months) Five months (groups E-H) and accordingly six months (groups A-D) after course completion, the students were once again assessed through an objectively structured practical test respective of the final test of the first aid course. The data collected concerning resuscitation was evaluated in this study. For reasons pertaining to the right of inspection, a different, already existing checklist containing eight items was used (see Figure 3 (Fig. 3)). The resuscitation was performed under the same conditions as in the first assessment. The collection of performance data regarding the phantom occurred using the same resuscitation model. The students had been informed about the test, were familiar with the contents of the course and were able to prepare for the test with the help of a resuscitation phantom. 2.3.Teaching methods For this study, the steps were modified as follows. 2.3.1.Group 1 (PEY): 4-Step method according to Peyton Since the original approach according to Peyton was conceived for one-on-one assistance in the operating theatre, we respectively appropriated the comprehension step (Step 3) to the group setting. Step 1 (Demonstration) and 2 (Deconstruction): Unaltered. Step 3 (Comprehension): 1. One student describes the course of action, the tutor then executes this course while the rest of the group observes. 2. Groups of three students each were formed around a resuscitation phantom. One student describes the course of action, another executes it while the third remaining student takes on the role of the observer. After the completion of the administration, the roles are switched in such a way that every subject has to have executed “Step 3 (Comprehension)”. The course tutor “monitors” group formation and steps in to correct mistakes if need be. Step 4 (Execution): Unaltered, groups of three practice resuscitation on a phantom. 2.3.2. Group 2 (PMOD): Omission of Step 3 according to Peyton Step 1 (Demonstration) und Step 2 (Deconstruction): Unaltered. Step 3 (Comprehension): Omitted. Step 4 (Execution): Groups of three practice resuscitation on a phantom without the division of roles described in 2.3.1. 2.3.3. Group 3 (STDM): Omission of Steps 1 and 3 Step 1 (Demonstration): Omitted. Step 2 (Deconstruction): Unaltered. Step 3 (Comprehension): Omitted. Step 4 (Execution): Groups of three practice resuscitation on a phantom without the division of roles described in 2.3.1. This complies with the so-called classic instruction “See one, do one”. The tutor explains and demonstrates the resuscitation (Step 2), after which all questions and uncertainties are clarified and practice on a resuscitation phantom in groups of three follows. 2.3.4. Evaluation, resources, data work flow The checklist items were collected on paper, digitalised using OMR-Office® Version 5 of the company Remark® and ported using Excel® versions 12 and 14 for Mac OSX of the company Microsoft. The CPR data was collected using the resuscitation model Resusci Anne Advanced Skill Reporter® of the company Laerdal®, assigned to the subjects per hardcopy and then digitalised. For statistical analysis we employed SPSS® versions 20 and 22 for Mac OSX of the company IBM®. With regards to the quality of resuscitation, we considered the parameters depicted in Table 1 (Tab. 1), on the one hand as actual measured value for direct comparison, but also categorised into dichotomous values (correct/ false). The basis of categorisation was provided by the resuscitation guidelines of 2005 [27] as well as the publications of Kern et al. 1992 [29] and Abella et al. 2005 [30] for the determination of the permitted average compression rate of 90-110 spm. The same targeted frequency range was later also employed by Sopka et al. [23] and Jenko et al. [20]. We arbitrarily determined a limit of 60% in the sense of a fail for the number of “correct compressions5” and checklist items. 2.4. Statistical methods The data was processed and analysed via SPSS (IBM® SPSS® version 22, Mac OSX). The testing of the data for normal distribution was performed through visual inspection via histogram and Q-Q-Plot, as well as mathematically with the aid of the Shapiro-Wilks-Test. Testing for significant differences in the epidemiologic data of the groups was calculated via Chi-Square-Test and Fischer’s Exact Test. Because not all the data was normally distributed, testing for differences in age distribution and CPR data was calculated according to Kruskal-Wallis. The Chi-Square-Test was used again for the comparison of categorised data. The level of significance was reduced to 0.0125 according to Bonferroni due to the comparison of four features. In order to simplify evaluation, the assumption was made that no interdependencies existed between the results of single subjects of a group. Therefore, all statements on statistical significance must be understood in the exploratory sense. Consistent values are given as arithmetic mean±standard deviation (MW±SD). 3. Findings For a tabular summary see Table 1 (Tab. 1), Table 2 (Tab. 2), Table 3 (Tab. 3), Table 4 (Tab. 4) and Table 5 (Tab. 5). 3.1. First Assessment, one week after intervention: The analysed sample consists of 134 subjects (68% female; age 22±4 years; PEY: n=62; PMOD: n=31; STDM: n=41). There was no significant difference between the groups in terms of gender distribution (p= 0,887 [χ2-Test]), pre-existing CPR experience (p=0,790 [Fischer’s Exact Test]), time of last CPR course (p=0,582 [χ2-Test]) and number of compressions during CPR (p=0,064 [Kruskal-Wallis], see Table 2 (Tab. 2) and Table 3 (Tab. 3)). For exclusion criteria see Figure 1 (Fig. 1). 3.1.1. Compression rate We calculated a mean compression rate in the first assessment for Group 1 (PEY): 99±17/minute, Group 2 (PMOD): 101±16/minute and Group 3 (STDM): 90±16/minute (p=0,007 for Group 3 vs. Group 1, as well as Group 3 vs. Group 2 in Mann-Whitney-U-Test). The percentage of subjects with the correct compression rate (defined a priori as 90-110 bpm) was 29 subjects for Group 1 (PEY) (47%), 13 subjects for Group 2 (PMOD) (42%) and 14 subjects for Group 3 (STDM) (34%). There was no significant difference between the groups, p=0,451 [χ2-Test]. 3.1.2. Compression depth There was no significant difference in mean compression depth between the three groups: Group 1 (PEY): 36±10 mm; Group 2 (PMOD): 38±8 mm, Group 3 (STDM): 38±11 mm (p=0,572 [Kruskal-Wallis]). The comparison of fractions with correct compression depth (defined a priori as ranging from 40-50 mm) showed the following: Group 1 (PEY) with 25 subjects (40%), Group 2 (PMOD) with 15 subjects (48%) and Group 3 (STDM) with 14 subjects (34%) within the correct range of compression depth. There was no significant difference between the groups, p= 0,457 [χ2-Test]. 3.1.3. Correct Compression When comparing “correct compressions”5 to all compressions in percentages, there is no visible significant difference between the groups (p=0,417 [Kruskal-Wallis]): Group 1 (PEY): 38%±34; Group 2 (PMOD): 44%±34; Group 3 (STDM): 43%±32. When comparing the number of subjects with at least 60% “correct compressions”, we found 18 subjects in Group 1(PEY) (29%), 13 subjects in Group 2 (PMOD) (42%) and 14 subjects in Group 3 (STDM) (34%) with 60% or more correct compressions (p=0,470 [χ2-Test]). 3.1.4. Checklist items When comparing the number of correct checklist items (at a maximum of 16), we found no significant difference between the groups (Group 1 (PEY): 13±2; Group 2 (PMOD): 13±2; Group 3 (STDM): 13±2), p=0,487 [Kruskal-Wallis]). When comparing the number of subjects with at least 60% correct checklist items, we found no significant difference either (p=0,479, [χ2-Test]). Group 1 (PEY) had 58 (94%) subjects, Group 2 (PMOD) had 31 (100%) subjects and Group 3 (STDM) had 39 (95%) subjects with 60% correct checklist items or more. 3.2. Second assessment, five/six months after intervention Compared to the first assessment, a loss of 4 subjects occurred because of insufficient data. 3.2.1. Compression rate The arithmetic mean of the average compression rate was 112±12/minute in Group 1 (PEY), 113±13/minute in Group 2 (PMOD) and 108±15/minute in Group 3 (STDM) (p=0,600 in the Kruskal-Wallis-Test). When comparing the percentage of subjects with a mean compression rate within the a priori determined range of 90-110/minute, there were 27 subjects in Group 1 (PEY) (44%), 12 subjects in Group 2 (PMOD) (40%), and 14 subjects in Group 3 (STDM) (37%). There was no evidence of significant difference between the groups (p=0,801 [χ2-Test]). 3.2.2. Compression depth We could not detect any significant difference concerning the average compression depth when comparing the individual groups, (p=0,942 in the 'Kruskal-Wallis-Test'). We measured an average compression depth of 42±8mm for Group 1 (PEY), 42±7mm for Group 2 (PMOD) and 42±8mm for Group 3 (STDM). Comparing the number of students with an average compression depth within the a priori determined target range from 40-50mm, we counted 33 subjects for Group 1 (PEY) (53%), 13 subjects for Group 2 (PMOD) (43%) and 19 subjects for Group 3 (STDM) (50%) (p= 0,669, [χ2-Test]). 3.2.3. Correct compressions Comparing the percentages of “correct compressions”5, we found 49%±32 for Group 1 (PEY), 53%±30 for Group 2 (PMOD) and 48%±31 for Group 3 (STDM) of correct compressions (p=0,840, in the “Kruskal-Wallis-Test”). Comparing the groups with regard to the number of subjects with at least 60% correct compressions, we found 27 subjects in Group 1 (PEY) (44%), 14 subjects in Group 2 (PMOD) (47%) and 17 subjects in Group 3 (STDM) (45%) (p=0,973, [χ2-Test]). 3.2.4. Checklist items The following mean values were produced for the different groups by the comparison of their number of correct checklist items (at max. 8): Group 1 (PEY) 7±1; Group 2 (PMOD) 8±1 and Group 3 (STDM): 8±1 (p=0,824, Kruskal-Wallis-Test). We could not detect a significant difference between the groups within the comparison of the fraction of subjects with at least 60% correct checklist items (p=1,000 [χ2-Test]). In Group 1 (PEY), this fraction was 61 subjects (98%), in Group 2 (PMOD) it was 30 (100%) and in Group 3 (STDM) it was 38 (100%). 4. Discussion In this study, we compared three different methods of practical instruction of basic life support with respect to medium-term learning achievement of resuscitation performance. This aspect is of importance because of the necessity of using sustainable instruction methods precisely for the instruction of basic life support. The accessibility of CPR skills decreases over time [31] and corresponds to roughly the level before a course after one to two years [32]. Jenko et al. [20] were able to illustrate the absence of advantages of the instruction of cardiopulmonary resuscitation via Peyton’s 4-Step method compared to the 2-Step standard model (“See one, do one”) concerning the short-term learning achievement (immediately after course completion). A further study, published as an abstract, found no difference between groups taught using Peyton’s method and those taught using the 2-Step method after a period of three months [19]. However, the aim of that course was the administration of the recovery position, which is why those findings can only be compared to ours to a certain extent. In the collective we researched, there was only one significant difference, namely the mean compression rate in the first assessment (one week after intervention). The subjects in Group 3 (STDM) instructed according to the “See one, do one” approach we call the standard model (Peyton’s Steps 2 and 4), were significantly slower in the first assessment one week after the intervention than the subjects in the other groups (PEY: Steps 1-4; PMOD: Steps 1, 2, and 4 (see Table 4 (Tab. 4))). The average compression rate in Group 3 (STDM) was still within the limits prescribed by the then current guidelines for resuscitation. There was no significant difference when comparing the percentages of subjects with the correct compression rate. Consequently, the practical relevance of the observed difference in mean compression rate dependent on the teaching method is improbable. The limitation is set higher in the recently released new guidelines for resuscitation by the ERC [33]: 100-120 spm. Nonetheless, these were not the basis of the course and therefore unknown by the students. Further studies must show which factors influence the learning of the correct compression rate. Instruction in conjunction with music, for example, supposedly leads to a positive effect on learning according to Hafner et al. [34]. The significant difference in mean compression rate was not evident in the assessment 5 to 6 months later. We attribute this most likely to a training and practice effect, due to the subjects’ knowledge of a second assessment and their subsequent chance at preparing themselves specifically for it. However, this could also be a reason for our inability to measure differences in the medium term, since deficits stemming from the instruction method could have been compensated through targeted training. The preparation time was not separately recorded, hence we do not know of the existence of differences in preparation time between different groups. The randomisation of the participants should ensure the rough equivalence of this “risk” in all groups. Apart from a systematic recording, we have no reason to believe that more training material (reanimation phantoms) was taken advantage of before the second assessment than in previous semesters, nor that corresponding training rooms were more highly frequented (loan of material and room keys occurs solely through our SkillsLab). The students were not informed about which group they were assigned to. It is, however, a safe assumption that the students exchanged details of the course contents amongst themselves and thus realised there were differences in instruction. This means we can only talk of blinding to a certain extent, which in turn means we cannot rule out the possibility of distortion. Whether this had an impact on the measurable performance of resuscitation, specifically on the compression rate, is questionable. A structured registration of preparation time, frequency and partners should therefore occur in a follow-up study. The division of subjects into two large subcategories (first and subsequent four weeks of semester) was necessary due to the prescribed schedule intending these time periods for the first aid course. The randomisation should have excluded the possibility of distortion in this case. We appointed student employees of the SkillsLab as tutors, who had been trained beforehand. The equal suitability of trained students as tutors to that of professional staff has been illustrated by Tolsgaard et al. [35], among others. The final analysable number of cases was significantly reduced due to an unexpectedly high drop-out rate. With respect to our recorded prevalence for the correct compression rate of Group 1 (PEY) and Group 3 (STDM), 294 subjects per group based on a power of 80% and α=0.05 would have to be researched in a new study. For the prevalence of the correct compression depth, the amount would be 985 subjects. Our study is therefore classified as underpowered and can only be considered a pilot study. It is possible that our transfer of Peyton’s 4-Step method from a one-on-one setting into a group setting decreased its effectiveness. Roughly only two thirds of the students in Group 1 (PEY) followed the correct order of Peyton’s 4-Step method, i.e. the verbalisation of the action prior to its execution. A third performed the action first contingent upon modification. Although this occurred following the instruction of the subject who was executing Step 3 rather than their own intention, the originally intended order was nevertheless not followed. Nikendei et al. also applied Peyton’s approach to a group context [36]. In their case, the originally intended order of steps was correctly followed for all students. Further studies must show, whether the order of Peyton’s steps has crucial impact on the result, or whether the third step, for example, still has an effect after the first motor performance of the action. In conclusion, we were unable to prove a superiority of Peyton’s 4-Steps-Approach (PEY) compared to the version modified through “omission of Step 3” (PMOD), nor compared to the “See one, do one” method (STDM) in our study with the limitations specified above. In the recently published new guidelines for resuscitation of the ERC, the recommendation for the application of the 4-Step method is revoked [31]. This retraction is, however, based on studies that focused their research on the instruction of other skills6 than cardiopulmonary resuscitation [17], [18]. In our opinion, the transfer of these study results onto the instruction of CPR is nonetheless not possible. Peyton’s 4-Steps-Approach was originally described as a way of acquiring and practicing complex clinical skills (specifically practical operative skills) [9]. There are studies that corroborate the effectiveness of Peyton’s method. Gradl et al. [16] illustrated the superiority of Peyton’s method in the publication of an abstract on the instruction of complex manual-therapeutic skills, wherein 100 dichotomous items were observed. Krautter et al. [12] were able to illustrate the superiority of results gained by the instruction of a central venous catheter (CVC) insertion via the complete 4-Step method according to Peyton, as a complex clinical skill encompassing 39 procedural treatment instructions, compared to only using parts of the 4-Step method. Contrary to these results, Bjørnshave et al. [19] showed in a published abstract that subjects who were taught the maneuvering of an unconscious person into the recovery position in 8 procedural steps did not profit from the 4-Step method. Greif et al. [17] also illustrated this with respect to the instruction of an emergency cricothyroidotomy (percutaneous needle-puncture cricothyroidotomy) with 10 procedural steps, as well as Orde et al. [18] for the instruction of the larynx mask insertion with 9 procedural steps. Peyton’s 4-Steps-Approach seems not to offer a measurable advantage in the context of less complex clinical (emergency) skills. Over the years, the complexity of basic life support has been steadily reduced so as to be easily accessible for a wide range of learners and quickly accessible in case of an emergency. This can be exemplified by the finding of a pressure point: In the resuscitation guidelines of 2000 [37], the finding of the correct pressure point alone consists of six successive instructions, five years later the same procedure consists of only one step [27]. Currently, the entirety of resuscitation is condensed into three words by the “Save a life” campaign of the professional organisation of German anaesthesiologists and the German Society of Anaesthesiology and Intensive Care Medicine (DGAI): “Check, call, press” [https://www.einlebenretten.de/]. Our study contained 19 procedural steps (16 dichotomous items an 3 resuscitation parameters, see Figure 2 (Fig. 2) and Table 4 (Tab. 4)). It is therefore conceivable, that the reduction in complexity renders Peyton’s 4-Steps-Approach less effective in higher selected subject collectives (such as medical students). The approach would nonetheless be interesting regarding wider education (medical laypersons). Krautter et al. 2011 [14] also illustrated the superiority of Peyton’s approach on another level in terms of the instruction of gastric-tube insertion: Procedurally compared (through checklists, 13 dichotomous items) the 4-Step method was not superior to the classic instruction method, it did however display advantages with respect to doctor-patient communication, professional manner and the speed of execution of the activity. This aspect was also shown by Lund et al. [15] for the instruction of an IV cannulation, although the intervention group (4-Step method) in this study was also superior to the control group (“See one, do one”) in procedural comparison (25 dichotomous items). The dimension of the Peyton method that transcends procedural comparison was not considered in our study, since communicative and hygienic skills as well as professional manner are not of utmost importance in basic life support. This aspect does, however, become important when it comes to the instruction of advanced resuscitation skills (Advanced Life Support, ALS), where team communication and (leading) role assignment are crucial. Further studies with larger numbers of cases should clarify, whether following the order of Peyton’s steps is obligatory, whether the choice of teaching method has an impact on the participants’ motivation, whether the assimilation of competence and sense of mastery can be achieved and whether the teaching method should be adjusted to the target audience. Furthermore, the suitability of the 4-Step method according to Peyton as a tool for inexperienced tutors to gain structure and security in the instruction of practical skills needs to be clarified, as well as the optimal subject group size for this instructional approach. More creativity and courage to try out new teaching methods should be shown in the search for teaching methods for basic life support in tutored courses, each of which should of course be scientifically accompanied and evaluated with respect to Best Evidence Medical Education (BEME). Peyton’s 4-Steps-Approach should be questioned in the sense of a “dogma” [38], which has already occurred through the recent release of the new resuscitation guidelines of the ERC, it should nevertheless not be completely taken out of the instructional repertoire for cardiopulmonary resuscitation. Notes 1 PEY=Peyton; PMOD=Peyton modified; STDM=standard model 2 Prior experience: first aid course at some point during the last two years, apprenticeship as paramedic, physiotherapist, nurse, in emergency rescue services 3 In the model course of studies in Cologne, students globally accept the personalised, pseudonymous and/ or anonymous collection and processing, but never the personalised publication, of their personal data for research purposes. 4 These time periods were reserved for the first aid course in the course of studies and were not freely changeable by us. 5 “Correct compressions” or “Compressions without faults” are compressions with the correct pressure point, the correct position and with complete decompression after each single compression. 6 Larynx mask insertion [18] and percutaneous needle-puncture cricothyroidotomy [17] Acknowledgements Many thanks go to the numerous students working at our SkillsLab, who supported this study and made it possible. Special thanks in particular go to: Elisabeth Sauer, Katharina Albrecht, Carsten Wessels, Patrick Lang, Daniel Weber, Traugott Gruppe and David Schwarz. Competing interests The authors declare that they have no competing interests. Table 1 CPR target value Table 2 Epidemiological data 1 Table 3 Epidemiological data 2: Previous experience, time of last course Table 4 CPR-Data Table 5 Binary CPR-Data Figure 1 Procedure and Drop-Out Figure 2 Checklist items first Assessment (16 items) Figure 3 Checklist items second Assessment (8 items) ==== Refs 1 Kouwenhoven WB Jude JR Knickerbocker GG Closed-chest cardiac massage JAMA 1960 173 1064–1067 10.1001/jama.1960.03020280004002 Available from: http://dx.doi.org/10.1001/jama.1960.03020280004002 14411374 2 Gräsner JT Seewald S Bohn A Fischer M Messelken M Jantzen T Wnent J German resuscitation registry Anaesthesist 2014 63 6 470–476 10.1007/s00101-014-2324-9 Available from: http://dx.doi.org/10.1007/s00101-014-2324-9 24895005 3 Gräsner JT Wnent J Gräsner I Seewald S Einfluss der Basisreanimationsmaßnahmen durch Laien auf das Überleben nach plötzlichem Herztod Notfall Rettungsmed 2012 15 7 593 599 10.1007/s10049-012-1584-7 Available from: http://dx.doi.org/10.1007/s10049-012-1584-7 4 Waalewijn RA Tijssen JG Koster RW Bystander initiated actions in out-of-hospital cardiopulmonary resuscitation: results from the Amsterdam Resuscitation Study (ARRESUST) Resuscitation 2001 50 3 273–279 10.1016/S0300-9572(01)00354-9 Available from: http://dx.doi.org/10.1016/S0300-9572(01)00354-9 11719156 5 Schnabel KP Boldt PD Breuer G Fichtner A Karsten G Kujumdshiev S Schmidts M Stosch C A consensus statement on practical skills in medical school - a position paper by the GMA Committee on Practical Skills GMS Z Med Ausbild 2011 28 4 Doc58 10.3205/zma000770 Available from: http://dx.doi.org/10.3205/zma000770 22205916 6 MFT Medizinischer Fakultätentag der Bundesrepublik Deutschland Nationaler Kompetenzbasierter Lernzielkatalog Medizin 2015 Berlin MFT Medizinischer Fakultätentag der Bundesrepublik Deutschland e. V 346 7 Peyton JW Teaching and Learning in Medical Practice 1998 Heronsgate Rickmansworth, Herts. Manticore Europe Ltd 8 Fabry G Medizindidaktik 2008 Bern Verlag Hans Huber 9 Walker M Peyton JW Teaching in the Theatre Teaching and Learning in Medical Practice. 1998 Heronsgate Rickmansworth, Herts. Manticore Europe Ltd 171–180 10 Banduras A Lernen am Modell: Ansätze zu einer sozial-kognitiven Lerntheorie 1976 Stuttgart Klett-Verlag 11 Jawhari J Krautter M Dittrich R Jünger J Nikendei C Instruktion im Skills-Lab: Differentielle Effekte der Peyton-Schritte auf die Gedächtnisleistung 2012 Jahrestagung der Gesellschaft für Medizinische Ausbildung (GMA) 27.-29.09.2012 Aachen Düsseldorf German Medical Science GMS Publishing House DOcV568 10.3205/12gma199 Available from: http://dx.doi.org/10.3205/12gma199 12 Krautter M Dittrich R Safi A Krautter J Peyton's four-step approach: differential effects of single instructional steps on procedural and memory performance–a clarification study Adv Med Educ Pract 2015 6 399 406 10.2147/AMEP.S81923 Available from: http://dx.doi.org/10.2147/AMEP.S81923 26060417 13 Reich K Systemisch-konstruktivistische Pädagogik – Einführung in Grundlagen einer interaktionitisch-konstruktivistischen Pädagogik 1997 2nd ed. Neuwied, Kriftel, Berlin Luchterhand 14 Krautter M Weyrich P Schultz JH Buss SJ Maatouk I Jünger J Nikendei C Effects of Peyton's four-step approach on objective performance measures in technical skills training: a controlled trial Teach Learn Med 2011 23 3 244–250 10.1080/10401334.2011.586917 Available from: http://dx.doi.org/10.1080/10401334.2011.586917 21745059 15 Lund F Schultz JH Maatouk I Krautter M Möltner A Werner A Weyrich P Jünger J Nikendei C Effectiveness of IV Cannulation Skills Laboratory Training and Its Transfer into Clinical Practice: A Randomized, Controlled Trial PLoS 2012 7 3 e32831 10.1371/journal.pone.0032831 Available from: http://dx.doi.org/10.1371/journal.pone.0032831 16 Gradl G Luebke C Muenker R Steinbusch J Pape HC Knobe M Die Vermittlung komplexer manualtherapeutischer Fertigkeiten – Konventionelle Lehre oder 4-Schritt-Methode nach Peyton: Eine prospektive randomisierte Studie 2015 Gemeinsame Jahrestagung der Gesellschaft für Medizinische Ausbildung (GMA) und des Arbeitskreises zur Weiterentwicklung der Lehre in der Zahnmedizin (AKWLZ) 30.09.-03.10.2015 Leipzig Düsseldorf German Medical Science GMS Publishing House DocV534 10.3205/15gma121 Available from: http://dx.doi.org/10.3205/15gma121 17 Greif R Egger L Basciani RM Lockey A Vogt A Emergency skill training—A randomized controlled study on the effectiveness of the 4-stage approach compared to traditional clinical teaching Resuscitation 2010 81 12 1692–1697 10.1016/j.resuscitation.2010.09.478 Available from: http://dx.doi.org/10.1016/j.resuscitation.2010.09.478 21067856 18 Orde S Celenza A Pinder M A randomised trial comparing a 4-stage to 2-stage teaching technique for laryngeal mask insertion Resuscitation 2010 81 12 1687–1691 10.1016/j.resuscitation.2010.05.026 Available from: http://dx.doi.org/10.1016/j.resuscitation.2010.05.026 20638765 19 Bjørnshave K Krogh LQ Hansen SB Nebsbjerg MA Rasmussen SE Thim T Løfgren B Abstract 19242: No Difference in Skill Acquisition and Retention When Teaching Laypersons Recovery Position Using Four-stage and Two-stage Teaching Technique: A Randomized Comparison Circulation 2015 131 A19242 20 Jenko M Frangež M Manohin A Four-stage teaching technique and chest compression performance of medical students compared to conventional technique Croat Med J 2012 53 5 486–495 10.3325/cmj.2012.53.486 Available from: http://dx.doi.org/10.3325/cmj.2012.53.486 23100211 21 Bullock I Skill acquisition in resuscitation Resuscitation 2000 45 2 139–143 10.1016/S0300-9572(00)00171-4 Available from: http://dx.doi.org/10.1016/S0300-9572(00)00171-4 10950322 22 Nolan J Baskett P Advanced life support course provider manual 2001 4 ed. London Resuscitation Council (UK) 23 Sopka S Biermann H Rossaint R Knott S Skorning M Brokmann JC Heussen N Beckers SK Evaluation of a newly developed media-supported 4-step approach for basic life support training Scand J Trauma Resusc Emerg Medc 2012 20 37 10.1186/1757-7241-20-37 Available from: http://dx.doi.org/10.1186/1757-7241-20-37 24 Perkins G ERC Course strategy 2008 2008 09.02.2016 genoplivning.dk Available from: http://genoplivning.dk/wp-content/uploads/2013/08/ERC-Course-strategy-2008.pdf 25 Harden MR Grant J Buckley G Hart IR BEME Guide No. 1: Best Evidence Medical Education Med Teach 2009 21 6 553–562 26 Fabry G Fischer MR Replication - The ugly duckling of science? GMS Z Med Ausbild 2015 32 5 Doc57 10.3205/zma000999 Available from: http://dx.doi.org/10.3205/zma000999 26604999 27 Handley AJ Koster R Monsieurs K Perkins GD Davies S Bossaert L Bahr J Lebensrettende Basismaßnahmen für Erwachsene und Verwendung automatisierter externer Defibrillatoren Notfall Rettungsmed 2006 9 1 10–25 10.1007/s10049-006-0792-4 Available from: http://dx.doi.org/10.1007/s10049-006-0792-4 28 Baskett P Nolan J Handley A Soar J Biarent D Richmond S Prinzipien des Trainings in der Wiederbelebung Notfall Rettungsmed 2006 9 1 164–70 10.1007/s10049-006-0799-x Available from: http://dx.doi.org/10.1007/s10049-006-0799-x 29 Kern KB A Study of Chest Compression Rates During Cardiopulmonary Resuscitation in Humans Arch Intern Med 1992 152 1 145–149 10.1001/archinte.1992.00400130153020 Available from: http://dx.doi.org/10.1001/archinte.1992.00400130153020 1728910 30 Abella BS Sandbo N Vassilatos P Alvarado JP O'Hearn N Wigder HN Hoffman P Tynus K Vanden Hoek TL Becker LB Chest Compression Rates During Cardiopulmonary Resuscitation Are Suboptimal A Prospective Study During In-Hospital Cardiac Arrest Circulation 2005 111 4 428–434 10.1161/01.CIR.0000153811.84257.59 Available from: http://dx.doi.org/10.1161/01.CIR.0000153811.84257.59 15687130 31 Greif R Lockey AS Conaghan P Lippert A De Vries W Monsieurs KG Education and Implementation of resuscitation section Collaborators Collaborators European Resuscitation Council Guidelines for Resuscitation 2015: Section 10. Education and implementation of resuscitation Resuscitation 2015 95 288–301 10.1016/j.resuscitation.2015.07.032 Available from: http://dx.doi.org/10.1016/j.resuscitation.2015.07.032 26477418 32 Moser DK Coleman S Recommendations for improving cardiopulmonary resuscitation skills retention Heart Lung 1992 21 4 372–380 1629007 33 Perkins GD Handley AJ Koster RW Castrén M Smyth MA Olasveengen T Monsieurs KG Raffay V Gärsner JT Wenzel V Ristagno G Soar J Adult basic life support and automated external defibrillation section Collaborators European Resuscitation Council Guidelines for Resuscitation 2015: Section 2. Adult basic life support and automated external defibrillation Resuscitation 2015 95 81–99 10.1016/j.resuscitation.2015.07.015 Available from: http://dx.doi.org/10.1016/j.resuscitation.2015.07.015 26477420 34 Hafner JW Jou AC Wang H Bleess BB Tham SK Death Before Disco: The Effectiveness of a Musical Metronome in Layperson Cardiopulmonary Resuscitation Training J Emerg Med 2015 48 1 43–52 10.1016/j.jemermed.2014.07.048 Available from: http://dx.doi.org/10.1016/j.jemermed.2014.07.048 25282123 35 Tolsgaard MG Gustafsson A Rasmussen MB HØiby P Müller CG Ringsted C Student teachers can be as good as associate professors in teaching clinical skills Med Teach 2007 29 6 553–557 10.1080/01421590701682550 Available from: http://dx.doi.org/10.1080/01421590701682550 17978968 36 Nikendei C Huber J Stiepak J Huhn D Lauter J Herzog W Jünger J Krautter M Modification of Peyton's four-step approach for small group teaching – a descriptive study BMC Med Educ 2014 14 68 10.1186/1472-6920-14-68 Available from: http://dx.doi.org/10.1186/1472-6920-14-68 24690457 37 Handley AJ Monsieurs KG Bossaert LL European Resuscitation Council Guidelines 2000 for Adult Basic Life Support Resuscitation 2001 48 3 199–205 10.1016/S0300-9572(00)00377-4 Available from: http://dx.doi.org/10.1016/S0300-9572(00)00377-4 11278083 38 Barelli A Scapigliati A The four-stage approach to teaching skills: The end of a dogma? Resuscitation 2010 81 12 1607 1608 10.1016/j.resuscitation.2010.09.010 Available from: http://dx.doi.org/10.1016/j.resuscitation.2010.09.010 20961678

PMC005xxxxxx/PMC5003128.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00104910.3205/zma001049Doc50urn:nbn:de:0183-zma0010491Article“Pass, fail” – On Standard Setting Procedures for the Assessment of Practical Skills at Medical Schools in Germany, Austria, and Switzerland „Bestanden, nicht bestanden“ – über den Einsatz von Standardsetzungsverfahren im Kontext praktischer Prüfungen an Medizinischen Fakultäten der D/A/CH-Länder Bauer Daniel *123Huwendiek Sören 23März Maren 431 Klinikum der Universität München, Institut für Didaktik und Ausbildungsforschung in der Medizin, München, Deutschland2 Universität Bern, Institut für Medizinische Lehre, Bern, Schweiz3 Gesellschaft für Medizinische Ausbildung e.V., Ausschuss Prüfungen, Erlangen, Deutschland4 Charité Universitätsmedizin Berlin, Referat für Studienangelegenheiten, Berlin, Deutschland*To whom correspondence should be addressed: Daniel Bauer, Universität Bern, Institut für Medizinische Lehre, Bern, Schweiz, E-mail: daniel.bauer@iml.unibe.ch15 8 2016 2016 33 4 Clinical skillsDoc5007 12 2015 29 6 2016 29 6 2016 Copyright © 2016 Bauer et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001049.shtml ==== Body Short report The members of the GMA committee on assessment have the common goal to make exams in the training and continuing education of health professionals in Germany, Austria and Switzerland (D/A/CH countries) as appropriate as possible. For this purpose, they engage in a discourse with each other, considering both scientific research and regional/local contexts. The committee is pleased to present in the present special issue its views on practical skills from the assessment perspective. The teaching of practical skills is now well-established in the medical programs in the D/A/CH region, reflected in the national catalogues of learning objectives: The Swiss Catalogue of Learning Objectives for Undergraduate Medical Training (SCLO) [1], the Austrian Competence Catalogue for Medical Skills (ÖKÄF) [2], and the German National Competence Based Catalogues of Learning Objectives for Undergraduate Medical Education (NKLM) [3]. The assessment of practical skills, too, is largely established. As early as 2007, over 70% of Germany’s medical programs featured practical skills assessment, most of which in OSCE format [4], and since 2011, an OSCE complements the Swiss Federal Licensing Examination and must be passed independently of the MCQ examination [5]. Besides OSCE, the oral-practical examination [6] is a common format, but also OSPEs (Objective Structured Practical/Preclinical Exams) are used [7]. This shows that practical examinations in medical education are indeed varied and widespread, but what about their quality? Yudkowsky lists various criteria compromising construct validity – i.e., the meaningful interpretability of the data obtained in an exam [8]. As is the case with exams of other formats, it is necessary in the assessment of practical skills such as the OSCE to prevent construct underrepresentation by providing sufficient and representative examination content and to minimize construct-irrelevant variance, i.e., to ignore, in interpreting exam performance, such information not relating to the subject matter under scrutiny. Quality-maintaining measures are being taken in many programs, such as blueprinting (selection of adequate, representative exam content), review (e.g., with respect to wording of tasks and checklists, and to anticipate task difficulty), as well as training of examiners and possibly, simulated patients involved. However, it seems that one aspect of quality assurance has been receiving only little attention: the deliberate setting of grades and pass/fail boundaries. In a commentary recently published, Tekian & Norcini advocate questioning the traditional, ubiquitous 60% pass boundary, and recommend to apply methods based on the assessment content in order to determine pass/fail and grades, for only this way, comparability can be established between test cohorts [9]. The challenge is, therefore, to operationalize these general requirements within the context of one’s actual practical skills assessment (the so-called standard setting) [10], [11]. So far, there is little data on the use of these criterion-oriented standard setting procedures in practical examinations at the medical schools of the D/A/CH countries. Accordingly, we exemplarily analysed 20 study and/or exam regulations or equivalent documents available on the internet relating to medical curricula in Germany, including all model programs. It turned out that the majority of programs apply the 60% pass boundary to practical exams [12], [13], [14], while some do not specifically address grades and pass criteria for practical exams [15], [16], [17], [18], [19] beyond perhaps referring to general directives (e.g., §13 (2) ÄAppO). Nevertheless, this common 60% rule does not rule out criterion-oriented standard setting approaches. Tekian & Norcini explicitly mention procedures under which a criterion-oriented pass/fail boundary can be rescaled to 60%, so that - if necessary - the traditional 60% can be retained. In fact, the Hannover medical program uses such a procedure in which an examination board determines, according to the [anticipated] difficulty and scope of the tasks, the maximum score from which at least 60% must be achieved to pass an exam [20]. Some regulations delegate the burden on the examiners [21], [22], [23], [24], occasionally with the caution to apply "proper methods" [25], or have an examination committee in place [26], [27], [28] to define pass and grade criteria. One faculty explicitly allows criterion-oriented methods for determining the pass boundaries [29], while others require only the disclosure of pass criteria at the beginning of the respective term, without the obligation to further specify them [30], [31]. The Medical Universities of Vienna [32] and Graz [33] use norm-oriented procedures while Innsbruck allows some adaptation depending on task difficulty [34]. The pass boundary for the clinical skills part of the Swiss Federal Licensing Examination [35] is determined with the borderline regression method, which is also true, e.g., for the OSCE in Basel’s medical program [36]. In summary, it is apparent from the examined documents that the regulations on pass and grade determination are heterogeneous. In some locations, detailed rules have been formulated, but many regulations handle pass/fail and grades rather superficially, which of course does not mean that decisions are not made very consciously in the local reality. Interpreting the documents was not always easy, and there may be passages that have been misinterpreted. Likewise, due to the selection of the studied faculties, interesting approaches may have been overlooked. For instance, dental and veterinary regulations were not analysed for this work. Still, our findings coincide well with those of Härtl et al., who conducted a survey on the assessment of communication skills at German-speaking medical schools. From 31 OSCEs with communication as test objective, 21 pegged the pass boundary to a fixed score or percentage, 5 applied the borderline regression method, 2 used Angoff’s method, and in 6 OSCEs, the method used was unknown (some multiple responses) [37]. On behalf of the committee we recommend to question the local regulations used to determine pass/fail and scoring criteria for practical exams, to seek the discourse, and to share and discuss experiences with the scientific community. Procedures used to define pass criteria and scoring should be based on examination content. This would add to the validity of pass/fail decisions of practical exams and better ensure fairness to examinees. The members of the GMA committee on assessment are gladly available for this discourse as discussion partners. Competing interests The authors declare that they have no competing interests. ==== Refs 1 Bloch R Bürgi H The Swiss catalogue of learning objectives Med Teach 2002 24 2 144 150 10.1080/01421590220120759 Available from: http://dx.doi.org/10.1080/01421590220120759 12098433 2 Schnabel K Boldt P Breuer G Fichtner A Karsten G Kujumdshiev S Schmidts M Stosch C Konsensusstatement "Praktische Fertigkeiten im Medizinstudium" - ein Positionspapier des GMA-Ausschusses für praktische Fertigkeiten GMS Z Med Ausbild 2011 28 4 Doc58 10.3205/zma000770 Available from: http://dx.doi.org/10.3205/zma000770 22205916 3 MFT Medizinischer Fakultätentag der Bundesrepublik Deutschland e. V. Nationaler Kompetenzbasierter Lernzielkatalog Medizin (NKLM) 2015 Berlin MFT Medizinischer Fakultätentag der Bundesrepublik Deutschland e. V 4 Möltner A Duelli R Resch F Schultz J Jünger J Fakultätsinterne Prüfungen an den deutschen medizinischen Fakultäten GMS Z Med Ausbild 2010 27 3 Doc44 10.3205/zma000681 Available from: http://dx.doi.org/10.3205/zma000681 5 Guttormsen S Beyeler C Bonvin R Feller S Schirlo C Schnabel K Schurten T Berendonk C The new licensing examination for human medicine: from concept to implementation Swiss Med Wkly 2013 143 w13897 24301323 6 Schickler A Brüstle P Biller S Mündlich-praktischer Teil des Zweiten Abschnitts der Ärztlichen Prüfung in Freiburg 2012 - Analyse der Notenvergabe zur Überprüfung von Qualitätssicherungsmaßnahmen GMS Z Med Ausbild 2015 32 4 Doc39 10.3205/zma000981 Available from: http://dx.doi.org/10.3205/zma000981 26483852 7 Boeckers A Fassnacht U Feneberg A Boeckers T Modifizierte objective structured practical examination (OSPE) als Leistungskontrolle im Kurs der Mikroskopischen Anatomie an der Universität Ulm GMS Z Med Ausbild 2006 23 4 Doc71 Available from: http://www.egms.de/static/de/journals/zma/2006-23/zma000290.shtml 8 Yudkowsky R Downing S Yudkowsky R Performance Tests Assessment in Health Professions Education 2009 New York Routledge 9 Tekian A Norcini J Overcome the 60% passing score and improve the quality of assessment GMS Z Med Ausbild 2015 32 4 Doc43 10.3205/zma000985 Available from: http://dx.doi.org/10.3205/zma000985 26483856 10 Friedman Ben-David M AMEE Guide No. 18: Standard setting in student assessment Med Teach 2000 22 2 120 130 10.1080/01421590078526 Available from: http://dx.doi.org/10.1080/01421590078526 11 Boursicot K Roberts T Pell G Standard setting for clinical competence at graduation from medical school: a comparison of passing scores across five medical schools Adv Health Sc Educ 2006 11 2 173 183 10.1007/s10459-005-5291-8 Available from: http://dx.doi.org/10.1007/s10459-005-5291-8 12 Prüfungsordnung des Modellstudiengang Medizin der Charité-Universitätsmedizin Berlin in der Fassung vom 13.04.2015 01.12.2015 Berlin Charité-Universitätsmedizin Berlin Available from: http://www.charite.de/fileadmin/user_upload/portal/charite/presse/publikationen/amtl-mitteilungsblatt/2015/AMB150417-144.pdf 13 Studien- und Prüfungsordnung für den Modellstudiengang Medizin der Rheinisch-Westfälischen Technischen Hochschule Aachen mit dem Abschluss "Ärztliche Prüfung" vom 05.11.2008 in der Fassung der Dritten Ordnung zur Änderung der Studien- und Prüfungsordnung vom 25.10.2013 veröffentlicht als Gesamtfassung. 01.12.2015 Aachen Rheinisch-Westfälisch Technische Hochschule Aachen Available from: http://www.rwth-aachen.de/global/show_document.asp?id=aaaaaaaaaagdhjq 14 Studienordnung für den Studiengang Humanmedizin an der Ernst Moritz Arndt Universität Greifswald vom 26. August 2004 01.12.2015 Greifswald Ernst Moritz Arndt Universität Greifswald Available from: http://www.uni-greifswald.de/fileadmin/mp/e_dez4/zpa/PO/Staatsexamen/Humanmedizin/Studienordnung_Humanmedizin_Lesefassung-13-AEndS.pdf 15 Studienordnung für das Medizinstudium an der Medizinischen Fakultät Heidelberg der Universität Heidelberg für das 3, 4. und 5. Studienjahr vom 22. 07.2010 01.12.2015 Heidelberg Medizinische Fakultät Heidelberg Available from: http://www.uni-heidelberg.de/md/studium/download/03-02-7a-05.pdf 16 Studienordnung für den Studiengang Medizin an der Universität Leipzig vom 8.Mai 2012 Leipzig Universität Leipzig 17 Studienordnung für den Studiengang Humanmedizin an der Medizinischen Fakultät Mannheim der Universität Heidelberg vom 16. Mai 2013 01.12.2015 Mannheim Medizinische Fakultät Mannheim der Universität Heidelberg Available from: http://www.umm.uni-heidelberg.de/studium/studienordnung.pdf 18 Prüfungsordnung des Modellstudiengangs Medizin an der Medizinischen Fakultät der Universität Hamburg vom 20. 06.2012 01.12.2015 Hamburg Medizinische Fakultät der Universität Hamburg Available from: https://www.uni-hamburg.de/campuscenter/studienorganisation/ordnungen-satzungen/pruefungs-studienordnungen/medizin/up-po-u-fak-u-med-u-modellstud-u-medizin-u-20150715.pdf 19 Studienordnung für den Studiengang Medizin an der Universität des Saarlandes vom 20. Februar 2003 01.12.2015 Homburg Universität des Saarlandes Available from: http://www.uni-saarland.de/fileadmin/user_upload/Campus/Service/Recht_und_Datenschutz/Recht_der_Universitaet/Ausbildungs-_Pruefungs-_Studienordnungen/DB03-106.pdf 20 Prüfungsordnung für den Studiengang Medizin an der Medizinischen Hochschule Hannover in der Fassung vom 15. 07.2015 01.12.2015 Hannover 07.2015 [Internet]. Medizinische Hochschule Hannover. https Available from: https://www.mh-hannover.de/fileadmin/mhh/download/studium_ausbildung/Medizin/Ordnungen/2015_07_15_SudOPruefO_Medizin_MHH.pdf 21 Technische Universität Dresden Medizinische Fakultät Carl Gustav Carus, Studienordnung für den Studiengang Medizin vom 26. 05.2010 01.12.2015 Dresden Technische Universität Dresden Available from: http://www.verw.tu-dresden.de/AmtBek/PDF-Dateien/2011-05/25ZsoSE08.09.2011.pdf 22 Nichtamtliche Lesefassung der Studienordnung für das Studium der Humanmedizin an der Universität Rostock unter Berücksichtigung der 1, 2. und 3. Satzung zur Änderung der Studienordnung vom Februar 2009 01.12.2015 Rostock Universität Rostock Available from: http://www.med.uni-rostock.de/fileadmin/Verwaltung/studiendekanat/humanmedizin/downloads/Lesefassung_Sto_Humanmedizin_2009.pdf 23 Studienordnung (StO) für den Modellstudiengang Humanmedizin an der Medizinischen Fakultät der Universität zu Köln mit dem Abschluss "Ärztliche Prüfung" vom 06. Januar 2014 01.12.2015 Köln Universität zu Köln Available from: http://am.uni-koeln.de/e10890/am_mitteilungen/@1/AM_02_2014.pdf 24 Studienordnung für den Studiengang Humanmedizin mit dem Abschluss "Ärztliche Prüfung" am Fachbereich Medizin der Justus-Liebig-Universität Gießen vom 03.Juli 2006 01.12.2015 Gießen Justus-Liebig-Universität Gießen Available from: https://www.uni-giessen.de/mug/6/pdf/6_60_11_2_neu.pdf 25 Studienordnung für den Studiengang Medizin an der Martin-Luther-Universität Halle-Wittenberg vom 21.04.2009 01.12.2015 Halle Martin-Luther-Universität Halle-Wittenberg Available from: http://www.verwaltung.uni-halle.de/KANZLER/ZGST/ABL/2009/09_08_01.pdf 26 Prüfungsordnung für den Modellstudiengang Humanmedizin an der Carl von Ossietzky Universität Oldenburg vom 24.02.2012 01.12.2015 Oldenburg Carl von Ossietzky Universität Oldenburg Available from: https://www.uni-oldenburg.de/uni/amtliche_mitteilungen/dateien/AM2013-03_09-Neufassung_EMS_PO_FKVI.pdf 27 Prüfungsordnung für den Brandenburger Modellstudiengang Medizin an der Medizinischen Hochschule Brandenburg Theodor Fontane [Internet] 01.12.2015 Neuruppin Medizinische Hochschule Brandenburg Theodor Fontane Available from: http://www.mhb-fontane.de/files/Dateien/Studiengang%20Humanmedizin/Pruefungsordnung_BMM_2015.pdf 28 Prüfungsordnung der Ruhr-Universität Bochum für den Modellstudiengang Medizin vom 20.April 2005 01.12.2015 Bochum Ruhr-Universität Bochum Available from: http://medizinstudium.ruhr-uni-bochum.de/medidek/infoszumstudium/amtliches/pruefungsordnungordnung_MSM_2005-04-20.pdf 29 Prüfungs- und Studienordnung für den Studiengang Medizin an der Ludwig-Maximilians-Universität München in der Fassung vom 24.09.2009 01.12.2015 München Ludwig-Maximilians-Universität München Available from: http://www.mecum-online.de/downloads/studienrecht/ordnungen/so_medizin24_11_09.pdf 30 Studien- und Prüfungsordnung für den Modellstudiengang Humanmedizin an der Heinrich-Heine-Universität Düsseldorf vom 7.10.2013 01.12.2015 Düsseldorf Heinrich-Heine-Universität Düsseldorf Available from: http://www.medizin.hhu.de/fileadmin/redaktion/Fakultaeten/Medizinische_Fakultaet/Studiendekanat/Dokumente/Medizin_Studium/131007_Studien-_und_Pruefungsordnung.pdf 31 Studienordnung für den Studiengang Humanmedizin mit dem Abschluss "Ärztliche Prüfung" an der Philipps-Universität Marburg vom 23.09.2015 01.12.2015 Marburg Philipps-Universität Marburg Available from: https://www.uni-marburg.de/administration/amtlich/60_2015.pdf 32 Information zur Bestehensgrenze und zur Notengebung bei den Gesamtprüfungen (SIP) für Humanmediziner (N202) und Zahnmediziner (N203, 1 u. 2. Studienjahr 01.12.2015 Wien Medizinische Universität Wien Available from: https://www.meduniwien.ac.at/homepage/fileadmin/HP-Relaunch/pdfstudien/HumanmedizinN202/allg-info-bestehensgrenzen-20101001.pdf 33 Beurteilungskriterien des OSCE 01.12.2015 Graz Medizinische Universität Graz Available from: https://www.medunigraz.at/themen-studieren/humanmedizin/osce-objective-structured-clinical-examination/beurteilungskriterien-des-osce/ 34 Information Famulatur OSCE 01.12.2015 Innsbruck Medizinische Universität Innsbruck Available from: https://www.i-med.ac.at/studium/studierende/docs/Information-OSCE_13-14-fin.pdf 35 Berendonk C Schirlo C Balestra G Bonvin R Feller S Huber P Jünger E Monti M Schnabel K Beyeler C Guttormsen S Huwendiek S The new final Clinical Skills examination in human medicine in Switzerland: Essential steps of exam development, implementation and evaluation, and central insights from the perspective of the national Working Group GMS Z Med Ausbild 2015 32 4 Doc40 10.3205/zma000982 Available from: http://dx.doi.org/10.3205/zma000982 26483853 36 Prüfungen 01.12.2015 Basel Universität Basel Available from: https://medizinstudium.unibas.ch/allgemeine-infos/pruefungen.html 37 Härtl A Bachmann C Blum K Höfer S Peters T Preusche I Raski B Rüttermann S Wagner-Menghin M Wünsch A Kiessling C Wunsch und Wirklichkeit - eine Umfrage im deutschsprachigen Raum zum Lehren und Prüfen kommunikativer Kompetenzen im Medizinstudium GMS Z Med Ausbild 2015 32 5 Doc56 10.3205/zma000998 Available from: http://dx.doi.org/10.3205/zma000998 26604998

PMC005xxxxxx/PMC5003129.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00105710.3205/zma001057Doc58urn:nbn:de:0183-zma0010578ArticleHow well do final year undergraduate medical students master practical clinical skills? Wie gut beherrschen Studierende im Praktischen Jahr klinisch-praktische Fertigkeiten? Störmann Sylvère *1Stankiewicz Melanie 1Raes Patricia 1Berchtold Christina 1Kosanke Yvonne 1Illes Gabrielle 1Loose Peter 1Angstwurm Matthias W. 11 Klinikum der Universität München, Medizinische Klinik und Poliklinik IV, München, Germany*To whom correspondence should be addressed: Sylvère Störmann, Klinikum der Universität München, Medizinische Klinik und Poliklinik IV, Ziemssenstraße 1, D-80336 München, Germany, Phone: +49 (0)89/4400-52318, Fax: +49 (0)89/4400-52192, E-mail: sylvere.stoermann@med.uni-muenchen.de15 8 2016 2016 33 4 Clinical skillsDoc5830 9 2015 10 5 2016 11 3 2016 Copyright © 2016 Störmann et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001057.shtmlIntroduction: The clinical examination and other practical clinical skills are fundamental to guide diagnosis and therapy. The teaching of such practical skills has gained significance through legislative changes and adjustments of the curricula of medical schools in Germany. We sought to find out how well final year undergraduate medical students master practical clinical skills. Methods: We conducted a formative 4-station objective structured clinical examination (OSCE) focused on practical clinical skills during the final year of undergraduate medical education. Participation was voluntary. Besides the examination of heart, lungs, abdomen, vascular system, lymphatic system as well as the neurological, endocrinological or orthopaedic examination we assessed other basic clinical skills (e.g. interpretation of an ECG, reading a chest X-ray). Participants filled-out a questionnaire prior to the exam, inter alia to give an estimate of their performance. Results: 214 final year students participated in our study and achieved a mean score of 72.8% of the total score obtainable. 9.3% of participants (n=20) scored insufficiently (<60%). We found no influence of sex, prior training in healthcare or place of study on performance. Only one third of the students correctly estimated their performance (35.3%), whereas 30.0% and 18.8% over-estimated their performance by 10% and 20% respectively. Discussion: Final year undergraduate medical students demonstrate considerable deficits performing practical clinical skills in the context of a formative assessment. Half of the students over-estimate their own performance. We recommend an institutionalised and frequent assessment of practical clinical skills during undergraduate medical education, especially in the final year. Zusammenfassung Einleitung: Die körperliche Untersuchung und klinisch-praktische Fertigkeiten sind wesentliche ärztliche Fähigkeiten, mit deren Hilfe die Diagnostik und Therapie von Krankheiten gesteuert werden. Durch curriculare Veränderungen wird der praktischen Ausbildung ein hoher Stellenwert eingeräumt. Wie gut beherrschen also Studierende im Praktischen Jahr (PJ) klinisch-praktische Fertigkeiten? Methoden: Wir führten eine freiwillige mündlich-praktische Prüfung mittels OSCE bei Studierenden im PJ durch. Inhalte der Prüfung waren u.a. die körperliche Untersuchung (Herz, Lunge, Abdomens, Gefäßsystem, Lymphsystem; neurologische, endokrinologische bzw. orthopädische Untersuchung) sowie elementare praktische Fertigkeiten (etwa EKG-Interpretation, Basis-Befundung Röntgen-Thorax). Die Teilnehmer füllten zudem vor Beginn der Prüfung einen Fragebogen aus, u.a. zur Einschätzung der eigenen Leistung. Ergebnisse: Insgesamt 214 PJ-Studierende nahmen teil und erreichten 72,8% der erreichbaren Punktzahl. Eine nicht ausreichende Leistung (<60%) zeigten 9,3% der Teilnehmer (n=20). Geschlecht, vorangegangene Ausbildung in einem Gesundheitsberuf sowie Studienort hatten keinen Einfluss auf die Leistung. Im Mittel schätzten sich die Studierenden 0,5 Notenstufen besser. 35,3% der Teilnehmer vermochten ihre Leistung richtig einzuschätzen. 30,0% überschätzten ihr Ergebnis um eine Notenstufe, 18,8% um zwei oder mehr Notenstufen. Diskussion: Studierende im Praktischen Jahr zeigen deutliche Defizite bei der Durchführung klinisch-praktischer Fertigkeiten im Rahmen einer mündlich-praktischen Prüfung. Dabei überschätzt knapp die Hälfte der Studierenden die eigene Leistung. Eine institutionalisierte, regelhafte Prüfung der mündlich-praktischen Fähigkeiten im Praktischen Jahr erscheint daher notwendig. practical examinationOSCEphysical examinationclinical skills ==== Body 1. Introduction The basic clinical examination is a fundamental skill of physicians that facilitates diagnosis and therapy [1], [2], [3]. Technical progress has changed the practice of medicine, which increasingly relies on laboratory assessments, diagnostic imaging, and other sources of technical examination. Some authors mourn that this has led to a deterioration of the ability to perform a systematic and focused hands-on clinical examination [4], [5]. Medical students and young physicians alike demonstrate deficits in performing a clinical examination [6], [7]. A legislative reform in Germany in 2002 (Medical Licensure Act) pushed the faculties to develop new curricula with an emphasis on practical training [8], [9], [10], [11], [12], [13], [14]. By specifically defining the type and amount, this law formalised clinical practical teaching and its curricular design [15]. This lead to novel teaching concepts as well as whole curricula centred on medical skills [16], [17], [18], [19], [20]. An intensive training of practical clinical skills during the first years of undergraduate medical education aims to prepare students well for their future role as physicians; the final year is an important landmark in undergraduate training and consists mainly of practical exercise of previously and newly acquired skills [21]. The Medical Curriculum (MeCuM) integrates clinical training into pre-clinical courses during year 2 of undergraduate studies, starting with history taking. Through a variety of diverse formats (lectures, bedside teaching, peer teaching, and blended learning) students learn the theoretical basis of the clinical examination and have the opportunity to put it into practice. Longitudinal internships in general medical practices and frequent bedside courses allow practical exercise and feedback discussions steer the learning. As students progress, further practical clinical skills are taught in the context of their respective system (such as writing and reading an ECG as part of the cardiovascular teaching block, interpretation of a chest X-ray during the respiratory teaching block). Higher-level skills such as clinical decision-making are part of the formation during the final year when students should be proficient in the practice of basic clinical skills. However, a comparison of the performance in the Licensure Examination before and after the reform at both medical faculties in Munich (LMU und TU) showed a statistically significant decline of scores in the oral and practical part of the exam [22]. Changes in medical curricula in Germany and their impact on the increase in medical knowledge during clinical training are well studied [23], [24]; investigations of learned practical skills and achievement of competence based learning objectives are lacking. It is therefore unclear, how well undergraduate medical students receiving training more focused on clinical skills effectively master these skills in their final year. Furthermore, the adequate self-assessment of performance and therefore one’s own limitations play a crucial role in the care for patients. Everyone involved in patient care should seek help in case of overload [25]. It is important for every (aspiring) physician to recognise limits of one’s own abilities and to prevent harm through erroneous action or even faulty omission. Danger lies in over-estimation (unconscious incompetence) as well as under-estimation (unconscious competence) of one’s abilities [26]. Multiple studies have shown that subjective self-assessment and objectively measured performance do not necessarily correlate [27], [28], [29]. Undergraduate medical students have a responsibility towards their patients as well as their teachers to estimate their skills adequately in order to improve on deficits and further develop strengths. This holds especially true as the physical examination acts as a cornerstone of diagnosis [2], [30]. Typically, final year medical students in Germany are first to see admitted patients. Therefore, a realistic self-assessment of examination proficiency is vital for the patients’ well-being and further course of hospital stay. We wanted to know how well undergraduate students in an advanced and critical part of their training could estimate their abilities to perform basic clinical skills. 2. Methods Undergraduate medical students in their final year could participate in a formative oral and practical examination (“mündlich-praktische Prüfung im PJ”, abbreviated: mP3) using the OSCE format (objective structured clinical examination) from mid-2011 through 2014. The intent of this examination was to offer the participants the possibility to objectively assess practical clinical skills and obtain individual feedback as to identify strengths and weaknesses. The examination consisted of four OSCE stations. The stations covered various aspects of the physical examination: heart, lungs, abdomen, and vascular/lymphatic system as well as neurological, endocrinological, and orthopaedic examination. Amongst others, the stations covered basic clinical skills such as writing and reading of a 12-lead ECG, basic interpretation of an abdominal CT scan, identifying normal and abnormal findings on a chest X-ray, outlining the management of an emergency in the ER, enumerate important laboratory parameters to aid differential diagnosis in specified clinical settings). Two thirds of each station is devoted to the physical examination, the remaining third assesses other clinical skills. Each instalment of the OSCE consisted of stations compiled from a pool of 12 different OSCE stations. An expert panel designed and validated all stations. Participants performed the physical examination on probands instructed not to give any feedback during the examination. Marks for specific steps of the physical examination were awarded only if that step was performed correctly in its entirety. Each station lasted precisely 12 minutes. Afterwards, students obtained 3 minutes feedback from the examiner. All 19 examiners were faculty staff members with experience in examining OSCEs as well as professional experience as clinicians. Frequently held workshops for faculty by our Institute for Didactics and Education Research ensure a high standard of quality in the implantation of assessments such as the OSCE. Participants voluntarily filled-out a questionnaire referring to personal and demographic details, the course of studies, prior training in healthcare (e.g. as paramedic or as nurse) as well as the assumed mark achieved in the examination (5 point scale as commonly used in Germany for school grading: exam mark 1 = “excellent” to 5 = “insufficient”). To allow for comparison of the self-assessment with the OSCE score (expressed as percentage of total achievable points), we converted the OSCE score into the same 5 point scale according to a conversion scheme common in Germany and used in the National License Examinations [8]. Students received a notification of their achieved score after the examination. All statistical analyses were performed using SPSS (IBM Corp., Armonk, NY, U.S.A.). For the difference between two means, t-tests were used; in cases with multiple groups, an analysis of variance was performed. Effect sizes were assessed using Cohen’s d. p values α=0.05 were considered statistically significant. The operational sequence, purpose, and intention of scientific interpretation of the data of this practical examination were announced to the local ethics committee, which deemed a formal ethical approval not necessary. The study was conducted according to principles of the World Medical Association’s Declaration of Helsinki and Declaration of Geneva. All undergraduate medical students in their final year could participate. Participation was voluntary and participants gave written consent to the scientific analysis of the examination and publication of results. Not consenting did not exclude students from the examination. 3. Results Study population 214 students participated in the study from mid-2011 until the end of 2014. Median age of participants was 26.3 (±4.5) years. Almost two thirds (64.0%; n=137) were female, thus corresponding to the gender distribution of all undergraduate medical students at the LMU Munich. There was no significant age difference between female and male participants (m=27.5±3.2 years; f=27.1±5.1 years; p=0.544). Most participants (n=156; 72.9%) had pursued their medical studies at the LMU Munich from the beginning; the other participants had joined the LMU Munich at later stages of their undergraduate studies. Total performance On average participants achieved 72.8%±10.1% of the maximum total score (see Figure 1 (Fig. 1)). After converting the performance score into a 5 point scale only 3.7% achieved an examination mark of “1” (“excellent”; score=90%; n=8), 23.4% a mark of “2” (“good”; score 80-90%; n=50), 34.6 a mark of “3” (“fair”; score 70-80%; n=74), and 29.0% a mark of “4” (“poor”; score 60-70%; n=62). Twenty students (9.3%) had an “insufficient” score (defined as<60%; mark of “5”). Confounding factors Female participants had a tendency towards slightly higher scores; however, this difference was not significant (73.7%±10.1% versus 71.1%±9.9%; p=0.069). There were no significant differences in scores between participants who had studied at the LMU Munich from the beginning vs. at later stages (p=0.349). A prior training in healthcare did not yield other scores than without prior training (p=0.363). Scores were homogenously distributed amongst participants from 2011 until 2014 (p=0.881). The majority of participants stated not having prepared themselves specifically for the exam (63.1%). They achieved significantly lower scores in comparison to prepared students (71.8%±9.3% vs. 78.6%±10,0%; p<0,001; d=0,27). For an overview of these results, cf. Table 1 (Tab. 1). Self-assessment 170 participants (79.4%) gave an estimate of their performance in the examination. Self-assessed performance and total examination score correlated positively and significantly (r=0.26; p<0.001). On average, students over-estimated their performance by half an examination mark. 60 participants (35.3%) correctly assessed their performance. 51 students (30.0%) over-estimated their performance by one, 32 participants (18.8%) by two marks. 21 students (12.4%) under-estimated their performance by one, 6 participants (3.5%) by two marks. Of the 20 participants with a total score below 60% (“insufficient”) 16 had self-assessed their performance of which 13 (81.3%) were over-estimating. Confer to Figure 2 (Fig. 2) for an overview of self-assessment in relation to total score. 4. Discussion Practical clinical skills such as the physical examination remain an important instrument in the physician’s armamentarium. Our analysis of a formative, oral-practical examination in undergraduate medical students in their final year showed a lack of these skills despite the advanced course of studies and immanent licensure. Our participants had trouble performing a physical examination as well as basic clinical procedures such as writing and reading an ECG. A comparable analysis in American students during the USMLE Step 2 Clinical Skills Examination yielded similar results [31]. Recently Schmidmaier et al. used a progress test to show that knowledge of internal medicine continuously increases at the LMU Munich [24]. However, these results are not generalizable onto practical clinical skills [32]. In Germany acquiring new and improving on existing skills during the final year of undergraduate medical studies relies heavily on the supervision and patronage of the ward’s physicians where students spend their final year. In practice, supervision is lacking and the acquisition and improvement of skills depends largely on chance and the individual commitment of the students [33], [34], [35]. A rather new approach is to follow the development of clinical skills with a progress test longitudinally [36], [37], [38]; so far, published data are lacking. Learning practical clinical skills requires complex interventions and a seamless interaction between all parties involved (medical faculty, teaching hospitals, and other hospitals/practices where students complete clinical traineeships). In reality, this is hardly controllable and students develop a large part of their “clinical practice” outside class [39]. Effectively this means that an important part of medical training is beyond the grasp of university structures and therefore escapes institutional quality standards. So far, teaching of practical clinical skills at faculty level focuses on the use of skills labs [40], [41] where peers mostly perform teaching (student tutors). Multiple studies have shown this concept to be effective [42], [43]. Structured formats improve practical clinical skills acquired in the skills lab lastingly [44]. Another mechanism is to perform intermittent formative examinations and make use of the “assessment drives learning” effect [45]. In respect to the data presented herein, it seems important to perform formative examinations assessing clinical skills as measures of quality assurance during the final year of undergraduate medical education [46]. Ideally, these examinations should be composed of assessments of diverse skills and compiled from an exhaustive catalogue [47]. In light of high costs of these examination formats [48], [49], faculties have to rethink how to find affordable solutions to improve the teaching of clinical skills, such as special tutorship programmes [50]. Important questions in this context are: Who profits from such formative examinations? How frequent should they be? At what point in time during the course of medical studies should the first examination take place? Is the OSCE the right format for such examinations? There are no answers to these questions derived from generalizable recommendations from the literature. To select specific students (and therefore to favour those) might seem ethically ill advised. However, additional interventions have proven helpful in those students at risk of attrition [51], [52]. From an economical stance, it could be justified to limit additional resources to those students at risk. A possible compromise (albeit increasing administrative overhead) could be to offer a certain minimum of such examinations to all students and to examine students at risk more frequently. How often and from which year of undergraduate education on is unclear as data is scarce [53]. We think that such examinations should begin early on to prevent giving feedback too late, i.e. when false manoeuvres have already become routine. Alternatives are other formats such as the Mini-CEX and others (CEC, DOPS) that can take place directly at the “workplace”. As such, they offer interesting possibilities to assess clinical skills intermittently with comparatively little à priori effort [54], [55], [56]. Conversely, more effort is required to instruct and train examiners correctly for these formats. One third of all students correctly self-assessed their performance. Almost half of our participants over-estimated their performance; nearly one in five to a vast degree. This is more than previously described [29], [57]. The phenomenon is not new and neither limited to medical studies nor students per se [58], [59]. Our students received structured and qualitative feedback after the exanimation. Many students were surprised when they realized how off they were in their self-assessment. Consequences of over-estimation can be serious, in particular when it leads to diagnostic and/or therapeutic errors [60], [61], [62]. The ability to correctly self-evaluate is difficult to teach or train. It was postulated that video feedback would be sufficient to improve self-assessment [63]. A study by Hawkins and colleagues achieved improvement only retrospectively when a video of the students’ performance was juxtaposed a video demonstrating the correct manoeuvres [64]. It is therefore important to give the students a good idea of their performance through structured feedback, but also to show the correct execution of the skill assessed. Criticism must always be communicated in a meaningful way and it has to be noted that responsiveness to feedback is modulated by expectations and attitudes [65]. We presented data from a formative and voluntary examination, for which students had to register actively. This may have biased our results. One would expect eager and especially motivated students to participate in an optional examination leading to false-positive results. In so far we deem our data and the conclusions derived from them to be plausible. Our examination is well accepted and students use it to prepare themselves for the final part of the Licensure Examination, which also takes place as an oral and practical exam. A strength of our data is the size of the study population that allows for reliable statements even when effect size is small. 5. Conclusions The performance of undergraduate medical students in their final year during a formative oral and practical clinical examination leaves ample room for improvement. Almost two thirds of our participants scored “fairly” or “poorly”; one in ten students fails. Practically half of our students over-estimated their own performance. An established, standardized, formative examination during the final year of undergraduate studies seems vital. Acknowledgements We thank our examiners and colleagues for their support and all the efforts put into implementing the mP3 examination: Prof. Dr. Ralf Schmidmaier, PD Dr. Peter Reilich, PD Dr. Stefan Grote, PD Dr. Peter Thaller, PD Dr. Philipp Baumann, Dr. Kathrin Schrödl, Dr. Philip von der Borch, Dr. Bert Urban, Dr. Michael Maier, Dr. Mark op den Winkel, Dr. Costanza Chiapponi, Anja Fischer, Caroline Strenkert, Andreas Sturm, and Christina Berr. Competing interests The authors declare that they have no competing interests. Table 1 Possible confounding factors: comparison of total score between various sub-groups Figure 1 Histogram of total score (4-station OSCEs) Figure 2 Radar chart of total examination performance (expressed as exam mark) and average self-assessment (same scale) ==== Refs 1 Kirch W Schafii C Misdiagnosis at a university hospital in 4 medical eras Medicine (Baltimore) 1996 75 1 29–40 10.1097/00005792-199601000-00004 Available from: http://dx.doi.org/10.1097/00005792-199601000-00004 8569468 2 Reilly BM Physical examination in the care of medical inpatients: an observational study Lancet 2003 362 9390 1100–1105 10.1016/S0140-6736(03)14464-9 Available from: http://dx.doi.org/10.1016/S0140-6736(03)14464-9 14550696 3 Kugler J Verghese A The physical exam and other forms of fiction J Gen Intern Med 2010 25 8 756–757 10.1007/s11606-010-1400-3 Available from: http://dx.doi.org/10.1007/s11606-010-1400-3 20502975 4 Feddock CA The lost art of clinical skills Am J Med 2007 120 4 374–378 10.1016/j.amjmed.2007.01.023 Available from: http://dx.doi.org/10.1016/j.amjmed.2007.01.023 17398236 5 Cook C The lost art of the clinical examination: an overemphasis on clinical special tests J Man Manip Ther 2010 18 1 3–4 10.1179/106698110X12595770849362 Available from: http://dx.doi.org/10.1179/106698110X12595770849362 21655416 6 Wilkerson L Lee M Assessing physical examination skills of senior medical students: knowing how versus knowing when Acad Med 2003 78 10 Suppl S30–S32 14557088 7 Ramani S Ring BN Lowe R Hunter D A pilot study assessing knowledge of clinical signs and physical examination skills in incoming medicine residents J Grad Med Educ 2010 2 2 232–235 10.4300/JGME-D-09-00107.1 Available from: http://dx.doi.org/10.4300/JGME-D-09-00107.1 21975626 8 Bundesministerium für Gesundheit Approbationsordnung für Ärzte 2002 Berlin Bundesministerium für Gesundheit 9 Eitel F Die neue Approbationsordnung verlangt tief greifende Änderungen in der Lehrorganisation Gesundheitswesen (Suppl Med Ausbild) 2002 19 Suppl1 1 2 Available from: https://gesellschaft-medizinische-ausbildung.org/files/ZMA-Archiv/2002/1/Eitel_F-ae.pdf 10 Steiner T Schmidt J Bardenheuer H Kirschfink M Kadmon M Schneider G Seller H Sonntag HG HEICUMED Planungsgruppe HEICUMED?: Heidelberger Curriculum Medicinale - Ein modularer Reformstudiengang zur Umsetzung der neuen Approbationsordnung Gesundheitswesen (Suppl Med Ausbild) 2003 20 Suppl2 87 91 Available from: https://gesellschaft-medizinische-ausbildung.org/files/ZMA-Archiv/2003/2/Steiner_T,_J%C3%BCnger_J,_Schmidt_J,_Bardenheuer_H,_Kirschfink_M,_Kadmon_M,_Schneider_G,_Seller_H,_Sonntag_HG.pdf 11 Schulze J Drolshagen S Nürnberger F Ochsendorf F Gestaltung des klinischen Studiums nach den Vorgaben der neuen Ärztlichen Approbationsordnung – Struktur und Organisation Gesundheitswesen (Suppl Med Ausbild) 2003 20 Suppl2 68 77 Available from: https://gesellschaft-medizinische-ausbildung.org/files/ZMA-Archiv/2003/2/Schulze_J,_Drolshagen_S,_N%C3%BCrnberger_F,_Ochsendorf_F.pdf 12 van den Bussche H Lehren und Lernen am UKE - Die Umsetzung der Approbationsordnung für Ärzte in Hamburg Z Allgemeinmed 2004 80 431–437 10.1055/s-2004-820414 Available from: http://dx.doi.org/10.1055/s-2004-820414 13 Huwendiek S Kadmon M Jünger J Kirschfink M Bosse HM Resch F Duelli R Bardenhauer HJ Sonntag HG Steiner T Umsetzung der deutschen Approbationsordnung 2002 im modularen Reformstudiengang Heidel- berger Curriculum Medicinale (HeiCuMed) ZfHE 2008 3 3 17 27 Available from: http://www.zfhe.at/index.php/zfhe/article/view/70/75 14 Zeuner S Henke T Achilles E Kampmeyer D Schwanitz P 36 Different Ways To Study Medicine GMS Z Med Ausbild 2010 27 2 Doc20 10.3205/zma000657 Available from: http://dx.doi.org/10.3205/zma000657 21818189 15 Haage H Das neue Medizinstudium 2003 Aachen Shaker Verlag 320 16 Fischer T Simmenroth-Nayda A Herrmann-Lingen C Wetzel D Chenot JF Kleiber C Staats H Kochen MM Medizinische Basisfähigkeiten – ein Unterrichtskonzept im Rahmen der neuen Approbationsordnung Z All Med 2003 79 432 436 Available from: https://www.online-zfa.de/media/archive/2003/09/10.1055-s-2003-43063.pdf 17 Öchsner W Forster J Approbierte Ärzte - kompetente Ärzte? Die neue Approbationsordnung für Ärzte als Grundlage für kompetenzbasierte Curricula GMS Z Med Ausbild 2005 22 1 Doc04 Available from: http://www.egms.de/static/de/journals/zma/2005-22/zma000004.shtml 18 Kulike K Hilgers J Störmann S Hornung T Dudziak J Weinmann P Kerncurriculum für die Medizinische Ausbildung in Deutschland: Ein Vorschlag der Medizinstudierenden Deutschlands GMS Z Med Ausbild 2006 23 4 Doc58 Available from: http://www.egms.de/static/de/journals/zma/2006-23/zma000277.shtml 19 Schnabel KP Boldt PD Breuer G Fichtner A Karsten G Kujumdshiev S Schmidts M Stosch C Konsensusstatement "Praktische Fertigkeiten im Medizinstudium" – ein Positionspapier des GMA-Ausschusses für praktische Fertigkeiten GMS Z Med Ausbild 2011 28 4 Doc58 10.3205/zma000770 Available from: http://dx.doi.org/10.3205/zma000770 22205916 20 Fischer MR Bauer D Mohn K Endlich fertig! Nationale Kompetenzbasierte Lernzielkataloge Medizin (NKLM) und Zahnmedizin (NKLZ) gehen in die Erprobung GMS Z Med Ausbild 2015 32 3 Doc35 10.3205/zma000977 Available from: http://dx.doi.org/10.3205/zma000977 26677513 21 Jilg S Möltner A Berberat P Fischer MR Breckwoldt J How do Supervising Clinicians of a University Hospital and Associated Teaching Hospitals Rate the Relevance of the Key Competencies within the CanMEDS Roles Framework in Respect to Teaching in Clinical Clerkships?? GMS Z Med Ausbild 2015 32 3 Doc33 10.3205/zma000975 Available from: http://dx.doi.org/10.3205/zma000975 26413171 22 Seyfarth M Reincke M Seyfarth J Ring J Fischer MR Neue ärztliche Approbationsordnung und Notengebung beim Zweiten Staatsexamen Dtsch Arztebl Int 2010 107 28-29 500–504 20703334 23 Raupach T Vogel D Schiekirka S Keijsers C Ten Cate O Harendza S WissenszuwachsimPraktischen Jahr des Medizinstudiums in Deutschland GMS Z Med Ausbild 2013 30 3 Doc33 10.3205/zma000876 Available from: http://dx.doi.org/10.3205/zma000876 24062813 24 Schmidmaier R Holzer M Angstwurm M Nouns Z Reincke M Fischer MR Querschnittevaluation des Medizinischen Curriculums München (MeCuM) mit Hilfe des Progress Tests Medizin (PTM) GMS Z Med Ausbild 2010 27 5 Doc70 10.3205/zma000707 Available from: http://dx.doi.org/10.3205/zma000707 21818215 25 Happel O Papenfuß T Kranke P Schockraummanagement – Simulation, Teamtraining und Kommunikation für eine bessere Traumaversorgung TT - Training for real: simulation, team-training and communication to improve trauma management Anästhesiol Intensivmed Notfallmed Schmerzther 2010 45 06 408–415 10.1055/s-0030-1255348 Available from: http://dx.doi.org/10.1055/s-0030-1255348 20539968 26 Ingham H Luft J The Johari Window: a graphic model for interpersonal relations 1955 Los Angeles Proceedings of the western training laboratory in group development 27 Barnsley L Lyon PM Ralston SJ Hibbert EJ Cunningham I Gordon FC Clinical skills in junior medical officers: A comparison of self-reported confidence and observed competence Med Educ 2004 38 4 358–367 10.1046/j.1365-2923.2004.01773.x Available from: http://dx.doi.org/10.1046/j.1365-2923.2004.01773.x 15025637 28 Mavis B Self-efficacy and OSCE performance among second year medical students. Adv Heal Sci Educ 2001 6 2 93–102 10.1023/A:1011404132508 Available from: http://dx.doi.org/10.1023/A:1011404132508 29 Jünger J Schellberg D Nikendei C Subjektive Kompetenzeinschätzung von Studierenden und ihre Leistung im OSCE GMS Z Med Ausbild 2006 23 3 Doc51 Available from: http://www.egms.de/static/de/journals/zma/2006-23/zma000270.shtml 30 Verghese A A doctor's touch 2011 Edinburgh TEDGlobal2011 31 Peitzman SJ Cuddy MM Performance in Physical Examination on the USMLE Step 2 Clinical Skills Examination Acad Med 2015 90 2 209–213 10.1097/ACM.0000000000000570 Available from: http://dx.doi.org/10.1097/ACM.0000000000000570 25406608 32 Vukanovic-Criley JM Criley S Warde CM Boker JR Guevara-Matheus L Churchill WH Nelson WP Criley JM Competency in cardiac examination skills in medical students, trainees, physicians, and faculty: a multicenter study Arch Intern Med 2006 166 6 610–616 10.1001/archinte.166.6.610 Available from: http://dx.doi.org/10.1001/archinte.166.6.610 16567598 33 Engel C Porsche M Roth S Ganschow P Büchler MW Kadmon M Welche Ausbildungsinhalte brauchen Studierende im Praktischen Jahr - Bedarfsanalyse als Basis für ein chirurgisches PJ-Curriculum GMS Z Med Ausbild 2008 25 3 Doc89 Available from: http://www.egms.de/static/de/journals/zma/2008-25/zma000573.shtml 34 Remmen R Derese A Scherpbier A Denekens J Hermann I Van Der Vleuten C Can medical schools rely on clerkships to train students in basic clinical skills? Med Educ 1999 33 8 600–605 10.1046/j.1365-2923.1999.00467.x Available from: http://dx.doi.org/10.1046/j.1365-2923.1999.00467.x 10447847 35 Köhl-Hackert N Krautter M Andreesen S Hoffmann K Herzog W Jünger J Nikendei C Lernen am späteren Arbeitsplatz: eine Analyse studentischer Erwartungen an den Stationseinsatz im Fachbereich der Inneren Medizin GMS Z Med Ausbild 2014 31 4 Doc43 10.3205/zma000935 Available from: http://dx.doi.org/10.3205/zma000935 25489343 36 Pugh D Touchie C Wood TJ Humphrey-Murto S Progress testing: is there a role for the OSCE? Med Educ 2014 48 6 623–631 10.1111/medu.12423 Available from: http://dx.doi.org/10.1111/medu.12423 24807438 37 Pugh D Touchie C Humphrey-Murto S Wood TJ The OSCE progress test – Measuring clinical skill development over residency training Med Teach 2016 38 2 168 173 10.3109/0142159X.2015.1029895 Available from: http://dx.doi.org/10.3109/0142159X.2015.1029895 25909896 38 Gold J DeMuth R Mavis B Wagner D Progress testing 2.0: clinical skills meets necessary science Med Educ Online 2015 20 39 Duvivier RJ van Geel K van Dalen J Scherpbier AJ van der Vleuten CP Learning physical examination skills outside timetabled training sessions: what happens and why? Adv Health Sci Educ Theory Pract 2012 17 3 339–355 10.1007/s10459-011-9312-5 Available from: http://dx.doi.org/10.1007/s10459-011-9312-5 21710301 40 Blohm M Lauter J Branchereau S Krauter M Köhl-Hackert N Jünger J Herzog W Nikendei C "Peer-Assisted Learning" (PAL) im Skills-Lab – eine Bestandsaufnahme an den Medizinischen Fakultäten der Bundesrepublik Deutschland GMS Z Med Ausbild 2015 32 1 Doc10 10.3205/zma000952 Available from: http://dx.doi.org/10.3205/zma000952 25699102 41 Profanter C Perathoner A DOPS (Direct Observation of Procedural Skills) im studentischen Skills-Lab: Funktioniert das? Eine Analyse der Performanz klinischer Fertigkeiten und der curricularen Nebeneffekte GMS Z Med Ausbild 2015 32 4 Doc45 10.3205/zma000987 Available from: http://dx.doi.org/10.3205/zma000987 26483858 42 Yu TC Wilson NC Singh PP Lemanu DP Hawken SJ Hill AG Medical students-as-teachers: a systematic review of peer-assisted teaching during medical school Adv Med Educ Pract 2011 2 157–172 23745087 43 Weyrich P Celebi N Schrauth M Möltner A Lammerding-Köppel M Nikendei C Peer-assisted versus faculty staff-led skills laboratory training: a randomised controlled trial Med Educ 2009 43 2 113–120 10.1111/j.1365-2923.2008.03252.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2008.03252.x 19161480 44 Herrmann-Werner A Nikendei C Keifenheim K Bosse HM Lund F Wagner R Celebi N Zipfel S Weyrich P "Best Practice" Skills Lab Training vs. a "see one, do one" Approach in Undergraduate Medical Education: An RCT on Students' Long-Term Ability to Perform Procedural Clinical Skills PLoS One 2013 8 9 1–13 10.1371/journal.pone.0076354 Available from: http://dx.doi.org/10.1371/journal.pone.0076354 45 Buss B Krautter M Mo¨ltner A Weyrich P Werner A Ju¨nger J Nikendei C Der "Assessment Drives Learning"-Effekt beim Training klinisch-praktischer Fertigkeiten - Implikationen für die Curriculumsgestaltung und die Planung von Ressourcen Zusammenfassung GMS Z Med Ausbild 2012 29 5 Doc70 10.3205/zma000840 Available from: http://dx.doi.org/10.3205/zma000840 23255965 46 Raes P Angstwurm M Berberat P Kadmon M Rotgans J Streitlein-Böhme I Burckhardt G Fischer MR Qualitätsmanagement der klinisch-praktischenAusbildung im Praktischen Jahr des Medizinstudiums – Vorschlag eines Kriterienkatalogs der Gesellschaft für Medizinische Ausbildung GMS Z Med Ausbild 2014 31 4 Doc49 10.3205/zma00094 Available from: http://dx.doi.org/10.3205/zma00094 25489349 47 Bouwmans GA Denessen E Hettinga AM Michels C Postma CT Reliability and validity of an extended clinical examination Med Teach 2015 37 12 1072 1077 10.3109/0142159x.2015.1009423 Available from: http://dx.doi.org/10.3109/0142159x.2015.1009423 25683172 48 Brown C Ross S Cleland J Walsh K Money makes the (medical assessment) world go round: The cost of components of a summative final year Objective Structured Clinical Examination (OSCE) Med Teach 2015 37 7 653 659 10.3109/0142159x.2015.1033389 Available from: http://dx.doi.org/10.3109/0142159x.2015.1033389 49 Rau T Fegert J Liebhardt H Wie hoch liegen die Personalkosten für die Durchführung einer OSCE? Eine Kostenaufstellung nach betriebswirtschaftlichen Gesichtspunkten GMS Z Med Ausbild 2011 28 1 Doc13 10.3205/zma000725 Available from: http://dx.doi.org/10.3205/zma000725 21818228 50 Kraus B Briem S Jünger J Schrauth M Weyrich P Herzog W Zipfel S Nikendei C Entwicklung und Evaluation eines Schulungsprogramms für Studenten im Praktischen Jahr in der Inneren Medizin GMS Z Med Ausbild 2006 23 4 Doc70 Available from: http://www.egms.de/static/de/journals/zma/2006-23/zma000289.shtml 51 Yates J Development of a "toolkit" to identify medical students at risk of failure to thrive on the course: an exploratory retrospective case study BMC Med Educ 2011 11 1 95 10.1186/1472-6920-11-95 Available from: http://dx.doi.org/10.1186/1472-6920-11-95 22098629 52 Winston KA van der Vleuten CP Scherpbier AJ Prediction and prevention of failure: An early intervention to assist at-risk medical students Med Teach 2014 36 1 25–31 10.3109/0142159X.2013.836270 Available from: http://dx.doi.org/10.3109/0142159X.2013.836270 24083365 53 Bosse HM Mohr J Buss B Krautter M Weyrich P Herzog W Jünger J Nikendei C The benefit of repetitive skills training and frequency of expert feedback in the early acquisition of procedural skills BMC Med Educ 2015 15 1 22 10.1186/s12909-015-0286-5 Available from: http://dx.doi.org/10.1186/s12909-015-0286-5 25889459 54 Van Der Vleuten CP Schuwirth LW Assessing professional competence: From methods to programmes Med Educ 2005 39 3 309–317 10.1111/j.1365-2929.2005.02094.x Available from: http://dx.doi.org/10.1111/j.1365-2929.2005.02094.x 15733167 55 Epstein RM Medical education - Assessment in medical education N Engl J Med 2007 356 4 387–396 10.1056/NEJMra054784 Available from: http://dx.doi.org/10.1056/NEJMra054784 17251535 56 Norcini J Burch V Workplace-based assessment as an educational tool: AMEE Guide No. 31 Med Teach 2007 29 9 855–871 10.1080/01421590701775453 Available from: http://dx.doi.org/10.1080/01421590701775453 18158655 57 Abadel FT Hattab AS How does the medical graduates' self-assessment of their clinical competency differ from experts' assessment? BMC Med Educ BMC Med Educ 2013 13 1 24 10.1186/1472-6920-13-24 Available from: http://dx.doi.org/10.1186/1472-6920-13-24 23402221 58 Falchikov N Student self-assessment in higher education: A meta-analysis Rev Educ Res 1989 59 359 394–430 10.3102/00346543059004395 Available from: http://dx.doi.org/10.3102/00346543059004395 59 Davis DA Mazmanian PE Fordis M Van Harrison R Thorpe KE Perrier L Accuracy of Physician Self-assessment Compared With Observed Measures of Competence JAMA 2006 296 9 1094 10.1001/jama.296.9.1094 Available from: http://dx.doi.org/10.1001/jama.296.9.1094 16954489 60 Berner ES Graber ML Overconfidence as a cause of diagnostic error in medicine Am J Med 2008 121 5 Suppl S2–23 18440350 61 Cavalcanti RB Sibbald M Am I right when I am sure? Data consistency influences the relationship between diagnostic accuracy and certainty Acad Med 2014 89 1 107–113 10.1097/ACM.0000000000000074 Available from: http://dx.doi.org/10.1097/ACM.0000000000000074 24280850 62 Verghese A Charlton B Kassirer JP Ramsey M Ioannidis JP Inadequacies of Physical Examination as a Cause of Medical Errors and Adverse Events: A Collection of Vignettes Am J Med 2015 128 12 1322 1324 10.1016/j.amjmed.2015.06.004 Available from: http://dx.doi.org/10.1016/j.amjmed.2015.06.004 26144103 63 Lane JL Gottlieb RP Improving the interviewing and self-assessment skills of medical students: is it time to readopt videotaping as an educational tool? Ambul Pediatr 2004 4 3 244–248 10.1367/A03-122R1.1 Available from: http://dx.doi.org/10.1367/A03-122R1.1 15153057 64 Hawkins SC Osborne A Schofield SJ Pournaras DJ Chester JF Improving the accuracy of self-assessment of practical clinical skills using video feedback – The importance of including benchmarks Med Teach 2012 34 4 279–284 10.3109/0142159X.2012.658897 Available from: http://dx.doi.org/10.3109/0142159X.2012.658897 22455696 65 Eva KW Armson H Holmboe E Lockyer J Loney E Mann K Sargeant J Factors influencing responsiveness to feedback: on the interplay between fear, confidence, and reasoning processes Adv Health Sci Educ Theory Pract 2012 17 1 15–26 10.1007/s10459-011-9290-7 Available from: http://dx.doi.org/10.1007/s10459-011-9290-7 21468778

PMC005xxxxxx/PMC5003130.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00105210.3205/zma001052Doc53urn:nbn:de:0183-zma0010526ArticleFeedback promotes learning success! Which kind of feedback for the faculty is given by an interdisciplinary OSCE with focus on decision-making? Feedback fördert den Lernerfolg – auch für Lehrende? Welches Feedback enthält ein fachübergreifender OSCE zum Thema „Entscheidungsfindung“ für die Fakultät? Stibane Tina *1Sitter Helmut 2Neuhof Despina 3Wiechens Helena 4Schönbauer Andrea 1Bösner Stefan 5Baum Erika 51 Universität Marburg, Fachbereich Medizin, Dr. Reinfried Pohl-Zentrum für medizinische Lehre (RPZ), Marburg, Germany2 Universität Marburg, Dekanat Medizin, Marburg, Germany3 Praxis für Allgemeinmedizin, Braunfels, Germany4 ehem. Universität Marburg, Fachbereich Medizin, Dr. Reinfried Pohl-Zentrum für medizinische Lehre (RPZ), Marburg, Germany5 Universität Marburg, Abteilung für Allgemeinmedizin, Präventive und Rehabilitative Medizin, Marburg, Germany*To whom correspondence should be addressed: Tina Stibane, Universität Marburg, Fachbereich Medizin, Dr. Reinfried Pohl-Zentrum für medizinische Lehre (RPZ), Conradistraße 9, D-35043 Marburg, Germany, E-mail: stibane@staff.uni-marburg.de15 8 2016 2016 33 4 Clinical skillsDoc5329 9 2015 31 5 2016 31 5 2016 Copyright © 2016 Stibane et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001052.shtmlClinical skills such as history taking, diagnostic reasoning, therapy planning, and giving advice are even more complex than practical skills like lung auscultation and have to be applied in complex clinical situations. We judged this competence in an interdisciplinary formative OSCE conducted with students of Marburg University. Results of 218 students passing 643 OSCE stations composed of 37 different scenarios were analyzed. As a competence based examination that reflects the practical skills gained during clinical training, the here presented analysis serves also as a feedback instrument for clinical teachers, their respective disciplines and the medical faculty as a whole. Zusammenfassung Praktische klinische Fertigkeiten, die klinisches Denken erfordern und über technisches Können hinausgehen, lassen sich schwer prüfen. Anamneseerhebung, zielgerichtetes diagnostisches Vorgehen, differentialdiagnostisches Denken, Therapieempfehlung und Beratung müssen in verschiedensten komplexen Anforderungssituationen angewendet werden und sind selber komplexer als isolierte Aufgaben wie Lungenauskultation. Bei Marburger Studierenden der Humanmedizin im 3. klinischen Studienjahr wird diese Kompetenz in einem fächerübergreifenden und formativen OSCE-„Entscheidungsfindung“ überprüft. Die Leistungen von 218 Studierenden im Studienjahr 2014/2015 wurden statistisch ausgewertet. Die Analyse der Ergebnisse gibt den beteiligten Lehrenden, ihren Fächern und auch der Fakultät als Ganzes ein Feedback über ihre Lehre. medical educationcompetency-based assessmentfeedback-role of assessmentfaculty development ==== Body 1. Introduction During the last years objective structured clinical examinations (OSCEs) are widely used in medical education in Germany [1]. They are well suited to test practical skills of students [2], [3], [4]. With OSCE at Philipps-University Marburg it was shown, that basic practical skills of physical examination could be improved by structured practical teaching [5]. The ability for a target oriented clinical examination, which students have to develop during the course of their studies, is much more complex: the performance of clinical examinations is accompanied by a comprehensive knowledge about diseases and diagnostic criteria. The doctor-patient-interaction is more complex, if it is not only a modular task for a test for students. During students’ education decision making is more and more in the focus and discussed in clinical courses. Explicitly this is the case in the integrated lecture of the third clinical year and in the seminar general practice, which accompanies the corresponding clinical course. This is the reason for establishing an interdisciplinary assessment in the curriculum of Marburg, which tests different competences of physicians: the “OSCE decision making”. Within the frame of two dissertations (Neuhof D, Wiechens H) and in cooperation with the department of general practice and a centre of medical education, belonging to the deanery of education, case vignettes, scripts for role plays, and evaluation criteria were developed, tested, and identified as suitable. For the choice of case vignettes, which had to be dealt with at the different stations, the frequency and relevance of different causes for consultations in the primary care setting were relevant. At the same time it should be possible to test a broad range of basic practical skills with the case vignettes. Especially their matching with practical skills in the curriculum of the department and with the National Competence Based Catalogues of Learning Objectives for Undergraduate Medical Education (NKLM) [6] was emphasized. Designed as an interdisciplinary examination in a setting of a general physician, during the course of time more and more clinical disciplines were motivated for participating at OSCE-decision making. In this way a pool of over 40 stations was generated, and every student in the third clinical year visits - according to a random principle - three stations with 20 minutes each, directly after the practical course “general practice”. During the first 15 minutes there is interaction between the “physician” (student) and the “patient” (standardised patient), during the last five minutes differential diagnoses and preventabale dangerous courses of disease are discussed. The reviewers give feedback on action, behaviour, and clinical reasoning of the students. During the last five minutes the standardised patients score according to a check list the behaviour of the students with respect to performance, shown empathy, and presentation of information during the interaction with an overall score each. After the examination the students have the chance to know their score by the examiner as well as the score of the interaction in written form as a scholar mark from 1 (very good) to 5 (unsatisfactory). Nevertheless this examination has no relevance for the students’ certificates. The examiners as well as the standardised patients are trained for applying this scoring and the majority of them is experienced. As feedback to the examiners the students could give a comment on the format of the examination with free text at the end. An examination which tests the presence of a practical skill which was taught in lessons before, directs the learning behaviour of the students and leads to practical exercise of students and acquirement of job specific and practice oriented competence [5]. The examination of complex decision making which consists of parts from the whole course of the study does not have – as a one shot event during the study - a directive function for learning, because focused rehearsal is not possible. This is one of the reasons why the OSCE - decision making was introduced as a formative examination without any danger of failing. Therefore an otherwise important aspect of an examination, the selection, is missing. Nevertheless this examination should fulfill a feedback function, where the trained standardised patients give additional feedback. Between January 2012 and March 2013 a total of 937 free text answers (cues, lists, whole sentences) from 272 participants from 11 OSCE examinations were evaluated. More than two third (70.43%) of all statements were categorised as positive feedback. Special emphasis is on an additional 1.28% of statements, which recurred on the format of the examination as a chance for learning: “Such examinations are good for self reflected learning! Unfortunately, this shows clearly the problem of theoretical studying. One is forced to think much more in terms of differential diagnoses, for this reason such examinations are great and should be implemented more often.” That this examination provides very good feedback to students was shown by the evaluation of the free texts [7]. However which kind of feedback is generated by this examination format for the involved medical disciplines? Is it possible to recognize the specific positive and negative aspects of the education as a whole and in different clinical disciplines through the results of the students in the OSCE decision making? 2. Description of the project Every parcours consists of a minimum of 6 and maximum of 10 stations for 14 to 30 students, who take the examination on one date. Every student passes three stations in well-equipped simulation rooms (home visit scenario and private practice), simulator technique is used in addition to the trained standardised patient. Every station is passed by 6 to 9 students. The students are informed about the interdisciplinary formative examination in general at the beginning of their general practice course and about the details and the procedure directly before the examination. At the time of the examination their practical course in private practices is just finished which entails that they are prepared for the setting of the cases of the examination. In contrast to the OSCE of the first clinical year, where a standard procedure for clinical examination techniques was developed, the selected situations here allow different possibilities to achieve the goal. A series of anamnestic questions, physical examinations and other diagnostic actions are necessary as standards in the given setting due to guidelines for general practicioners and expert opinions, but some students relatively quickly get a target-oriented diagnosis in their mind and therefore ask and examine in a clear direction, while others do not, but arrive as well at the right diagnosis. Some steps of a clinical examination are adequate, but not always necessary. Searching techniques for clarifying differential diagnoses belong to this. This is important for the evaluation criteria. A checklist with items, often used in OSCE for practical skills, does not take into account the variance, therefore on the evaluation sheet relevant items of the case are given which include the exclusion of differential diagnoses because of clinical reasoning. However five permanent evaluation criteria were determined for every single step of the interaction and of the differential diagnostic procedure in the course of the development process and with their case specific associated items global scores were introduced. For the scores 1, 3, and 5 requirements are described, the scores 2, and 4 can be used for further differentiation. Evaluated items are (partly word by word) [7]: History taking. In order to avoid the definition of single questions a question-category is built with every case. “Pain” for example, consists completely of all questions about pain, i.e. character of pain, first appearance, course, cause etc. “Diseases and drugs” consists of all important questions for the main symptoms about accompanying, previous, and family diseases and drugs taken. Target-oriented selection of physical examinations and target-oriented administration of diagnostic actions. For each case there is a list of necessary findings, which are on the check list of the examiner. Evaluated is which physical examinations and requests for medical findings are ordered by the student. Physical examination and interpretation of findings. Evaluated are the procedural quality of the performed investigations on the standardised patient or the simulator and the interpretation of the other findings present from the previous requested clinical examination. Proposal of therapy, consulting and termination of interaction. The termination of each case has to make sense, there exist big differences: According to the case the proposal of therapy should be correct, risks for the patient or the future actions should be addressed. Differential diagnostic weighing. The consideration of “preventable adverse courses” as well as the correct diagnosis, which is given on the checklist lead to the score. 2.1. Data analysis The performance of all students, who participated during winter term 2014/2015 and during summer term 2015 at the OSCE decision making was analysed in the criteria “history taking”, “target-oriented selection of physical examinations and target-oriented administration of diagnostic actions”, “quality of taking findings”, “termination of interaction and future actions”, and “differential diagnostic thinking” with respect to distribution of scores, average scores and standard deviation at the different stations as well as the comparison of means at all stations and all disciplines by ANOVA and Scheffe-test in order to test the influence of differences in performance statistically. The significance level was chosen as 0.05. The results of the students at different stations were treated as independent data, because data analysis with dependent data was not necessary for the addressed questions. The evaluation, which the standardised patients gave on the competence of communication and behavior were analysed, according to their distribution in the categories “performance”, “empathy”, and “information” for all students and stations. The data analysis (figure 1 (Fig. 1), table 1 (Tab. 1), frequencies, means, standard deviations, significances, ANOVA)) was performed with the statistics software SPSS (IBM SPSS Statistics 22) and for the figure 2 (Fig. 2), figure 3 (Fig. 3) and figure 4 (Fig. 4) the software EXCEL (Microsoft Office Excel 2007) was used. 3. Results Between November 2014 and July 2015 altogether 218 students who had just finished their practical course general practice participated in an interdisciplinary OSCE decision making. 208 students attended three stations each (altogether 624 stations), 9 students attended only two stations (examiner was late) (altogether 18 stations). One student cancelled the examination due to illness after one station. During this period of time 37 different stations were used, these stations are mapped to 18 clinical disciplines (see table 1 (Tab. 1)). 3160 evaluations were documented on the evaluation sheets of the examiners in the categories history taking, request, findings, therapy and differential diagnosis. Considering all stations 55 single scores (1.71%) are missing (see table 2 (Tab. 2)). The frequency distribution of the average scores in figure 1 (Fig. 1) has a mean of 2.33 with a left handed skewnees for better scores. The distribution of the mean values of the scores range for the five categories from a minimum value of 2.17 for history taking as best result to a maximum value of 2.48 for recommendation of therapy as worst result, shown in table 2 (Tab. 2) and figure 2 (Fig. 2). The number of the scores 4 and 5 was only 61 (9.5%) for history taking, but 125 (19.4%) for therapy. For physical examination and interpretation of results the students got in 61% (N=388) of cases the score 1 or 2, but in 15% (N=95) of cases the score 4 or 5. Looking at the distribution of scores only by comparison of means, the post-hoc-test resulted in a significant difference between the categories history taking and request. Considering the distribution of scores in the disciplines, descriptively there are marked differences, which are significant in the comparison of means by ANOVA, but if every discipline is tested against every other discipline by post-hoc test (Scheffe) the overall difference rests only on a few significant differences between two disciplines. Statistically significant differences of the scores between the five disciplines selected here for illustration purposes are only seen between general practice and neurology as well as between cardiology and neurology. All other differences are statistically not significant (see figure 3 (Fig. 3)). The standardised patients evaluated for 643 stations 590 valid evaluation sheets with altogether 1729 evaluations of the interaction “doctor” and “patient” (53 times there was no filled in evaluation sheet by the standardised patient, i.e. because of lack of time). The score is missing or equivocal 10 times in the category performance, 15 times in the category empathy, and 16 times in the category information (missing values altogether: 2.31%). Figure 4 (Fig. 4) shows that the standardised patients evaluate the students much more often than average with the scholar marks 1 or 2 (80.34%) and only in 3.2% of the evaluations with 4 or 5. The situation is similar with empathy, 78.08% receive 1 or 2 and only 4.43% receive 4 or 5. The evaluation of “information” was worse, with two third of the evaluations (68.84%) 1 or 2, but also 11.32% of evaluations with 4 or 5. In 10 evaluation sheets students were rated in all three categories with 4 or worse. 4. Discussion The students get an individual feedback on their observed clinical competence by the patient cases from different clinical disciplines, due to the systematics of the stations, and due to the post-examination discussion with the examiner. The performance of the totality of students however expresses how much the students are prepared for the whole spectrum of practical skills of their future job by the teaching of reference disciplines. The average results of all stations in the categories history taking, decision for request of clinical examinations and investigations, termination of “doctor-patient” interaction, and therapy, as well as differential diagnostic feedback give information about the whole teaching process. The reliability of the evaluation was proven to be sufficient in a previous test [Neuhof D], whereas a formal evaluation of validity was not performed. The overall result was, that the students showed a good level of education before the start of the “practical year”. This does not apply to everybody to the same amount: It is apparent, that with positive overall performance, a small part of students (3.6%) at one station in all five categories is inferior to 4.0 and 2 students (<1%) are at all stations inferior to 4.0. In addition the standardised patient´s state marked shortcomings in 10 cases of all evaluated aspects of doctor-patient-communication. The fact that history taking is part of the curriculum in clinical as well as in communicative teaching in several lectures, resembles the good results for this part of the task at the stations. Here the task to recommend a therapy or risk reducing behaviour to the patient or to explain necessary actions gets the most inferior marks. One reason for this might be, that in this step of the examination a decision about an intervention is warranted and this is addressed in the teaching not in detail, one trusts that this area is taught in detail during the “practical year”. The difference in the performance of the students according to disciplines can be recognised descriptively, but significant differences in the comparison of the means are present only in a few cases. Consideration of the distribution of scores at a single station give each discipline a chance for detailed reflexion about the causes of evaluation scores and for adaptation of their teaching contents. The evaluations of the doctor-patient-communication by the standardised patients are very favourable; nevertheless the standardised patients also identify single students, by whom they feel treated not adequately in several aspects. The results in total show a positive image of the performance of the students. Furthermore the OSCE decision making as a competence oriented examination is not only well suited as feedback for students but also for the reference disciplines, to check for theoretical teaching content and how it can be applied in a reality-near situation. In addition the OSCE decision making can be used as orientation by the faculty to decide, whether the education of students is sufficient successful with respect to measurable competences in the examination. A shortcoming might be that students perhaps take a formative examination not serious enough, and therefore show not their maximum performance. The low number of stations, which are conducted by the students, might be another source of bias of the individual results with respect to objective performance. Chance plays an important role, if previous experience from “practice” with the addressed differential diagnoses is present, or if a specific skill is needed which has been taught only in one lecture. Furthermore success or failure at the first station might influence the future performance and for sure the length of time from lecture to examination has an influence on the result. A restriction with respect to measuring performance is the fact that not all pathologic findings can be simulated well (i.e. blood pressure and heart rate) and the change of standardised patients and simulators pose problems for accomplishing the task by the students. In spite of this potential bias concerning measurement of individual performance and quality of test, the authors share the opinion, that the results altogether can be used as feedback to the faculty. In this sense the OSCE decision making in its present form is not a reliable instrument for measurement of performance, but an instrument for reflexion of individual and discipline specific - and moreover by the spectrum of 18 clinical disciplines also faculty-related – teaching related strengths and weaknesses and is therefore able to contribute to the “learning success” of a faculty. 5. Conclusions The OSCE decision making has been worked up to an important feedback tool in the Marburg course of studies in human medicine. The fact, that it is interdisciplinary is for many disciplines a chance to participate, because they do not have to fill several stations from a certain discipline, but they can assess a part of their learning objectives. At the same time the totality of stations tests the majority of practical skills defined in the curriculum, and a lot of competence oriented learning targets are evaluated, which so far are not addressed in any other examination. These results encourage to present regularly a summary of analysed examination data to the disciplines and to stimulate the discussion of their specific results. Being one of the few competence oriented examinations at the faculty of medicine in Marburg, it is desirable to spread this format to more stations. Within the planned restructuring of the curriculum and the future consideration of reimbursement of educational trained medical examiners this possibility seems to come true in practice. In addition, it makes sense and it is feasible, to consult and to supervise students, who performed unsatisfactory in this examination during their further course of studies and their practical year. In one case this has already been done. Competing interests The authors declare that they have no competing interests. Table 1 Reference disciplines (N=18) and corresponding cases/stations (N=37) for “OSCE decision making“ during the teaching year 2014/2015 Table 2 Distribution, Mean and Standard deviation for the 5 categories of target oriented clinical action for all students (N=218) and stations (N=37) Figure 1 Frequency of mean scores (D) for all students and stations Figure 2 Distribution of scores (1=very good – 5=unsatisfactory) in percent in the five categories history taking, request, findings, therapy, and differential diagnosis (DD) Figure 3 Distribution of mean scores (1=very good – 5=unsatisfactory) in selected disciplines Figure 4 Distribution of scores for doctor-patient-interaction in the categories performance, empathy and information (1=very good - 5=unsatisfactory) due to the judgement of standardised patients ==== Refs 1 Möltner A Duelli R Resch F Schultz JH Jünger J Fakultätsinterne Prüfungen an den deutschen medizinischen Fakultäten GMS Z Med Ausbild 2010 27 3 Doc44 10.3205/zma000681 Available from: http://dx.doi.org/10.3205/zma000681 2 Harden RM Gleeson FA Assessment of clinical competence using an objective structured clinical examination (OSCE) Med Educ 1979 13 1 41 54 10.1111/j.1365-2923.1979.tb00918.x Available from: http://dx.doi.org/10.1111/j.1365-2923.1979.tb00918.x 763183 3 Nikendei C Jünger J OSCE - praktische Tipps zur Implementierung einer klinisch-praktischen Prüfung GMS Z Med Ausbild 2006 3 3 Doc47 Available from: http://www.egms.de/static/de/journals/zma/2006-23/zma000266.shtml 4 Van der Vleuten C Swanson DB Assessment of clinical skills with standardized patients: State of the art Teach Learn Med 1990 2 58 76 10.1080/10401339009539432 Available from: http://dx.doi.org/10.1080/10401339009539432 5 Stibane T Schönbauer A Jerrentrup A Pressel T Baum E Bösner S Sytematischer praktischer Unterricht führt zu mehr praktischer Kompetenz Z Allg Med 2012 88 4 184 191 6 Schnabel KP Boldt PD Breuer G Fichtner A Karsten G Kujumdshiev S Schmidts M Stosch C Konsensusstatement "Praktische Fertigkeiten im Medizinstudium" – ein Positionspapier des GMA-Ausschusses für praktische Fertigkeiten GMS Z Med Ausbild 2011 28 4 Doc58 10.3205/zma000770 Available from: http://dx.doi.org/10.3205/zma000770 22205916 7 Stibane T Neue Herausforderungen an Prüfungsformate durch das kompetenzorientierte Paradigma. Die Implementierung von kompetenzorientierten Prüfungsinstrumenten am Beispiel des Studiengangs der Humanmedizin an der Philipps-Universität Marburg 2015 Gießen Universität Gießen-Marburg

PMC005xxxxxx/PMC5003131.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00106510.3205/zma001065Doc66urn:nbn:de:0183-zma0010654ArticlePractical Skills en route to Professionalism Praktische Fertigkeiten auf dem Weg zur Professionalität Schnabel Kai P. *1Stosch Christoph *21 Universität Bern, Institut für medizinische Lehre, Abteilung für Unterricht und Medien, Bern, Schweiz2 Universität zu Köln, Medizinische Fakultät, Referat für Lehre, Studium & Studienreform, Kölner Interprofessionelle Sklls Labs (KISS), Köln, Deutschland*To whom correspondence should be addressed: Christoph Stosch, Universität zu Köln, Medizinische Fakultät, Referat für Lehre, Studium & Studienreform, Kölner Interprofessionelle Sklls Labs (KISS), Joseph-Stelzmann-Straße 20, D-50931 Köln,, Deutschland, E-mail: c.stosch@uni-koeln.de15 8 2016 2016 33 4 Clinical skillsDoc6617 7 2016 17 7 2016 17 7 2016 Copyright © 2016 Schnabel et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001065.shtml ==== Body Foreword and annotations The acquisition of practical skills has always been in the shadows of medical education. Until about 100 years ago, the pragmatic on model-oriented education was at the forefront and the teachers relied largely, on that the apprentices could require the needed practical skills through observation of the experts with the following practice on patients (see Figure 1 (Fig. 1)). The chair and his clinic, at which the apprentices are employed, guarantee the qualitative mediation of the practical skills. Publications about practical skills are found in the field of training [1]. Since the ‘90s, practical skills during education have earned internationally an important value [2]. Simulations are also progressively winning value in the acquisition of practical skills and are in several areas being considered superior to traditional training methods [3]. Where once patients as training objects served, the role can often be taken now by the simulators or somewhat prepare the students technically better than the plain reading from textbooks and manuals. This doesn’t mean that they will not continue to train on the patients. Just like in the aviation, in which they have flight simulators to prepare the apprentices and assure qualitative outcomes, is now also imaginable in the fields of medicine. Practical procedures like blood taking or lumbar puncturing are being practiced in a safe environment under supervision on a model, before they are executed on live patients. Since over 20 years, starting out in the fields of surgery, the OSCE (Objective Structured Clinical Examination) was established to inspect and assure practical skills [4]. Not only to certify but also to recertify in a way of uninterrupted quality assurance. Consistent standardization through testing of practical skills and also the test-statistical requirements for the inspection (validity, objectivity and reliability) has called forth worldwide a variety of learn-goal catalogs such as the Swiss Catalogue of Learning Objectives [5], the Canadian CANMEDS acting model [6], the Dutch Blueprint [7] and Scottish learn-goal catalog [8], all that are being used as guiding principles in various fields of medical education and further training and are continually evolving [9]. In German-speaking countries, Skills-Labs have been founded at medical faculties since the nineties to live up to the increased requirements of this sector [10]. The changes in the approbation regulation of 2002 [https://www.gesetze-im-internet.de/_appro_2002/ cited 13.07.2016], the initiation of the tuition fees in many German states in 2007 and the initiation of a national practical exam in Switzerland in 2011 [11] supported this development at the faculties to exert themselves more for the practical and communicative skills of their graduates and to establish supportive Skills-Labs [12]. The GMA-committee Practical Skills [https://gesellschaft-medizinische-ausbildung.org/ausschuesse/praktische-fertigkeiten.html cited 13.07.2016] was founded in 2007 to foster the mediation of practical skills and to strengthen the research on this topic [https://gesellschaft-medizinische-ausbildung.org/ausschuesse/praktische-fertigkeiten.html cited 13.07.2016]. It is also worth mentioning that these activities and the caused paradigm shift in the medical education, have also lead to the event of deciding the National competence-based learning-goal catalog (NKLM) [http://www.nklm.de/kataloge/nklm/lernziel/uebersicht cited 13.07.2016], [13] and the national learning-goal catalog dentistry (NKLZ) [http://www.nklz.de/kataloge/nklm/lernziel/uebersicht cited 13.07.2016] by a large majority (29:3 votes, one abstention) on the medical faculty day on the fourth of July 2015 in Kiel. Basis of this chapter 14b (clinical-practical skills) was the, by the GMA-committee for practical skills developed, consensus statement practical skills [14]. These twenty-year-old developments deserve it to release a special edition on the topic of the “GMA-committee for practical skills”. In this booklet, you can find the original work of different general topics of the mediation of practical skills, project reports and statements of the GMA-committee for exams, veterinary medicine and dentistry as to the GMA-committee for practical skills adjacent committees partly assigned with overlapping contents. In the beginning, there are comprehensive papers [15], [16], [17], [18], the following articles are the original papers picking up the recent scientific questions [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29]. In the end they are followed by statements of the “adjacent” committees [30], [31], [32]. We would like to thank all the contributors in the field of practical skills, especially the authors and reviewers of the articles in this special edition and we wish the readers an inspiring read. Notes Dedicated to the Pioneers who committed themselves to the teachings of practical skills during and after medical education Comepting interests The authors declare that they have no competing interests. Figure 1 Prof. Ernst v. Bergmann, Demonstration operation, Ziegelstrasse, Berlin around the turn of the century. ==== Refs 1 Bannister SL Hilliard RI Regehr G Lingard L Technicalskills in pediatrics: a qualitative study of acquisition, attitudes and assumptions in the neonatal intensive care unit Med Educ 2003 37 12 1082 1090 10.1111/j.1365-2923.2003.01711.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2003.01711.x 14984114 2 Remmen R Derese A Scherpbier A Denekens J Hermann I van der Vleuten C Van Royen P Bossaert L Can medical schools rely on clerkships to train students in basic clinical skills? Med Educ 1999 33 8 600 605 10.1046/j.1365-2923.1999.00467.x Available from: http://dx.doi.org/10.1046/j.1365-2923.1999.00467.x 10447847 3 McGaghie WC Issenberg SB Cohen ER Barsuk JH Wayne DB Does simulation-based medical education with deliberate practice yield better results than traditional clinical education? A meta-analytic comparative review of the evidence Acad Med 2011 86 6 706 711 10.1097/ACM.0b013e318217e119 Available from: http://dx.doi.org/10.1097/ACM.0b013e318217e119 21512370 4 Sloan DA Donnelly MB Schwartz MD Strodel WE The objective structured clinical examination. The new gold standard for evaluation postgraduate clinical performance Ann Surg 1995 222 6 735–742 10.1097/00000658-199512000-00007 Available from: http://dx.doi.org/10.1097/00000658-199512000-00007 8526580 5 Bloch R Bürgi H The Swiss Catalogue of Learning Objectives Med Teach 2002 24 2 144 150 10.1080/01421590220120759 Available from: http://dx.doi.org/10.1080/01421590220120759 12098433 6 Frank JR CanMEDS 2005 Physician Competency Framework 2005 Ottawa The Royal College 23 24 Available from: http://www.ub.edu/medicina_unitateducaciomedica/documentos/CanMeds.pdf 7 Metz JC Verbeek-Weel AM Huisjes HJ Blueprint 2001: training of doctors in The Netherlands 2001 Utrecht NFU Available from: http://www.bibliosgam.ch/pdf/blueprint.pdf pdf 8 The Scottish Deans' Medical Curriculum Group Learning Outcomes for the Medical Undergraduate in Scotland: A foundation for competent and reflective practitioners 2000 Edinburgh University of Edinburgh Available from: http://www.scottishdoctor.org/resources/scotdoc1.pdf 9 Michaud PA Jucker-Kupper P Profiles working group The "Profiles" document: a modern revision of the objectives of undergraduate medical studies in Switzerland Swiss Med Wkly 2016 146 w14270 26829005 10 Segarra LM Schwedler A Weih M Hahn EG Schmidt A Der Einsatz von medizinischen Trainingszentren für die Ausbildung zum Arzt in Deutschland, Österreich und der deutschsprachigen Schweiz GMS Z Med Ausbild 2008 25 2 Doc80 Available from: http://www.egms.de/static/de/journals/zma/2008-25/zma000564.shtml 11 Guttormsen S Beyeler C Bonvin R Feller S Schirlo C Schnabel K Schurter T Berendonk C The new licencing examination for human medicine: from concept to implementation Swiss Med Wkly 2013 143 w13897 10.4414/smw.2013.13897 Available from: http://dx.doi.org/10.4414/smw.2013.13897 24301323 12 Damanakis A Statusreport von Skills Labs in der D-A-CH-Region und Aufbau einer Informationsplattform zur Katalogisierung und Bewertung von Simulatoren zur medizinischen Ausbildung, Inaugural-Dissertation 2015 Marburg Philipps-Universität-Marburg Available from: http://d-nb.info/1071865412/04 13 Hahn EG Fischer MR Nationaler Kompetenzbasierter Lernzielkatalog Medizin (NKLM) für Deutschland: Zusammenarbeit der Gesellschaft für Medizinische Ausbildung (GMA) und des Medizinischen Fakultätentages (MFT) GMS Z Med Ausbild 2009 26 3 Doc35 10.3205/zma000627 Available from: http://dx.doi.org/10.3205/zma000627 14 Schnabel K Boldt P Breuer G Fichtner A Kujumdshiev S Karsten G Schmidts M Stosch C Konsensusstatement Praktische Fertigkeiten im Medizinstudium" - ein Positionspapier des GMA-Ausschusses für praktische Fertigkeiten GMS Z Med Ausbild 20011 28 4 Doc58 10.3205/zma000770 Available from: http://dx.doi.org/10.3205/zma000770 22205916 15 Gerhard-Szép S Güntsch A Pospiech P Söhnel A Scheutzel P Wassmann T Zahn T Assessment formats in dental medicine: An overview GMS J Med Educ 2016 33 4 Doc65 10.3205/001064 Available from: http://dx.doi.org/10.3205/001064 16 Vogel D Harendza S Basic practical skills teaching and learning in undergraduate medical education – a review on methodological evidence GMS J Med Educ 2016 33 4 Doc64 10.3205/001063 Available from: http://dx.doi.org/10.3205/001063 17 Bugaj TJ Nikendei C Practical Clinical Training in Skills Labs: Theory and Practice GMS J Med Educ 2016 33 4 Doc63 10.3205/001062 Available from: http://dx.doi.org/10.3205/001062 18 Dannenberg KA Stroben F Schröder T Thomas A Hautz WE The future of practical skills in undergraduate medical education – an explorative Delphi-Study GMS J Med Educ 2016 33 4 Doc62 10.3205/001061 Available from: http://dx.doi.org/10.3205/001061 19 Schmitt L Möltner A Rüttermann S Gerhard-Szép S Study on the Interrater Reliability of an OSPE (Objective Structured Practical Examination ) – Subject tot the Evaluation Mode in the Phantom Course of Operative Dentistry GMS J Med Educ 2016 33 4 Doc61 10.3205/001060 Available from: http://dx.doi.org/10.3205/001060 20 Münster T Stosch C Hindrichs N Franklin J Matthes J Peyton's 4-Steps-Approach in comparison: Medium-term effects on learning external chest compression – a pilot study GMS J Med Educ 2016 33 4 Doc60 10.3205/001059 Available from: http://dx.doi.org/10.3205/001059 21 Fünger SM Lesevic H Rosner S Ott I Berberat P Nikendei C Sonne C Improved self- and external assessmentof the clinical abilities of medical students through structured improvement measures in an internal medicine bedside course GMS J Med Educ 2016 33 4 Doc59 10.3205/001058 Available from: http://dx.doi.org/10.3205/001058 22 Störmann S Stankiewicz M Raes P Berchtold C Kosanke Y Illes G Loose P Angstwurm MW How well do final year undergraduate medical students master practical clinical skills? GMS J Med Educ 2016 33 4 Doc58 10.3205/001057 Available from: http://dx.doi.org/10.3205/001057 23 Weber U Constantinescu MA Woermann U Schmitz F Schnabel K Video-based instructions for surgical hand disinfection as a replacement for conventional tuition?A randomised, blind comparative study GMS J Med Educ 2016 33 4 Doc57 10.3205/001056 Available from: http://dx.doi.org/10.3205/001056 24 Friederichs H Brouwer B Marschall B Weissenstein A Mastery learning improves students skills in inserting intravenous access: a pre-post-study GMS J Med Educ 2016 33 4 Doc56 10.3205/001055 Available from: http://dx.doi.org/10.3205/001055 25 Karthaus A Schmidt A "PERLE bedside-examination-course for candidates in state examination" – Developing a training program fort he third part of medical state examination (oral examination with practical skills) GMS J Med Educ 2016 33 4 Doc55 10.3205/001054 Available from: http://dx.doi.org/10.3205/001054 26 Nikendei C Ganschow P Groener JB Huwendiek S Köchel A Köhl-Hackert N Pjontek R Rodrian J Scheibe F Stadler AK Steiner T Stiepak J Tabatabai J Utz A Kadmon M "Heidelberg standard examination " and "Heidelberg standard procedures" – Development of faculty-wide standards for physical examination techniques and clinical procedures in undergraduate medical education GMS J Med Educ 2016 33 4 Doc54 10.3205/001053 Available from: http://dx.doi.org/10.3205/001053 27 Stibane T Stitter H Neuhof D Wiechens H Schönbauer A Bösner S Baum E Feedback promotes learning success! Which kind of feedback fort he faculty is given by an interdisciplinary OSCE with focus on decision-making? GMS J Med Educ 2016 33 4 Doc53 10.3205/001052 Available from: http://dx.doi.org/10.3205/001052 28 Vajda C "Peer2Peer" – A university program for knowledge transfer and consultation in dealing with psychosocial crises in med-school and medical career GMS J Med Educ 2016 33 4 Doc52 10.3205/001051 Available from: http://dx.doi.org/10.3205/001051 29 Tim A Polack C Commentary: Clincial skills teaching in UK medical education as exemölified by the BM5 curriculum GMS J Med Educ 2016 33 4 Doc51 10.3205/001050 Available from: http://dx.doi.org/10.3205/001050 30 Bauer D Huwendiek S März M "Pass, fail" – On Standard Setting Procedures fort he Assessment of Practical Skills at Medical Schools in Germany, Austria, and Switzerland GMS J Med Educ 2016 33 4 Doc50 10.3205/001049 Available from: http://dx.doi.org/10.3205/001049 31 Dilly M Gruber C Committee on Veterinary Medicine at the Society for Medical Education: Skills Labs in Veterinary Medicine – a brief overview GMS J Med Educ 2016 33 4 Doc49 10.3205/001048 Available from: http://dx.doi.org/10.3205/001048 32 Scheutzel P Gerhard-Szép S "Practical skills" – Positioning of the GMA committee for dentistry GMS J Med Educ 2016 33 4 Doc48 10.3205/001047 Available from: http://dx.doi.org/10.3205/001047

PMC005xxxxxx/PMC5003132.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00105010.3205/zma001050Doc51urn:nbn:de:0183-zma0010509ArticleCommentary: Clinical skills teaching in UK medical education as exemplified by the BM5 curriculum, Faculty of Medicine, University of Southampton Kommentar: Vermittlung von klinischen Fähigkeiten bei der medizinischen Ausbildung im Vereinigten Königreich, am Beispiel des BM5 Curriculum der Medizinischen Fakultät, Universität Southampton Timm Anja *1Polack Clare 11 University of Southampton, Faculty of Medicine, Academic Unit of Medical Education, Medical Education Development Unit (MEDU), Southampton, UK*To whom correspondence should be addressed: Anja Timm, University of Southampton, Faculty of Medicine, Academic Unit of Medical Education, Medical Education Development Unit (MEDU), Life Sciences Building (B85), Highfield Campus, Southampton SO17 1BJ, UK, E-mail: a.timm@soton.ac.uk15 8 2016 2016 33 4 Clinical skillsDoc5130 9 2015 09 5 2016 02 2 2016 Copyright © 2016 Timm et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001050.shtmlThis commentary seeks to enable comparisons about clinical skills teaching in Germany and the UK. It outlines the British regulatory environment and its impact on programme design. Through the example of the University of Southampton we show how clinical skills teaching is integrated both vertically and horizontally. Zusammenfassung Dieser Kommentar soll Vergleiche über die Vermittlung klinischer Fertigkeiten in Deutschland und Großbritannien ermöglichen. Er skizziert das regulatorische Umfeld in Großbritannien und seine Auswirkung auf die Programmgestaltung an den medizinischen Fakultäten. Am Beispiel der University of Southampton zeigen wir, wie die Vermittlung klinischer Fertigkeiten sowohl vertikal als auch horizontal integriert ist. ==== Body Introduction As a European medical school running an undergraduate programme in cooperation with a German healthcare provider, Gesundheit Nordhessen, we have been following the recent developments in German medical education with great interest, especially the recently published recommendations of the Wissenschaftsrat [1] and the establishment of the national catalogue of competencies (http://www.nklm.de citid 28 September 2015). We think it is helpful to explore developments in other countries and we seek to facilitate a comparative perspective with this commentary. UK context For UK medical schools, all curricula are governed by the national guidance of the General Medical Council (GMC). In “Tomorrow’s Doctors” (2009), the GMC highlights the importance of three domains: The doctor as - ‘a scholar and scientist’, ‘a practitioner’ and ‘a professional’ (http://www.gmc-uk.org/education/undergraduate/undergrad_outcomes.asp cited 29 September 2015). This essentially provides the learning outcomes for all aspects of medical programmes, including non-technical skills (NTS), such as communication and teamwork as well as professional attitudes and behaviours. An appendix outlines 32 practical competencies (http://www.gmc-uk.org/Outcomes_for_graduates_Jul_15.pdf_61408029.pdf cited 29 September 2015). All graduates must achieve both the learning outcomes and the ‘practical procedures’. The overall UK emphasis is on the integration of science/knowledge and its application in clinical practice. Hence most UK medical schools apply an integrated approach to the teaching of practical skills (PS). Traditional UK undergraduate programmes tend to have the first two years mainly taught at the university, followed by three years mainly in clinical settings. However, a strict division into a pre-clinical and clinical phase has largely been eroded by the ever greater prominence of early patient contact initiatives. This also means that PS and NTS are introduced in year one, initially in quite controlled skills lab settings, then increasingly taught, and practiced in the clinical environment. Each medical school designs its curriculum according to its own interpretation of the GMC guidance and local circumstances. However, the GMC quality assures the delivery of each study programme through regular visits and public reporting. University of Southampton example At Southampton we have an intake of about 290 students annually on five different undergraduate programme streams. The majority undertake the Bachelor of Medicine (BM5) programme, which runs over five years and is divided into four phases, which are demarcated by different colours in the figure below (see Figure 1 (Fig. 1)). In each phase PS and NTS are integrated into students’ learning and made relevant by their connection with the science, the patient and the clinical environment. 1. Fundamentals of medicine phase (yellow) In year 1 PS training is integrated mainly within Medicine in Practice (MiP) but links to the content of the systems-based teaching in the other modules. Whilst students study the Nervous and Locomotor 1 module, they take a history of someone with a musculoskeletal problem. The MiP sessions are delivered to small groups of students by general practitioners in their surgeries. NTS, such as time-management, giving and receiving feedback are also facilitated, highlighted and assessed. In year 2 MiP continues as before, although settings alternate between General Practice and hospital. Additionally, all students undertake facilitated health care support work (HCSW). This entails working shifts on a ward where they learn PS and NTS. For example, students are taught basic infection control and they participate in multi-professional teams. During the accompanying facilitated small group tutorials, students are asked to reflect on the nature of teams and how they can work effectively. 2. Progression into clinical practice phase (pink) At the start of year 3 students concentrate on a research project and learn many PS including accessing and critically appraising evidence, which are vital in both scientific and clinical practice. In parallel to undertaking their research module, they also undertake PS training (e.g. basic life support) as well as separate communication sessions with simulated patients to prepare for full-time clinical placements (for 4 of 5 days a week), which start in semester 2 of year 3. In the clinical modules of year 3 – Medicine and Elderly Care / Surgery and Orthopaedics / Long Term Conditions and Primary Medical Care – students gradually progress towards meeting their practical competencies; starting from basic familiarisation using models and practicing in-vitro to observation in practice followed by supervised doing. At this stage, students have to record their progress in a log book to show engagement and facilitate discussion with the clinical supervisor at the end of placement. In subsequent years, the requirements on the students increase and their competencies are logged on their e-portfolio (completion is compulsory). 3. Developing clinical practice phase (green) Immersion in the clinical environment allows students to learn and consolidate PS and NTS. To evidence competency in the 32 skills mandated by the GMC students must perform each 3 times independently whilst being observed by an experienced clinician. Both PS and NTS are also assessed in OSCE exams, and Assessment of Clinical Competence (ACC). The latter is an observed clinical assessment that includes diagnosis and management and is an adaptation of the postgraduate work-based placed assessment [2]. The ACC allows for a much more realistic in-vivo assessment of a student’s capability, i.e. the “doing” rather than ‘showing’ in Millers pyramid [3]. Continuous supervisor assessment on placements also allow PS and NTS to be assessed in an integrated and holistic manner. Students must pass all four types of assessment (as well as written papers). 4. Preparing for independent practice (blue) phase This phase occurs after the major examinations, but is assessed and must be passed to allow graduation. In this final phase students put all the PS and NTS into practice by shadowing and assisting a junior doctor in their daily work. The aim is to ensure that all graduates are properly prepared to take on the responsibilities of their first job. In the UK all graduates enter a two year foundation programme which is also quality assured by the GMC and follows a carefully designed curriculum. At the end of the foundation programme junior doctors should possess all the generic PS and NTS to enable them to start their specialist training. Discussion At Southampton we believe that the knowledge, skills and attitudes required to be a good doctor are equally important. PS and NTS are introduced early in the curriculum and continually revisited in an integrated manner both horizontally within the year and vertically with previous learning [4]. The prominence of these skills in the curriculum and their testing in important assessments leaves students in little doubt how important they are to practicing in the UK. The list of competencies in Germany seems very comprehensive but it is not yet clear (to us) how this translates in practice to a graduate’s confidence and skills in these areas. With the free movement of doctors across Europe this is an important area to examine and collaborate on. More shared knowledge could contribute to better induction and supervision of doctors from other European countries and ultimately improve patient safety. Competing interests The authors declare that they have no competing interests. Figure 1 BM5 curriculum map ==== Refs 1 Wissenschaftsrat Empfehlungen zur Weiterentwicklung des Medizinstudiums in Deutschland auf Grundlage einer Bestandsaufnahme der humanmedizinischen Modellstudiengänge 2014 Dresden Wissenschaftsrat Available from: http://www.wissenschaftsrat.de/download/archiv/4017-14.pdf 2 Hill FJ Kendall K Galbraith K Crossley J Implementing the undergraduate mini-CEX: a tailored approach at Southampton University Med Educ 2009 43 4 326 334 10.1111/j.1365-2923.2008.03275.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2008.03275.x 19335574 3 Norcini JJ Work based assessment BMJ 2003 326 7392 753 755 10.1136/bmj.326.7392.753 Available from: http://dx.doi.org/10.1136/bmj.326.7392.753 12676847 4 Harden RM What is a spiral curriculum? Med Teach 1999 21 2 141 143 10.1080/01421599979752 Available from: http://dx.doi.org/10.1080/01421599979752 21275727

PMC005xxxxxx/PMC5003133.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00105410.3205/zma001054Doc55urn:nbn:de:0183-zma0010543Article“PERLE bedside-examination-course for candidates in state examination” – Developing a training program for the third part of medical state examination (oral examination with practical skills) „Perle Bettenprüfungskurs für Examenskandidaten“ – Entwicklung eines Trainings für den dritten Abschnitt der ärztlichen Prüfung (mündlich-praktische Prüfung) Karthaus Anne *1Schmidt Anita 21 Friedrich-Alexander-University Erlangen-Nuremberg, Tutor at the Skills Lab PERLE, Erlangen, Germany2 Friedrich-Alexander-University Erlangen-Nuremberg, Head of the Skills Lab PERLE, Erlangen, Germany*To whom correspondence should be addressed: Anne Karthaus, Friedrich-Alexander-University Erlangen-Nuremberg, Tutor at the Skills Lab PERLE, Krankenhausstraße 12, D-91054 Erlangen, Germany, E-mail: akarthaus@gmx.de15 8 2016 2016 33 4 Clinical skillsDoc5523 9 2015 20 5 2016 13 5 2016 Copyright © 2016 Karthaus et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001054.shtmlIntroduction: In preparation for the state examination, many students have open questions and a need for advice. Tutors of the Skills Lab PERLE-„Praxis ERfahren und Lernen“ (experiencing and learning practical skills) have developed a new course concept to provide support and practical assistance for the examinees. Objectives: The course aims to familiarize the students with the exam situation in order to gain more confidence. This enables the students to experience a confrontation with the specific situation of the exam in a protected environment. Furthermore, soft skills are utilized and trained. Concept of the course: The course was inspired by the OSCE-model (Objective Structured Clinical Examination), an example for case-based learning and controlling. Acquired knowledge can be revised and extended through the case studies. Experienced tutors provide assistance in discipline-specific competencies, and help in organizational issues such as dress code and behaviour. Evaluation of the course: An evaluation was conducted by the attending participants after every course. Based on this assessment, the course is constantly being developed. In March, April and October 2015 six courses, with a total of 84 participants, took place. Overall 76 completed questionnaires (91%) were analysed. Discussion: Strengths of the course are a good tutor-participants-ratio with 1:4 (1 Tutor provides guidance for 4 participants), the interactivity of the course, and the high flexibility in responding to the group's needs. Weaknesses are the tight schedule, and the currently not yet performed evaluation before and after the course. Conclusion: In terms of “best practise”, this article shows an example of how to offer low-cost and low-threshold preparation for the state examination. Zusammenfassung Einführung: Viele Studierende haben Fragen und Beratungsbedarf bei der Vorbereitung für das letzte mündliche Examen. Um sie bei den Examensvorbereitungen zu begleiten und praktische Hilfestellung zu geben, wurde von Tutorinnen und Tutoren des SkillsLabs Perle- „Praxis ERfahren und LErnen“ der FAU Universität Erlangen-Nürnberg ein neues Kursmodell konzipiert. Ziele: Der Kurs soll den Studierenden helfen, sich mit der Prüfungssituation vertraut zu machen und Sicherheit zu erlangen. So ist die Auseinandersetzung mit dieser spezifischen Situation in einem geschützten Rahmen möglich. Außerdem findet eine Anwendung und eine Vertiefung von Softskills statt. Kurskonzept: Dieser Kurs ist angelehnt an das OSCE Modell (Objective Structured Clinical Examination), einem Fallbasierten Lern- und Prüfungsmodell. Anhand von Fallbeispielen wiederholen und vertiefen die Studierenden bereits gelerntes Wissen. Dabei werden sie von erfahrenen Tutoren begleitet, die Hilfestellung zu fachspezifischen und organisatorischen Fragen wie Kleiderordnung und Verhaltensweisen geben. Auswertung des Kurses: Der Kurs wurde von den Teilnehmern am Ende jedes Kurses evaluiert. Anhand der Evaluation wird der Kurs stetig weiterentwickelt. Im März, April und Oktober 2015 fanden sechs Kurse mit in Summe 84 Teilnehmern statt. 76 Fragebögen (91%) wurden vollständig ausgefüllt und ausgewertet. Diskussion: Als Stärken des Kurses sind ein guter Tutoren-Teilnehmer Schlüssel mit 1:4 (1 Tutor betreut 4 Teilnehmer), die Interaktivität des Kurses und die hohe Flexibilität auf die einzelne Gruppe eingehen zu können, zu nennen. Schwächen sind der limitierte Zeitrahmen und die bisher nicht erhobene Leistungsüberprüfung vor und nach dem Kurs. Schlussfolgerung: Dieser Artikel zeigt im Sinne eines „best practice“-Beispiels eine Möglichkeit auf, niederschwellig und mit geringen Kosten eine Vorbereitung für das praktische Examen anzubieten. Case-based learningcourse in preparation for the state examinationcourse conceptionpractical examination ==== Body 1. Introduction In a newly developed course based on the OSCE model, practical year students are prepared for the last part of the state examination, the oral examination. Tutors of the Skills Lab Perle developed the course “PERLE-Bettenprüfung” (“PERLE bed-side examination”) in order to meet the needs of students just before their third part of the medical state examination. Impetus was the examinees` strong demand for a revision in practical skills and tips for the oral examination. The name of the course is derived from our Skills Lab's name: PERLE – “Praxis ERfahren und LErnen”, which stands for PERLE – “experiencing and learning practical skills”. The other part of the name was derived from the colloquial word for one part of the last oral examination: “Bettenprüfung” (bed-side examination). In Germany the third and also last state examination is an oral examination, in groups of four students, which takes two days. Four auditors evaluate the student's skills in their own field: Surgery, internal Medicine, one subject chosen by the student and one randomly assigned subject. Besides the theoretical knowledge, practical knowledge is often assessed. 2. Objectives Objectives of the course are training and deepening the use of soft skills, such as talking to patients, giving notice of the following examinations and explaining the further procedure. The course aims to support students through familiarizing them with the exam situation and to gain more confidence. Skills that have been trained during the curriculum are now being performed in a simulated assessment. This enables the students to experience a confrontation with the specific situation of the exam in a protected environment. It was specially developed to provide support to those students who yet feel insecure about the practical part of the third state examination. 3. Concept of the course The structure of the course was inspired by the OSCE model (Objective Structured Clinical Examination). Participants pass through several stations, working on different case studies at each station. The case studies were specially prepared for the course by tutors, taking recent guidelines into account. Medical specialists did the counterchecking. Skills in physical examination and anamnesis are assessed. Auditors follow the check-lists to ensure that all the necessary steps are being performed at each station. In our course we modified the model of OSCE so that student pass through the stations in groups of four. Thirty minutes are planned for each station, so there is time to talk over the case studies and to respond to individual questions. Every small group consists of four students (student [1], [2], [3], [4]), which pass through four stations (station A, B, C, D) together. At each station one of the students assumes the role of the doctor and tries to solve the case. The other participants observe the situation and give advice if necessary. Afterwards they give feedback to the “doctor”. In the end every participant is supposed to have worked on one case himself and to have given advice and feedback at the other three stations (see Figure 1 (Fig. 1)). At each station the tutor guides through the case and provides assistance if necessary. The tutor reads out the description of the case report and then starts with the assessment by ticking off the check-list. Afterwards the check-list is discussed (see Attachment 1), and constructive feedback is given to the “doctor”. The different stations cover a variety of medical fields, with a case from neurology, surgery/orthopaedics and internal medicine in each course. Basic components of the physical examination are the same at each and every case-study. At the beginning of the examination the hand disinfection is performed, followed by welcoming the patient and taking the anamnesis. Furthermore, informative sheets with information on the dress code (see Attachment 2) and a list with “Do´s and Don´ts” (see Attachment 3) are handed out. Tutors of the PERLE have collected information and statements in meetings with professors that have been asked about dress code and behaviour of the examinees. The idea came up, because many students in former courses had questions about some practical advice on these issues. 4. Evaluation of the Course In the end of every course a written evaluation (see Attachment 4) was performed, in which we asked for the opinion of every student on the following points: feedback received from the tutor help received from other students information given on dress code and “Do´s and Don’ts” simulation of the state examination assessment of the ease/difficulty and the usefulness of each case study wishes and ideas for following courses The evaluations were given to the participants directly after the course. In numbers, there were six courses, with a total number of 84 students, in March, April and October of 2015. 76 questionnaires were filled in and analysed. 55 were filled in by female students (72,4%) and 21 by male students (27,6%). Item 1 was determined as “too easy/ doesn’t make sens /not helpful” and 5 was determined as “too difficult/ makes sense/helpful”. The evaluation stated the following: The lists with “Do´s and Don’ts” were evaluated with an average value of 4,55 to be very helpful. Likewise, the information about the dress code was helpful with a mean of 4,1. The simulation of the state examination scored an average value of 4,7, the help received from other students scored 4,4 and feedback received from the tutor 4,8 (see Table 1 (Tab. 1)). The case study of acute pancreatitis was evaluated to be very useful with a score of 4,85, although it was rated as easy with 2,95. The case study for hyperthyreosis was said to be very beneficial, with 4,78, and a normal difficulty level of 3,05. Furthermore, the case study of Morbus Bechterew scored 4,4, but was the only one rated a little bit too difficult with 3,5 (see Table 2 (Tab. 2)). In the free comment section, the students could remark compliments, critics and wishes for the further development of the course. On the one hand, there has been the wish for more case studies from the field of neurology and orthopaedics, especially the clinical examination of major joints. On the other hand, the students wanted to have more time for the case studies, for the demonstration of the clinical examination by the tutor, and for more questions about theoretical knowledge. Furthermore, they wanted to be able to choose a special field of medicine for their case studies themselves. It was also mentioned that this course should start in earlier years of study with less difficult case studies and then build up to the last course in the practical year. Besides the participants liked the checklists being handed out to them. Moreover, they liked the opportunity to test their skills and knowledge and the chance to enhance and consolidate them. 5. Discussion The course “PERLE bed-side examination” was developed for students prior to the second state examination. This course has been held for one year now, and has been improved steadily with the help of evaluations, to better fit the students' needs. A high demand to practice the clinical examination, especially in the field of neurology and orthopaedics was shown by the inquiry. Because of this, we have developed special courses only for students right before the final state examination. These courses recapitulate practical and theoretical skills in neurology and orthopaedics. The strong points in favour of the course “PERLE bed-side examination” are the ratio of tutors and students (1 tutor teaches 4 students), the interactivity of the course and the flexibility to answer to every single group. In this particular setting, the special benefit of peer-teaching can be shown, which is repeatedly stated in the literature (e.g. [5]): Students rather dare to try something new and ask a lot of questions in this “sheltered” situation. In the tense situation the students are in just before the exam, this is especially valuable. The weak points are the limitations in time that only allow an extract of the examinated topics to be conducted. Besides, it is not possible to demonstrate and practice a complete, structured clinical examination of the whole body in the setting of the course. Therefore, we are designing another course, which aims to teach a structured physical examination from head to toe. This course will be open for earlier semesters and will be the basis for “PERLE bed-side examination”. Another weak point is the lack of tracking further development and learning enhancement of the participants. To collect data about the students' progress, a group assessment is planned. We want to see whether the students make progress in the course of the four stations. The course was created for the students to achieve more self-assurance, so in our opinion there is no need for it to be integrated into the curriculum. But it is essential that a structured clinical examination is taught in the first clinical years. Therefore, it is important to integrate the course “clinical examination from head to toe” into the curriculum. Strikingly, the number of women participating decreased from 72,4% to 60,6% in the course of the following exams (spring and autumn 2015 together). The reasons for this are to be assessed, in future courses and evaluations, by asking for the students' individual reasons to visit the course. 6. Conclusion With this new course concept, it was possible to meet some of the students' needs, right before their state examination. Through the evaluations we ascertained the need for a broad variety of case studies and the need for more opportunities to practice the clinical examination, especially the neurological and orthopaedic clinical examination. We will put this into action in future semesters, and we will record the students progress by further evaluation. Competing interests The authors declare that they have no competing interests. Supplementary Material Attachment 1 Attachment 2 Attachment 3 Attachment 4 Table 1 Overall evaluation; evaluation in Likert-Scale: 1=not helpful at all – 5=very helpful Table 2 Evaluations of the case study; 1=not useful at all – very useful; 1=too easy – 5=too difficult Figure 1 Course structure ==== Refs 1 Chenot JF Simmenroth-Nayda A Koch A Fischer T Scherer M Emmert B Stanske B Kochen MM Himmel W Can student tutors act as examiners in an objective structured clinical examination? Med Educ 2007 41 11 1032 1038 10.1111/j.1365-2923.2007.02895.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2007.02895.x 17973763 2 Brazeau C Boyd L Crosson J Changing an existing OSCE to a teaching tool: the making of a teaching OSCE Acad Med 2002 77 9 932 10.1097/00001888-200209000-00036 Available from: http://dx.doi.org/10.1097/00001888-200209000-00036 12228103 3 ten Cate O A teaching rotation and a student teaching qualification for senior medical students Med Teach 2007 29 6 566 571 10.1080/01421590701468729 Available from: http://dx.doi.org/10.1080/01421590701468729 17917986 4 Dandavino M Snell L Wiseman J Why medical students should learn how to teach Med Teach 2007 29 6 558 565 10.1080/01421590701477449 Available from: http://dx.doi.org/10.1080/01421590701477449 17922358 5 Lockspeiser TM O'Sullivan P Teherani A Muller J Understanding the experience of being taught by peers: the value of social and cognitive congruence Adv Health Sci Educ Theory Pract 2008 13 3 361 372 10.1007/s10459-006-9049-8 Available from: http://dx.doi.org/10.1007/s10459-006-9049-8 17124627

PMC005xxxxxx/PMC5003134.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00106110.3205/zma001061Doc62urn:nbn:de:0183-zma0010614ArticleThe future of practical skills in undergraduate medical education – an explorative Delphi-Study Die Zukunft praktischer Fertigkeiten im Medizinstudium – eine explorative Delphi-Studie Dannenberg Katja Anne *12Stroben Fabian 1Schröder Therese 3Thomas Anke 3Hautz Wolf E. 41 Charité – Universitätsmedizin Berlin, Lernzentrum (Skills Lab), Berlin, Germany2 Charité – Universitätsmedizin Berlin, Department of Emergency Medicine at Campus Benjamin Franklin, Berlin, Germany3 Charité – Universitätsmedizin Berlin, Gynecology and Obstetrics Clinic, Berlin, Germany4 Inselspital Bern, University Emergency Center, Bern, Switzerland*To whom correspondence should be addressed: Katja Anne Dannenberg, Charité – Universitätsmedizin Berlin, Lernzentrum (Skills Lab), Charitéplatz 1, D-10117 Berlin, Germany, Phone: +49 (0)30/450-576403, Fax: +49 (0)30/450-576922, E-mail: katja-anne.dannenberg@charite.de15 8 2016 2016 33 4 Clinical skillsDoc6225 6 2015 16 3 2016 29 2 2016 Copyright © 2016 Dannenberg et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001061.shtmlBackground: 64% of young medical professionals in Germany do not feel adequately prepared for the practical requirements of the medical profession. The goal of “outcome-orientated training” is to structure medical curricula based on the skills needed when entering the workforce after completing undergraduate medical education, and thus to bridge the gap between the skills graduates have attained and those necessary for a career in the medical profession. Outcome frameworks (OFs) are used for this purpose. In preparation for developing the National Competence-Based Catalogue of Learning Objectives for Medicine (NKLM) – the German OF – the “Consensus Statement of Practical Skills in Undergraduate Medical Education” (which structures the teaching and acquisition of practical skills in Germany and which strongly influenced the “Clinical-Practical Skills” chapter of the NKLM) was published in 2011. It is not uncommon for at least a decade to elapse between the definition and implementation of an OF and the students’ graduation, which can further increase the gap between necessary and acquired skills. Thus, the purpose of this paper is to posit theses for future development in healthcare and to apply these theses to a current OF. Methodology: Partially structured interviews with experts were used to generate theses pertaining to general, future development in healthcare. These theses were assessed by physician experts based on the likelihood of implementation by the year 2025. The 288 learning goals of the consensus statement were assessed for their relevance for medical education in the interim. Results: 11 theses were generated for the development of medicine, and these theses were assessed and discussed by 738 experts. These theses include the increase in diseases associated with old age, the increasing significance of interprofessional cooperation, and the growing prevalence of telemedicine applications. Of the 288 learning goals of the consensus statement, 231 of the goals were assessed as relevant, and 57 were deemed irrelevant for the short-term future. Discussion: The theses on the future of healthcare, which were generated in this study and which were validated by numerous experts, provide indications of future developments of overall requirements for medical school graduates. For example, when applied to the content of the “Clinical-Practical Skills” NKLM chapter, they largely validate the future relevance of developing practical skills while also providing indications for their further development as applied to the consensus statement. Zusammenfassung Hintergrund: 64% der Berufsanfänger in Deutschland fühlen sich nicht ausreichend auf die praktischen Anforderungen des Arztberufes vorbereitet. Ziel „ergebnisorientierter Ausbildung“ ist es, medizinische Curricula von den am Ende des Studiums notwendigen Kompetenzen ausgehend zu strukturieren und so die Lücke zwischen nötigen und erworbenen Fähigkeiten der Absolventen zu reduzieren. Dazu dienen insbesondere Lernzielkataloge (outcome frameworks OF). Als Vorarbeit zur Entwicklung des Nationalen Kompetenzbasierten Lernzielkatalogs Medizin (NKLM) – des deutschen OF - wurde 2011 das „Konsensusstatement Praktische Fertigkeiten im Medizinstudium“ publiziert, welches Lehre und Erwerb praktischer Fertigkeiten in Deutschland strukturiert und größtenteils in das Kapitel „Klinisch-Praktische Fertigkeiten“ des NKLM einfloss. Von der Definition über die Implementation eines OF bis zum Abschluss der Studierenden vergeht oftmals jedoch wenigstens ein Jahrzehnt, was die Lücke zwischen nötigen und erworbenen Kompetenzen wieder vergrößern kann. Ziel dieser Arbeit ist es daher, Thesen zur zukünftigen Entwicklung des Gesundheitswesens zu generieren und auf ein bestehendes OF anzuwenden. Methodik: Mittels halbstrukturierter Experteninterviews wurden Thesen zur allgemeinen, zukünftigen Entwicklung im Gesundheitswesen generiert. Diese Thesen wurden durch ärztliche Experten auf die Wahrscheinlichkeit ihres Eintretens bis zum Jahr 2025 hin bewertet. Die 288 Lernziele des Konsensusstatements, wurden von derselben Expertenkohorte in einer 2-stufigen, explorativen Delphi-Befragung auf ihre mittelfristige Relevanz für die ärztliche Ausbildung hin bewertet. Ergebnisse: Es wurden 11 Thesen zur Entwicklung der Medizin generiert und von 738 Experten bewertet und diskutiert. Die Thesen umfassen die Zunahme altersassoziierter Erkrankungen, die zunehmende Bedeutung von interprofessioneller Zusammenarbeit sowie eine steigende Verbreitung von telemedizinischen Anwendungen. Angewendet auf die 288 Lernziele des Konsensusstatements wurden 231 dieser Ziele als relevant und 57 als irrelevant für die mittelfristige Zukunft bewertet. Diskussion: Die in dieser Studie generierten und durch zahlreiche Experten validierten Thesen zur Zukunft des Gesundheitswesens liefern Anhaltspunkte zur zukünftigen Entwicklung der Anforderungen an Absolventen der Medizin insgesamt. Exemplarisch angewandt auf das Konsensusstatement praktische Fertigkeiten als inhaltgebendes Kapitel „Klinisch-Praktische Fertigkeiten“ des NKLM validieren sie dessen zukünftige Relevanz in großen Teilen, während sie gleichzeitig Hinweise für dessen Weiterentwicklung geben. Skillspractical skillsclinical skillsmedical trainingconsensus methodDelphi surveylearning goalsoutcomescompetenciesNKLM ==== Body Introduction On the one hand, the significance of obtaining practical skills during undergraduate medical studies has increased significantly in recent years [1], [2]. On the other hand, 64.7% of young medical professionals in Germany state that they do not feel adequately prepared for the practical requirements of the medical profession [3], a figure which is startlingly high, even compared to international data [4], [5]. Possible causes identified by the graduate survey in Cologne (Stosch C et al., unpublished) and a national survey (partially published in [6]), were both the narrow scope of practical training and the inadequate or lacking integration of this training in curricula and examinations. In order to bridge the gap between education and training, medical curricula are increasingly oriented toward national framework curricula, known as “outcome frameworks” (OF) [7], [8], which – generally speaking – describe the skills and knowledge which should be obtained during a training period in a competence-oriented fashion. Various outcome definitions exist internationally [9], [10], [11]. The Tuning Project [12] in Europe is an attempt to synchronize the many national OFs currently in existence. The German “National Competence-Based Catalogue of Learning Objectives for Medicine” (NKLM) was developed by the medical faculty association in cooperation with the Society for Medical Education (GMA) [13] and was initially published in June of 2015 after a six-year period of development [14]. In preparation for developing the NKLM, the “Consensus Statement of Practical Skills in Undergraduate Medical Education” was developed by the committee for practical skills of the GMA in 2011 [15]. This consensus statement “can and should have a formative effect on faculties to adjust their curricula in accordance with guidelines” [15] and strongly influenced the “Clinical-Practical Skills” chapter of the NKLM. The recommendations of the consensus statement have been implemented and validated within at least one faculty department [16]. In addition, the statement serves to assist the simulator network – a merger of the DACH region Skillslabs – to structure its simulator database [17]. There are, however, notable differences in content and structure between different OFs [18], [19], which raises the question of which OF should reasonably be referenced for teaching proficiency. In addition, developing medical curricula is generally a lengthy process: the six stages of the Kern cycle as a widely taught model of curriculum development [20], for example, require a considerable period of time between the initial definition of requirements, implementation, evaluation, and adaptation. Furthermore, an average of 6.4 years [21] elapse between beginning undergraduate medical education and beginning to practice medicine [21]. This contrasts starkly with rapid developments in medicine and the use of new technologies which have become ubiquitous. Consequently, there is a risk that the contents of curricula developed based on current OFs are no longer up-to-date when the medical professionals educated accordingly enter the medical profession. 1. Object of the Study The object of this study is to examine the “Consensus Statement of Practical Skills in Undergraduate Medical Education,” and thus an important part of the NKLM, for medium-term sustainability. The results should, on the one hand, serve to provide details for the further development of the NKLM; and on the other hand, help enhance the future stability of OF and curricula by means of overarching trends in healthcare which must yet be identified. The applied explorative Delphi method, as well as its results, can also serve to further develop local and national curricula. 2. The explorative Delphi method Originally developed in the 1950s as a technique for exploring technical developments in a military context [22], this method had been continually developed in the intervening decades [23] and is now considered an established method for analyzing uncertain developments and identifying strategic treatment options [22], [24], [25]. In principle, the Delphi method serves to collect group opinions and to focus group communications [24], as well as to qualitatively and quantitatively assess uncertain facts [24]. Although widely varied definitions of the Delphi method exist [24], certain common basic principles can be identified: anonymity of experts, multiple repetitions of the survey, statistical summary of group opinion, and controlled feedback [22]. The use of the Delphi method has been tested in various contexts, though here it is predominantly of interest to sufficiently documented applications in medical education research, such as for developing guidelines [26], [27], [28], [29]. Methods The project was structured into preparatory and working phases. During preparations, literary research followed by partially structured stakeholder interviews was used to develop theses about developments in healthcare. These theses were then assessed by means of an expert survey. In addition, the same expert cohort assessed the 288 learning goals of the “Consensus Statement of Practical Skills in Undergraduate Medical Education” within the framework of a 2-level, explorative Delphi survey. The course of the study is depicted in Figure 1 (Fig. 1). 1. Preparatory Phase: Theses on healthcare development Guidelines for partially structured interview with various healthcare practitioners were developed by means of selective literature research. The topics discussed in the interview included the following: The future development of healthcare Potential changes to care and to the disease spectrum Changes in medical technology and telemedicine Interdisciplinarity and cooperation with other occupational groups Future changes to undergraduate medical education Medical occupations in Germany and abroad Medical skills needed in the future During the preparatory phase, 9 interviews were conducted with experts in the fields of public health, medical technology and pedagogy, clinical and outpatient, practical skills, and students of human medicine (cf. Table 1 (Tab. 1) for details). Interview partners were chosen by means of a “purposive sampling strategy” [30] with the goal of obtaining as broad a spectrum of perspectives on these topics as possible. All interviewed experts were prepared for the interview. Interviews lasted an average of 22 minutes. The interviewees’ answers were then grouped by topic using a qualitative text analysis and summarized as theses by means of inductive categorization per Mayring [31] by an interdisciplinary research group comprised of two students of futures studies, including one nurse and one student of human medicine with paramedic training, one practicing physician, and one computer scientist. The goal of the Mayring analysis is to systematically process the written communications at hand, and to identify similarities and differences [32]. The principles of categorization were a) category selectivity, and b) a high level of abstraction of the same. 2. Working Phase: Expert interviews on future theses generated, and on consensus statement learning goals After generating theses, their probability of occurrence was assessed within the framework of an expert interview. The individual learning goals of the “Practical Skills in Undergraduate Medical Education” consensus statement were then assessed by the same experts within the framework of a two-level Delphi study. Physicians in all German medical university hospitals whose email addresses were available on the internet, as well as established physicians, were contacted via email to request their participation in the study. In addition to these 8,000 physicians contacted, others were approached at conferences (e.g., the Skills-Lab Symposium 2012) to request their participation in the study. Each participant then assessed the probability of occurrence of these theses on the future of medical care (generated during the preparatory phase) using a 4-level Likert scale (1 – very likely to 4 – very unlikely). Afterward, each participant was then assigned randomly to a group of ten in order to assess the future relevance of the consensus statement learning goals. This statement defines 288 learning goals assigned to one of 16 organ systems. There is a statement for three different training stages (“clinical traineeship, practical year, advanced training”) based on a three-tiered scale for each learning goal, to what extent this should be mastered (“seen demonstrated, performed under supervision, performed repeatedly”), and the survey further distinguishes between core and elective goals [15]. Each group was asked to assess a portion of the consensus statement learning goals (ca. 30/group) vis-à-vis their relevance for general medical training up to completing undergraduate medical education in the year 2025 using a 4-level Likert scale (1 – highly relevant to 4 – not at all relevant). The learning goal assessments were depicted based on the degree of mastery required by the advanced training stage, as stipulated in the consensus statement. After round 1 assessments and individual review of the results by the research group, the round 2 learning goals to be assessed were determined by consensus. Selection criteria included a wide distribution of assessment and the estimated significance of each learning goal as assessed in round 1. The learning goals were then re-evaluated by physicians participating in round 2, who were provided with the result of round 1 evaluations. Registered participants were assigned randomly to two groups for this purpose. Each group re-evaluated circa 50 learning goals. 3. Data evaluation The online interview was conducted using LimeSurvey (http://www.limesurvey.org). Data evaluation was conducted with Microsoft Excel (Microsoft Corporation, Redmond, WA, USA) and IBM SPSS Statistics 21.0 (SPSS Inc., Chicago, IL, USA). Results 1. Participants 738 experts registered online for round 1 of the study, and 594 complete data sets (19.5% dropout rate) were usable. 314 experts were registered for round 2, and 188 complete data sets (40.1% dropout rate) were usable. Since the learning goals were assessed within the context of the organ system assigned to them, we only took complete data sets into consideration. Participants were only asked to assess theses about the future during round 1. Partially-completed questionnaires were taken into consideration here, as the theses about the future can logically be interpreted on an individual basis. For this purpose, 651 expert opinions (11.8% dropout rate) were available. A large portion of experts in round 1 had more than one year of work experience (96.0%), and 137 experts had been practicing for more than 15 years. The overwhelming majority of physicians worked on an inpatient basis in maximum-care hospitals (87.9%). Study participants represented a total of 26 disciplines. The number of experts in round 2 was smaller, though the characteristics of their working environments were similar. In comparison to German Medical Association (BÄK) statistics from 2014, the proportion of inpatient physicians is large (88.6% in round 1 and 79.3% in round 2, compared to 51.0% by BÄK figures), and the exact distribution of specialties also differed somewhat. Overall, 57.8% (round 1) and 53.7% (round 2) of experts work in disciplines such as surgery, internal medicine, or anesthesia, in addition to general medicine. This is consistent with BÄK data, which lists this figure at 48.8% [http://www.bundesaerztekammer.de/fileadmin/user_upload/downloads/pdf-Ordner/Statistik2014/Stat14AbbTab.pdf]. An exact analysis of the participant cohort in comparison to BÄK figures from 2014 is depicted in Table 2 (Tab. 2). 2. Theses on the future of healthcare A total of 11 theses on the future of health care were derived from the partially structured interviews (cf. Table 3 (Tab. 3)). The theses on the future of healthcare trends generated in the preparatory phase are listed below. It was assumed that these theses were listed in accordance with expert assessment (cf. Table 3 (Tab. 3) and Figure 2 (Fig. 2)): Aspects of managing dementia are becoming significantly more important in physician communication. Increasingly balanced patient-physician relationships are emphasizing non-authoritarian forms of discussion and reasoning. Increasing mechanization also poses a barrier to entry in the medical field: the relevance of purely manual skills is decreasing, yet IT technologies still cannot determine medical history or diagnoses, requiring the skills of doctors. Diagnostics and patient monitoring will lead to less physical contact, and patients prefer internet and smartphones for this purpose. The physician remains the personal point of contact in established practice concepts. Duties once performed purely by physicians are, however, increasingly being delegated or substituted. Mobile treatment concepts of primary care are not catching on. Business economic considerations are slowly moving into focus: while business economic and organizational aspects are included in training, patients’ financial concerns also play a role in the type of care and treatment. 3. Assessment of learning goals 288 learning goals were assessed by experts in the 2 rounds of the study. The average of all expert assessments was used to determine whether a learning goals was deemed relevant or irrelevant. In round 1 of the Delphi study, 240 learning goals were assessed as relevant or highly relevant (average<2.5) and 57 were assessed as somewhat relevant or irrelevant (average>2.5), while 1 learning goal was assessed as neither (average=2.5). After reviewing the results of round 1, 103 learning goals were selected for assessment in round 2 based on the distribution of round 1 results; of these, 71 were assessed as relevant, 31 were assessed as irrelevant, and 1 was assessed as neutral in round 1. In round 2, experts assessed 62 learning goals as relevant and 41 as irrelevant. In comparing the two rounds, 13 learning goals (12.6%) were rated less relevant and 4 (3.9%) were rated as more relevant in round 2. A total of 231 learning goals were considered relevant and 57 learning goals were considered irrelevant. Figure 1 (Fig. 1) depicts an overview of these results. A more extensive review based on organ systems revealed that a large portion of learning goals for the sensory organ system (65.0%) were assessed as irrelevant for the future. Likewise, the future relevance of numerous learning goals associated with the skin, urogenital, and GI tract organs (≥30.0% each) was called into question. 31 of the 55 (56.4%) elective learning goals were assessed as irrelevant for the future – on the other hand, more than half (54.4%) of non-relevant learning goals are elective. Only 8% of the learning goals and skills which should be mastered upon completing advanced training were assessed as irrelevant for the future, yet 42% of skills were assessed at the lowest level of skill (“seen demonstrated”). The exact results are depicted in Table 4 (Tab. 4). The online appendix of this study shows an overview of all consensus statement learning goals and their assessments in the two rounds of the Delphi study. Discussion This paper attempts, on the one hand, to anticipate future global requirements for medical school graduates; and on the other, to concretely examine the future relevance of practical medical skills as an example of a limited scope of competency in undergraduate medical studies. For the latter, the learning goals of the consensus statement of practical skills in undergraduate medical studies was assessed by means of an explorative Delphi study as preliminary work to the NKLM “clinical-practical skills” chapter. The panel of experts in the Delphi study possesses many years of work experience and represents nearly all medical specialties. The majority of experts work in maximum-care hospitals, including university hospitals in which undergraduate training is primarily conducted in Germany. Though the experts’ more intensive knowledge has influenced the contents and requirements of the study, outpatient physicians (who are needed primarily for contributing opinions on these theses about the future of medical care) are underrepresented. We can only speculate on the reasons for which so few outpatient physicians participated in the study. At the beginning of this paper, 11 theses for relevant topics for future medical education were identified. We deliberately chose not to provide an OF as a basis for the interviews, as national OFs differ substantially in structure [18] as well as content [19]. An American group has already published a similar approach [33] which did not, however, take any further validation steps for its theses in comparison to our study. Below, we will discuss a few theses of this study in the context of their assessment by participating experts and derive implications for medical education: The current demographic shift has caused an increase in diseases associated with advanced age, such as mild cognitive impairment and dementia [34], [35]. The learning goals catalog lists two learning goals which could be attributed to dementia illnesses. Determining the medical history of elderly patients and performing simple test procedures such as geriatric assessments or falling risk tests were both assessed as relevant for the future. However, emphasizing geriatric test procedures was listed in the catalog as an elective goal, though it should be a core learning goals according to the experts in this study. The state of medical care, especially in rural areas with their own specific requirements [36], is in need of improvement [37]. Experience obtained in these places during voluntary training during undergraduate medical studies seemed to have a positive effect on students’ learning and career choices [38] and could improve primary care. The mandatory primary care physician clinical traineeship [39] was recently introduced, and a GMA position paper emphasizes the significance of primary care during undergraduate medical studies [40]. In addition to these structural changes, working on an (interprofessional) team and using telemedicine or E-health applications will become more important in the future: delegating work to non-medical personnel increases the effectiveness of primary care [41], [42]. At the same time, learning goals which do not apply solely to physicians, such as applying plaster casts or demonstrating functional taping, have also been assessed as relevant for the future. Beginning to promote interprofessional cooperation during undergraduate medical studies, such as combined courses with trainees or students of other medical care professions, could be one method of implementing the interpersonal aspect of care more intensively during medical education. In addition, electronic support system (health information technology) resources that are currently available could be used more effectively [42], [43]. Some positive effects of this technology, such as increased activity for COPD patients [44] or improved control of chronic asthma symptoms [45], have already been demonstrated. Integrating this growing field into training and education seems crucial, and could take place by means of telemedicine modules [46]. A dedicated learning goals catalog for e-health and telemedicine has already been published and can be consulted for future developments [47]. In consideration of this knowledge and the high relevance of soft skills and communication ability in this Delphi study, telephone- or internet-based physician-patient interaction could also grow more significant [48]. In order to do this field justice, training for communication skills (e.g., via telephone) should be intensified [49], as has been implemented in individual cases [50], [51]. Older patients in Germany see telemedicine methods more critically, however, and miss personal contact with and direct feedback from their physician [52]. Prioritization of core and elective learning goals in the OF original publication [15] (which were, in part, determined by means of the Delphi methodology and our results) mutually validate each other. More than 90% of the consensus statement learning goals assessed as needing to be mastered and nearly all border area learning goals were assessed by our participants as especially relevant for the future of the medical profession. On the other hand, more than 50% of skills listed in the consensus statement as elective were also assessed as less relevant by our study. The experts assessed practical learning goals overwhelmingly as relevant for the future, primarily in the large categories of communication skills, soft skills, interinstitutional skills, cardiovascular, and emergency, but also in narrower disciplines such as mental health and the endocrine system. The portion of learning goals not relevant for the future is largest for the sensory organs and for the urogenital and GI tract systems. This could be related to the choice of experts, but could also be due to the fact that the learning goals were phrased in a very specific manner, and thus there are a great many of them. In other catalog categories, more learning goals tended to be summarized as one, which made it difficult for experts to provide a differentiated assessment. In addition, it cannot be determined whether rejection of a learning goal was due to a general lack of future relevance, or because experts considered the learning goal relevant for the future, but believed that the learning goal should be a part of specialty training rather than general medical training. In a further step, the detailed results of this study (cf. online appendix) could be used, just like other validation studies [16], to re-assess the individual learning goals of the consensus statement and the NKLM, which would contribute to a review of the consensus statement and the NKLM. Preparing future physicians to practice medicine should be done, in parallel, on as many levels as possible. Numerous OFs, unlike this study, currently name the “self-directed learning” method resulting from “self-assessment” [53] as a significant method of improving the results of undergraduate medical studies [54]. At the same time, there are significant doubts as to the accuracy of self-assessments [53], [55], [56]. Thus, “lifelong learning” based on self-assessment cannot be the only methods of anticipating and addressing future developments. This should be done at the level of the OF. In addition to instructions on effective, self-directed, and lifelong learning, optimizing current OFs could make significant contributions. Anticipating future developments in conjunction with current research results could, on the one hand, provide important incentives for new content, and on the other hand, explorative Delphi studies could examine current learning goals and OFs with respect to their sustainability and possibly identify deficiencies. In concrete terms, the results of the Delphi study could serve to justify specific revisions to and implementation of the NKLM in different departments. Broad trend-setting decisions on possible future trends in undergraduate medical education can be derived from the assessment of these theses on the future. 1. Limitations Cognitive bias must be accepted as a significant limitation of any expert survey. This is of particular importance for the application of the explorative Delphi method for assessing issues that are uncertain, per se, as they take place in the future, since the line between rational assessment and the experts’ personal desires or fears could be blurred [57]. The structure of the survey could also have influenced expert opinions. After first assessing the likelihood of implementing theses in the future, experts were then asked to assess the future relevance of learning goals. This could have led to a bias. Though experts were asked to base their assessments on general education leading up to the medical exam, it cannot be determined whether this was actually done, and to what extent experts assessed general education as opposed to specialty training. The study population of this paper is comprised primarily of inpatient physicians at maximum-care facilities. Consequently, a bias against outpatient treatment methods cannot be ruled out. In addition, colleagues practicing general medicine were underrepresented (just 2.9% of experts), while anesthesiology and intensive care – two highly specialized subjects – were overrepresented, which could explain the strong emphasis on technological trends in the theses generated. One possible explanation for the large number of anesthesiology practitioners represented in the study could be their disproportionate involvement in conveying practical skills. A follow-up survey with primarily outpatient physician seems reasonable. Conclusions The explorative Delphi method provides an adequate opportunity to allow a current outcome framework to be assessed by experts on the basis of its future relevance. In addition, future trends can be anticipated by means of generating and assessing theses. It is important to continually review and adapt current OF and curricula to future developments in order to provide optimal preparation for medical studies graduates for their future daily professional lives. Acknowledgements We would like to thank Sascha Dannenberg for his help and support with technical questions about futures studies, and for his assistance in carrying out the study. In addition, we would like to thank Nicole Ambacher and Daniel Knapp for their support in data collection. Finally, we would like to sincerely thank all of the experts who took part in this study. Data Data for this article are available from the Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.q4sc8 [58] Competing interests The authors declare that they have no competing interests. Supplementary Material Online attachment – only in german Table 1 Overview of interview partners Table 2 Work experience, environment, and specialties for participants in rounds 1 and 2. Table 3 Expert responses (round 1) to the 11 stakeholder theses Table 4 Results of Delphi Rounds 1 + 2 Figure 1 Study design and results overview. Figure 2 Graphical depiction of the experts’ assessment of the 11 theses on the future of healthcare. See tab. 3 for allocation of these theses. ==== Refs 1 Nicky Hudson J Tonkin AL Clinical skills education: outcomes of relationships between junior medical students, senior peers and simulated patients Med Educ 2008 42 9 901 908 10.1111/j.1365-2923.2008.03107.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2008.03107.x 18694405 2 Yudkowsky R Otaki J Lowenstein T Riddle J Nishigori H Bordage G A hypothesis-driven physical examination learning and assessment procedure for medical students: initial validity evidence Med Educ 2009 43 8 729 740 10.1111/j.1365-2923.2009.03379.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2009.03379.x 19659486 3 Ochsmann EB Zier U Drexler H Schmid K Well prepared for work? Junior doctors' self-assessment after medical education BMC Med Educ 2011 11 99 10.1186/1472-6920-11-99 Available from: http://dx.doi.org/10.1186/1472-6920-11-99 22114989 4 Cave J Woolf K Jones A Dacre J Easing the transition from student to doctor: how can medical schools help prepare their graduates for starting work? Med Teach 2009 31 5 403 408 10.1080/01421590802348127 Available from: http://dx.doi.org/10.1080/01421590802348127 19142797 5 Goldacre MJ Taylor K Lambert TW Views of junior doctors about whether their medical school prepared them well for work: questionnaire surveys BMC Med Educ 2010 10 78 10.1186/1472-6920-10-78 Available from: http://dx.doi.org/10.1186/1472-6920-10-78 21070622 6 Damanakis A Ein kleiner Statusreport zu Skills Labs in der D-A-CH-Region 2011 Jahrestagung der Gesellschaft für Medizinische Ausbildung (GMA) 05.-08.10.2011 München Düsseldorf German Medical Science GMS Publishing House Doc11gma133 10.3205/11gma133 Available from: http://dx.doi.org/10.3205/11gma133 7 Harris P Snell L Talbot M Harden RM Competency-based medical education: implications for undergraduate programs Med Teach 2010 32 8 646 650 10.3109/0142159X.2010.500703 Available from: http://dx.doi.org/10.3109/0142159X.2010.500703 20662575 8 Busing N Slade S Rosenfield J Gold I Maskill S In the spirit of Flexner: working toward a collective vision for the future of medical education in Canada Acad Med 2 2010 85 2 340 348 10.1097/ACM.0b013e3181c8880d Available from: http://dx.doi.org/10.1097/ACM.0b013e3181c8880d 20107365 9 Bloch R Burgi H The Swiss catalogue of learning objectives Med Teach 3 2002 24 2 144 150 10.1080/01421590220120759 Available from: http://dx.doi.org/10.1080/01421590220120759 12098433 10 Frank JR Danoff D The CanMEDS initiative: implementing an outcomes-based framework of physician competencies Med Teach 2007 29 7 642 647 10.1080/01421590701746983 Available from: http://dx.doi.org/10.1080/01421590701746983 18236250 11 Metz JC 'Blueprint 1994': common objectives of medical education in The Netherlands Neth J Med 1999 55 4 165 167 10.1016/S0300-2977(99)00067-4 Available from: http://dx.doi.org/10.1016/S0300-2977(99)00067-4 10555432 12 Cumming A Ross M The Tuning Project for Medicine-learning outcomes for undergraduate medical education in Europe Med Teach 2007 29 7 636 641 10.1080/01421590701721721 Available from: http://dx.doi.org/10.1080/01421590701721721 18236249 13 Hahn EG Fischer MR Nationaler Kompetenzbasierter Lernzielkatalog Medizin (NKLM) für Deutschland: Zusammenarbeit der Gesellschaft für Medizinische Ausbildung (GMA) und des Medizinischen Fakultätentages (MFT) GMS Z Med Ausbild 2009 26 3 Doc35 10.3205/zma000627 Available from: http://dx.doi.org/10.3205/zma000627 14 Fischer MR Bauer D Mohn K NKLM-Projektgruppe Finally finished! National Competence Based Catalogues of Learning Objectives for Undergraduate Medical Education (NKLM) and Dental Education (NKLZ) ready for trial GMS Z Med Ausbild 2015 32 3 Doc35 10.3205/zma000977 Available from: http://dx.doi.org/10.3205/zma000977 26677513 15 Schnabel KP Boldt PD Breuer G Fichtner A Karsten G Kujumdshiev S Schmidts M Stosch C A consensus statement on practical skills in medical school - a position paper by the GMA Committee on Practical Skills GMS Z Med Ausbild 2011 28 4 Doc58 10.3205/zma000770 Available from: http://dx.doi.org/10.3205/zma000770 22205916 16 Blaum WE Dannenberg KA Friedrich T Jarczewski A Reinsch AK Ahlers O The practial use of the consensus statement on practical skills in medical school-a validation study GMS Z Med Ausbild 2012 29 4 Doc58 10.3205/zma000828 Available from: http://dx.doi.org/10.3205/zma000828 22916084 17 Damanakis A Blaum WE Stosch C Lauener H Richter S Schnabel KP Simulator Network project report: a tool for improvement of teaching materials and targeted resource usage in Skills Labs GMS Z Med Ausbild 2013 30 1 Doc4 10.3205/zma000847 Available from: http://dx.doi.org/10.3205/zma000847 23467581 18 Hautz SC Hautz WE Feufel MA Spies CD Comparability of outcome frameworks in medical education: Implications for framework development Med Teach 2015 20 1 9 10.3109/0142159x.2015.1012490 Available from: http://dx.doi.org/10.3109/0142159x.2015.1012490 19 Hautz SC Hautz WE Keller N Feufel MA Spies C The scholar role in the National Competence Based Catalogues of Learning Objectives for Undergraduate Medical Education (NKLM) compared to other international frameworks GMS Ger Med Sci 2015 13 Doc20 10.3205/000224 Available from: http://dx.doi.org/10.3205/000224 26609287 20 Kern DE Curriculum development for medical education: a six-step approach 1998 Baltimore u.a. Johns Hopkins Univ. Press 21 Wijnen-Meijer M Burdick W Alofs L Burgers C ten Cate O Stages and transitions in medical education around the world: clarifying structures and terminology Med Teach 2013 35 4 301 307 10.3109/0142159X.2012.746449 Available from: http://dx.doi.org/10.3109/0142159X.2012.746449 23360484 22 Rowe G Wright G The Delphi technique as a forecasting tool: issues and analysis I J Forecast 1999 15 4 353 375 10.1016/S0169-2070(99)00018-7 Available from: http://dx.doi.org/10.1016/S0169-2070(99)00018-7 23 Grunwald A Modes of orientation provided by futures studies: making sense of diversity and divergence Eur J Futures Res 2013 2 1 24 Häder M Delphi-Befragungen: ein Arbeitsbuch 2009 2. ed. Wiesbaden VS 10.1007/978-3-531-91926-3 Available from: http://dx.doi.org/10.1007/978-3-531-91926-3 25 Gönder D KiTA 2030: Eine Delphi-Befragung zur Zukunft von Kindertageseinrichtungen in Deutschland 2014 Berlin Freie Universität Berlin, Institut Futur 26 Williams PL Webb C The Delphi technique: a methodological discussion J Adv Nurs 1994 19 1 180 186 10.1111/j.1365-2648.1994.tb01066.x Available from: http://dx.doi.org/10.1111/j.1365-2648.1994.tb01066.x 8138622 27 Houwink EJ Henneman L Westerneng M van Luijk SJ Cornel MC Dinant JG Vleuten Cv Prioritization of future genetics education for general practitioners: a Delphi study Genet Med 2012 14 3 323 329 10.1038/gim.2011.15 Available from: http://dx.doi.org/10.1038/gim.2011.15 22241093 28 De Villiers MR de Villiers PJ Kent AP The Delphi technique in health sciences education research Med Teach 2005 27 7 639 643 10.1080/13611260500069947 Available from: http://dx.doi.org/10.1080/13611260500069947 16332558 29 Hasson F Keeney S McKenna H Research guidelines for the Delphi survey technique J Adv Nurs 2000 32 4 1008 1015 11095242 30 Tavakol M Sandars J Quantitative and qualitative methods in medical education research: AMEE Guide No 90: Part II Med Teach 2014 36 10 838 848 10.3109/0142159X.2014.915297 Available from: http://dx.doi.org/10.3109/0142159X.2014.915297 24845954 31 Mayring P Qualitative Inhaltsanalyse. Grundlagen und Techniken 2010 11., aktualisierte und überarbeitete Auflage Weinheim, Basel Beltz Verlag 32 Flick U von Kardorff E Steinke I Qualitative Forschung. Ein Handbuch 2010 Orig.-Ausg. 8 Reinbek bei Hamburg Rowohlt-Taschenbuch-Verl 33 Pershing S Fuchs VR Restructuring Medical Education to Meet Current and Future Health Care Needs Acad Med 2013 88 12 1798 1801 10.1097/ACM.0000000000000020 Available from: http://dx.doi.org/10.1097/ACM.0000000000000020 24128642 34 Pater C Mild cognitive impairment (MCI) - the novel trend of targeting Alzheimer's disease in its early stages - methodological considerations Curr Alzheimer Res 2011 8 7 798 807 10.2174/156720511797633250 Available from: http://dx.doi.org/10.2174/156720511797633250 21592047 35 Langa KM Levine DA The diagnosis and management of mild cognitive impairment: a clinical review JAMA 201417 312 23 2551 2561 25514304 36 Weeks WB Wallace AE Rural-urban differences in primary care physicians' practice patterns, characteristics, and incomes J Rural Health 2008 24 2 161 170 10.1111/j.1748-0361.2008.00153.x Available from: http://dx.doi.org/10.1111/j.1748-0361.2008.00153.x 18397451 37 Rosenblatt RA Andrilla CH Curtin T Hart LG Shortages of medical personnel at community health centers: implications for planned expansion JAMA 2006 295 9 1042 1049 10.1001/jama.295.9.1042 Available from: http://dx.doi.org/10.1001/jama.295.9.1042 16507805 38 Barrett FA Lipsky MS Lutfiyya MN The impact of rural training experiences on medical students: a critical review Acad Med 2011 86 2 259 263 10.1097/ACM.0b013e3182046387 Available from: http://dx.doi.org/10.1097/ACM.0b013e3182046387 21169781 39 Bundesministerium für Gesundheit Erste Verordnung zur Änderung der Approbationsordnung für Ärzte vom 17. Juli 2012 Bundesgesetzbl 2012 I 34 1539 Available from: http://www.bgbl.de/xaver/bgbl/start.xav?start=%2F%2F*[%40attr_id%3D%27bgbl112s1539.pdf%27]#__bgbl__%2F%2F*[%40attr_id%3D%27bgbl112s1539.pdf%27]__1436264296102 40 Huenges B Gulich M Böhme K Fehr F Streitlein-Böhme I Rüttermann V Baum E Niebling WB Rusche H Recommendations for undergraduate training in the primary care sector-position paper of the GMA-Primary Care Committee GMS Z Med Ausbild 2014 31 3 Doc35 10.3205/zma000927 Available from: http://dx.doi.org/10.3205/zma000927 25228937 41 Altschuler J Margolius D Bodenheimer T Grumbach K Estimating a reasonable patient panel size for primary care physicians with team-based task delegation Ann Fam Med 2012 10 5 396 400 10.1370/afm.1400 Available from: http://dx.doi.org/10.1370/afm.1400 22966102 42 Chen PG Mehrotra A Auerbach DI Do we really need more physicians? Responses to predicted primary care physician shortages Med Care 2014 52 2 95 96 10.1097/MLR.0000000000000046 Available from: http://dx.doi.org/10.1097/MLR.0000000000000046 24309673 43 Urban E Ose D Joos S Szecsenyi J Miksch A Technical support and delegation to practice staff - status quo and (possible) future perspectives for primary health care in Germany BMC Med Inform Decis Mak 2012 12 81 10.1186/1472-6947-12-81 Available from: http://dx.doi.org/10.1186/1472-6947-12-81 22853799 44 Lundell S Holmner A Rehn B Nyberg A Wadell K Telehealthcare in COPD: a systematic review and meta-analysis on physical outcomes and dyspnea Respir Med 2015 109 1 11 26 10.1016/j.rmed.2014.10.008 Available from: http://dx.doi.org/10.1016/j.rmed.2014.10.008 25464906 45 Fathima M Peiris D Naik-Panvelkar P Saini B Armour CL Effectiveness of computerized clinical decision support systems for asthma and chronic obstructive pulmonary disease in primary care: a systematic review BMC Pulm Med 2014 14 189 10.1186/1471-2466-14-189 Available from: http://dx.doi.org/10.1186/1471-2466-14-189 25439006 46 Brockes MC Wirth F Schmidt-Weitmann S Battegay E Gerke W Implementierung des Moduls Klinische Telemedizin/E-Health" in das Medizinstudium der Universität Zürich GMS Z Med Ausbild 2010 27 1 Doc14 10.3205/zma000651 Available from: http://dx.doi.org/10.3205/zma000651 47 Rohrig R Stausberg J Dugas M GMDS project group Medical Informatics Education in Medicine" Development of national competency-based learning objectives "Medical Informatics" for undergraduate medical education Methods Inf Med 2013 52 3 184 188 10.3414/ME13-04-0001 Available from: http://dx.doi.org/10.3414/ME13-04-0001 23615960 48 Car J Sheikh A Telephone consultations BMJ 2003 326 7396 966 969 10.1136/bmj.326.7396.966 Available from: http://dx.doi.org/10.1136/bmj.326.7396.966 12727771 49 Car J Freeman GK Partridge MR Sheikh A Improving quality and safety of telephone based delivery of care: teaching telephone consultation skills Qual Saf Health Care 2004 13 1 2 3 10.1136/qshc.2003.009241 Available from: http://dx.doi.org/10.1136/qshc.2003.009241 14757786 50 Fisher J Martin R Tate D Hands on + hands free: simulated on-call interaction Clin Teach 2014 11 6 425 428 10.1111/tct.12180 Available from: http://dx.doi.org/10.1111/tct.12180 25212926 51 Dickinson M Pimblett M Hanson J Davis M Reflecting reality: pager simulations in undergraduate education Clin Teach 2014 11 6 421 424 10.1111/tct.12185 Available from: http://dx.doi.org/10.1111/tct.12185 25212925 52 Terschuren C Mensing M Mekel OC Is telemonitoring an option against shortage of physicians in rural regions? Attitude towards telemedical devices in the North Rhine-Westphalian health survey, Germany BMC Health Serv Res 2012 12 95 10.1186/1472-6963-12-95 Available from: http://dx.doi.org/10.1186/1472-6963-12-95 22507694 53 Eva KW Regehr G Self-assessment in the health professions: a reformulation and research agenda Acad Med 2005 80 10 Suppl S46 S54 10.1097/00001888-200510001-00015 Available from: http://dx.doi.org/10.1097/00001888-200510001-00015 16199457 54 Hautz SC Hautz WE Feufel MA Spies CD What makes a doctor a scholar: a systematic review and content analysis of outcome frameworks BMC Med Educ 2015 55 Davis DA Mazmanian PE Fordis M Van Harrison R Thorpe KE Perrier L Accuracy of physician self-assessment compared with observed measures of competence: a systematic review JAMA 2006 296 9 1094 1102 10.1001/jama.296.9.1094 Available from: http://dx.doi.org/10.1001/jama.296.9.1094 16954489 56 Li ST Paterniti DA Co JP West DC Successful self-directed lifelong learning in medicine: a conceptual model derived from qualitative analysis of a national survey of pediatric residents Acad Med 2010 85 7 1229 1236 10.1097/ACM.0b013e3181e1931c Available from: http://dx.doi.org/10.1097/ACM.0b013e3181e1931c 20592521 57 Gerhold L Standards und Gütekriterien der Zukunftsforschung: ein Handbuch für Wissenschaft und Praxis 2015 Wiesbaden Springer VS 10.1007/978-3-658-07363-3 Available from: http://dx.doi.org/10.1007/978-3-658-07363-3 58 Dannenberg KA Stroben F Schröder T Thomas A Hautz WE Data from: The future of practical skills in undergraduate medical education - an explorative Delphi-Study 2016 Dryad Digital Repository 10.5061/dryad.q4sc8 Available from: http://dx.doi.org/10.5061/dryad.q4sc8

PMC005xxxxxx/PMC5003135.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00106610.3205/zma001066Doc67urn:nbn:de:0183-zma0010665ArticleDidactic, practical, good! 20 years of clinical skills training in the German speaking countries Didaktisch, praktisch, gut! 20 Jahre Training praktischer Fertigkeiten in der D-A-CH-Region Stosch Christoph *1Schnabel Kai P. *21 Universität zu Köln, Medizinische Fakultät, Referat für Lehre, Studium & Studienreform, Kölner Interprofessionelle Skills Labs (KISS), Köln, Deutschland2 Universität Bern, Institut für medizinische Lehre, Abteilung für Unterricht und Medien, Bern, Schweiz*To whom correspondence should be addressed: Kai P. Schnabel, Universität Bern, Institut für medizinische Lehre, Abteilung für Unterricht und Medien, Konsumstraße 13, CH-3010 Bern, Schweiz, E-mail: kai.schnabel@iml.unibe.ch15 8 2016 2016 33 4 Clinical skillsDoc6713 7 2016 13 7 2016 13 7 2016 Copyright © 2016 Stosch et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001066.shtml ==== Body Editorial In electrical engineering, a transformer controls the level of an AC voltage. “Transformers” are mechanical beings from the planet Cybertron who can continuously transform their body shape. In genetics, transformation describes a process of transferring DNA to another cell, and in medicine, “malignant transformation” means the transition from normal to tumor cell growth. Social transformation refers to the fundamental metamorphosis of a political system into another; in architecture, it means the conversion of existing buildings, and the term “digital transformation” keeps popping up in talk shows. It almost seems as if "transformation" is a ubiquitous phenomenon inherent to the world of things which tags standstill as a failure. The world of medical education is no exception, although if looked at cursorily, the players’ willingness to embrace change may not always and not immediately be noticeable [https://de.wikipedia.org/wiki/Transformation cited 2016 Jan 24] One particularly successful major project in this area is documented in the present special issue “Clinical skills”: 20 years of development of "skills labs" in the D-A-CH region (Germany/Austria/Switzerland)! Clinical practical skills have been taught before, according to training regulations and laws. However, the fact that an impressive number of laboratories (from Latin laborare = to labor, work, suffer [https://de.wikipedia.org/wiki/Labor cited 2016 Jan 24] have been set up for the systematic and longitudinal training of skills can only be interpreted as part of a transformation process, i. e. the professionalization of medical education. 1996 marked the beginning of the establishment of learning centers for the teaching of patient-centered skills. A hospital rich in history, the Allgemeines Krankenhaus (General Hospital) of Vienna, was first, followed by Berlin (1999), Heidelberg and Frankfurt a.M. (2000), Cologne (2003), Mainz (2003), and others. The accompanying transformation of education in human medicine can be described in many ways: historically (when, what, why), economically (costs, how financed), outcome-oriented (objectives achieved, patients saved), or with regard to the supported processes of change in medical education: The transition from pure knowledge-based higher education to a competency-based orientation without systematic inclusion of skills training is simply unthinkable. The systematics inherent in action-oriented skills training (“see one – do one”) offers the transition “from teaching to learning” an inherent field for assisted self-regulation in the learning process by providing specific contexts, scenarios and – often formative – feedback. The democratization of knowledge through peer teaching often employed in the skills labs finds resonance in the medical faculties by virtue of the changed roles in the learning process. The training of practical skills can be seen as practiced, active patient protection. If form follows function, the transformation of the organization of the skills labs must be - and has been - understood as a response to the professionalization happening in medical education. While it often started as a students‘ initiative, full-time management is today’s standard. In the beginning, the skills part of the curriculum could often only be offered as an added option, but nowadays, curricular integration seems adequate. While at first, skills labs would be housed in buildings set for demolition on the edge of hospital grounds in recognition of the effort spent, new, designated buildings are now being built on prime real estate. Formerly, training centers for patient-centered skills developed one-dimensionally toward the fulfillment of tasks in the study of medicine, but today, they are in demand by continuing education and training programs both in medicine and other health professions; hence, they are becoming true nuclei of interprofessional education. While formerly, skills lab employees were without outside support, since 2007, they have been able to rely on an established Committee on Practical Skills of the Society for Medical Education which hosts an annual international skills lab symposium (iSLS) of up to 300 participants, and which may be considered as the engine that keeps the described transformation going. Besides providing a platform for scientific exchange, the iSLS focuses on skills lab staff‘s continuing education and networking, with meetings for the student employees, regular skills lab manager meetings (sLiT), and the Committee meetings ultimately serving the professionalization of skills labs in the D-A-CH region. There, the content of the “Consensus Statement on Practical Skills in Medical School” [1] was defined, which had in several respects an unmistakable influence on the development of the National Competence-Based Learning Objectives Catalog in Medicine [http://www.nklm.de cited 2016 April 14]: as a source of ideas for the formulation of “competence levels” and the “milestones of skill acquisition” as well as the basis of chapter 14b, “Clinical practical skills”. Currently, the accreditation regulations appropriate to higher education and concerning education, advanced and continuing training of clinical practical skills are being developed, to be become effective this year. Finally, this transformation raises the question of further development of centers for learning clinical practical skills. Although the future is open by its very nature, key questions can already be derived whose answers will shape tomorrow‘s discourse: Following “competence”, what metaphor will govern the discussion on further development of medical education, and what - deliberate - role will the skills labs play in it? What are the effects of the profanation of the skills labs accompanying their professionalization in the context of medical schools (success orientation, competition of means)? Furthermore, what is the role of hands-on experience (experience-based learning with models and simulators, and simulated patients) with regard to “augmented reality”, or, in other words: will the digital revolution transform the - as yet physical - skills labs into “holodecks” in the future? “They say that the ‘little man’ cannot achieve anything. But, if everybody finds another little man, many little men can achieve a lot.” (Thomas Häntsch (*1958), photographer [https://www.aphorismen.de/zitat/108609 cited 2016 July 5]. Competing interests The authors declare that they have no competing interests. ==== Refs 1 Schnabel KP Boldt PD Breuer G Fichtner A Karsten G Kujumdshiev S Schmidts M Stosch C Konsensusstatement "Praktische Fertigkeiten im Medizinstudium" - ein Positionspapier des GMA-Ausschusses für praktische Fertigkeiten. A Consensus Statement on Practical Skills in Medical School - a position paper by the GMA Committee on Practical Skills GMS Z Med Ausbild 2011 28 4 Doc58 10.3205/zma000770 Available from: http://dx.doi.org/10.3205/zma000770 22205916

PMC005xxxxxx/PMC5003136.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00105310.3205/zma001053Doc54urn:nbn:de:0183-zma0010537Article“Heidelberg standard examination” and “Heidelberg standard procedures” – Development of faculty-wide standards for physical examination techniques and clinical procedures in undergraduate medical education „Heidelberger Standarduntersuchung“ und „Heidelberger Standardprozeduren“ – Entwicklung fakultätsweiter Standards für körperliche Untersuchungstechniken und klinische Prozeduren im Medizinstudium Nikendei C. *1Ganschow P. 23Groener J. B. 4Huwendiek S. 56Köchel A. 1Köhl-Hackert N. 7Pjontek R. 28Rodrian J. 2Scheibe F. 29Stadler A.-K. 2Steiner T. 10Stiepak J. 11Tabatabai J. 12Utz A. 213Kadmon M. 141 Heidelberg University Hospital, University Medical Center, Internal Medicine II, Department of General Internal Medicine and Psychosomatics, Heidelberg, Germany2 Heidelberg University Hospital, Department of General, Visceral and Transplantation Surgery, Heidelberg, Germany3 University Hospital of Munich, Department of General, Visceral, Transplantation, Vascular and Thoracic Surgery, Munich, Germany4 Heidelberg University Hospital, University Medical Center, Internal Medicine I, Department of Endocrinology, Metabolism and Clinical Chemistry, Heidelberg, Germany5 University of Bern, Institute of Medical Education, Department of Assessment and Evaluation, Bern, Switzerland6 Heidelberg University Hospital, Center for Child and Adolescent Medicine, Clinic 1, Heidelberg, Germany7 Heidelberg University Hospital, Department of General Medicine and Health Services Research, Heidelberg, Germany8 University Hospital RWTH Aachen, Department of Diagnostic and Interventional Neuroradiology, Department of Neurology, Aachen, Germany9 Ortenau Hospital Offenburg-Gengenbach, Department of Cardiology, Pneumology, Angiology and Intensive Care Medicine, Offfenburg-Gengenbach, Germany10 Klinikum Frankfurt Höchst, Department of Neurology, Frankfurt/Main, Germany11 Heidelberg University Hospital, University Medical Center, Internal Medicine III, Department of Cardiology, Angiology and Pneumology, Heidelberg, Germany12 Heidelberg University Hospital, Department of Pediatrics I, Center for Child and Adolescent Medicine, Heidelberg, Germany13 Ortenau Hospital Offenburg-Gengenbach, Department of Gynecology, Offenburg-Gengenbach, Germany14 Carl von Ossietzky University of Oldenburg, School of Medicine and Health Sciences, Oldenburg, Germany*To whom correspondence should be addressed: C. Nikendei, Heidelberg University Hospital, University Medical Center, Internal Medicine II, Department of General Internal Medicine and Psychosomatics, Im Neuenheimer Feld 410, D-69120 Heidelberg, Germany, Phone: +49 (0)6221/56-38663, E-mail: christoph.nikendei@med.uni-heidelberg.de15 8 2016 2016 33 4 Clinical skillsDoc5429 9 2015 25 2 2016 22 2 2016 Copyright © 2016 Nikendei et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001053.shtmlThe competent physical examination of patients and the safe and professional implementation of clinical procedures constitute essential components of medical practice in nearly all areas of medicine. The central objective of the projects “Heidelberg standard examination” and “Heidelberg standard procedures”, which were initiated by students, was to establish uniform interdisciplinary standards for physical examination and clinical procedures, and to distribute them in coordination with all clinical disciplines at the Heidelberg University Hospital. The presented project report illuminates the background of the initiative and its methodological implementation. Moreover, it describes the multimedia documentation in the form of pocketbooks and a multimedia internet-based platform, as well as the integration into the curriculum. The project presentation aims to provide orientation and action guidelines to facilitate similar processes in other faculties. Zusammenfassung Die kompetente körperliche Untersuchung von Patientinnen und Patienten und die sichere und professionelle Durchführung klinischer Prozeduren stellen in nahezu allen medizinischen Fachbereichen einen essentiellen Bestandteil des ärztlichen Handelns dar. Die zentrale Zielsetzung der durch Studierende initiierten Projekte „Heidelberger Standarduntersuchung“ und „Heidelberger Standardprozeduren“ war die Etablierung eines einheitlichen interdisziplinären Standards zur körperlichen Untersuchung und zu klinischen Prozeduren und deren Distribution in Abstimmung mit allen klinischen Disziplinen an der Universitätsklinik Heidelberg. Der vorgestellte Projektbericht beleuchtet den Hintergrund der Initiative, die methodische Umsetzung und stellt die multimediale Dokumentation in Form von Kitteltaschenbüchern und einer multimedialen internetbasierten Plattform sowie die curriculäre Integration vor. Die Projektvorstellung soll eine Orientierung und Handlungsleitlinie bieten, um ähnliche Prozesse an anderen Fakultäten zu erleichtern. undergraduate medical educationphysical examinationclinical proceduresfaculty development ==== Body Introduction The competent physical examination of patients constitutes an essential component of medical practice in nearly all areas of medicine [1]. Together with history-taking, it leads to crucial action-guiding hints for a targeted diagnosis and treatment of the patient [2], [3]. Upon entering the profession, every physician therefore needs to have mastered physical examination to a high level [4]. The situation is similar with respect to the implementation of clinical procedures such as drawing blood, IV cannulation, or mask ventilation. Here, the thorough preparation of materials, knowledge about the order of relevant sub-steps, focused work under sterile conditions, and the correct implementation of the clinical-technical skill are crucial factors for the success of the treatment measure and for patient safety [5]. To achieve this, students not only have to develop the necessary complex psychomotor skills, but also learn the required communicative and affective competences in their interaction with patients, as well as the concluding assessment of findings. Investigations in medical students show that considerable deficits exist, both in conducting the physical examination and evaluating findings, and in carrying out clinical procedures [6], [7], [8], [9]. A study among students in the Practical Year (PY) at a German university revealed that in the physical examination of four important organ systems, only 40% of the students (thyroid 38%, heart 37%, lungs 42%, abdomen 43%) correctly implemented the a priori defined relevant sub-steps [10]. Likewise, in the interpretation of cardiac and pulmonary auscultation findings [11], pathological findings were only correctly classified in 20 - 45% of cases, which can be seen as posing a considerable danger for misdiagnosis [6]. The integration of focused physical examination techniques into complex clinical procedures such as ward rounds appears to be particularly marked by difficulties [7]. Even with respect to frequent clinical procedures such as IV cannulation, PY students show serious shortcomings, meaning that an independent implementation in the sense of an “entrustable professional activity” [8] is called into question even at the time when students are completing their medical education [9]. Physical examination techniques and clinical procedures are learned among the students themselves, with standardized patients, mannequins, part-task trainers, or directly at the patient’s bedside. With regard to the acquisition of physical examination competences, students prefer to learn with standardized patients or real patients; students only prefer learning with standardized patients for genital examinations [12]. Despite the continued widespread practice of the “see one – do one” approach [13], learning according to this approach appears to be ethically untenable for most clinical procedures due to their invasive nature. In this regard, in the framework of simulation-based medical education (SMBE) [14], simulation settings such as the “Skills Lab” [15] have established themselves and have proven to be effective for learning clinical procedures [5], [16], [17]. To enable this learning process to take place at a high level, continuous training using uniform faculty standards across all stages and areas of study is indispensable for the students. This is not least the case because a congruence has to exist between the assessment goals and criteria on the one hand, and the imparted learning goals and standards on the other, in the sense of a “constructive alignment” [18]. This should enable the learning incentives triggered by clinical assessments to be optimally utilized (“assessment drives learning”; [19]). The central objective of the project “Heidelberg standard examination” was to establish uniform interdisciplinary standards for physical examination and clinical procedures in coordination with all clinical disciplines at the Heidelberg University Hospital. Further aims were to provide multimedia materials which would be available to assist all students from the first to the last day of their medical training, and to integrate the examination standards into the curriculum. At the same time, to foster the implementation into the curriculum, all lecturers working in the area of clinical teaching should have the same multimedia materials at their disposal. The standardization of medical examination techniques and the integration of the new joint standards into the curriculum should meet the needs of students as frequently expressed in student evaluations, sustainably increase the quality of student training and the acquisition of competences, and improve patient care. The following sections describe this process of interdisciplinary development of the teaching standards and the integration into the curriculum. Methods Project initiation and project team The project arose in 2007 from a student initiative which the last author (MK) assisted and supervised with regard to content. Initially, a central objective of the project was to provide films on physical examination and clinical procedures created by students for students. These should serve as a helpful orientation in the preparation for clinical-technical examinations (objective structured clinical examination – OSCE; [20]) at the Heidelberg Medical Faculty. This student initiative led to plans to establish faculty-wide standards for these areas. As it developed further, the project was medically guided mainly by the University Hospital Department of Surgery (MK) and the University Medical Center (CN) and the project plans were implemented within a project and editorial team (co-authors). Tasks of the project and editorial team The composition and personal strengths of the project and editorial team varied over the running time of the project. Mostly, two to three medical personnel from the Departments of Surgery and Internal Medicine acted together to develop the project plans further. The tasks of the project and editorial team included coordinating team meetings, producing checklists, leading the interdisciplinary meetings with subject experts, developing layout suggestions, creating graphics and pictures, and implementing the texts in InDesign©. With regard to the filming of physical examination techniques and clinical-technical procedures, the project team was responsible for coordinating the film shoots, cutting and sound, and the final inspection by the experts. Advertising, clarification of legal aspects and funding applications should also not be forgotten. Faculty-wide interdisciplinary coordination of learning goals and establishment of teaching standards In an iterative coordination process between the subject areas of general medicine, ophthalmology, surgery, orthopedic medicine and trauma surgery, dermatology, geriatrics, gynecology, ENT medicine, emergency medicine, hygiene, internal medicine, neurology, neurosurgery, pediatrics, psychiatric and psychosomatic medicine, pathology, forensic medicine and urology, authorized medical representatives of the individual disciplines (“subject experts”) agreed on a uniform process for a standardized “head-to-toe” basic examination of entrusted patients, which was valid across all faculties. Moreover, for the individual physical examination techniques as well as clinical procedures which are used in several disciplines, all subject experts from these disciplines were included in the coordination process for agreeing upon learning goals and establishing teaching standards. For those physical examination measures or procedures which are only conducted in one of the disciplines, the respective subject expert developed, in coordination with his clinical department, the learning goals and the teaching standards to be established together with the project and editorial team. The iterative coordination process completed in each case is illustrated in figure 1 (Fig. 1). Multimedia documentation and distribution of the faculty-wide standards In consultation with the Dean of Studies of the Medical Faculty of the University of Heidelberg, the project and editorial team resolved and strived to distribute the developed teaching standards for physical examination techniques (“Heidelberg standard examination”) and for clinical procedures (“Heidelberg standard procedures”) in the form of a book and likewise to make them accessible in the form of film material. In this respect, each one of the 1,926 medical students of the Heidelberg Medical Faculty should be provided with these standards at no charge in the form of a pocketbook, and the films should equally be freely accessible on a homepage, together with the text and picture material in the book and the contents of the book chapters. At the same time, all medical lecturers involved in teaching should also receive a copy of the pocketbook. Integration into the curriculum The “Heidelberg standard examination” – represented by the developed pocketbook and the corresponding film sequences – was to be integrated into the curriculum from the preclinical part of studies up to the final examinations. To achieve this, a close consultation regarding the joint standards occurred with the course leaders of all existing examination courses: AaL-Plus (living anatomy) in the pre-clinical stage, interdisciplinary clerkship in surgery and internal medicine at the beginning of the clinical stage, clinical examination courses using standardized patients, bedside teaching (UaK) in all subject areas in the further clinical stage, and training in the clinical environment, self-directed learning and feedback conversations with the students during the practical year (PY). At the same, the clinical-technical examinations (OSCE; [21]), which had been revised and modified according to the faculty-wide standards, should serve the purpose of reviewing learning goals of the individual clinical subject areal. Funding Project funding was ensured by tuition fees and quality assurance funds of the Baden-Württemberg State Ministry for Sciences and Arts. From the beginning of the project in August 2009, two half-time physician positions as well as one student assistant position were continuously available, and from 2014 one half-time position for a film technician was provided. Moreover, the printing costs for the student copies of the pocketbooks “Heidelberg standard examination” and “Heidelberg standard procedures” (each approx. 35,000 €) as well as the costs for producing the homepage (approx. 8,000 €) were also covered. Sales of the pocketbooks by Heidelberg Medical Faculty also brought further funding opportunities for the project. The administration of the Heidelberg University Hospital pledged to fund the copies made available to the medical lecturers. Gathering and integrating feedback Feedback on the pocketbook “Heidelberg standard examination” can be provided through email addresses listed in the book. These user comments form the basis for a continuous revision and adaptation of the pocketbook. Results Pocketbook “Heidelberg standard examination” Figure 2 (Fig. 2) shows, by way of example, an excerpt from the pocketbook “Heidelberg standard examination” [22]. For presenting the physical examination techniques, a double-page format was chosen. Unique in its presentation, on the left-hand side there is a description of the relevant action steps, and on the right-hand side a comparison of possible normal and pathological findings. This provided the students, beyond the implementation of techniques, with support in using the correct nomenclature, description and documentation of normal and pathological findings. Additionally, “tips” were integrated, with advice on facilitating the practical procedure, warnings (“CAVE”) regarding complicating behavior, and boxes with clinically relevant information. Pocketbook “Heidelberg standard procedures” The graphical representation within the pocketbook “Heidelberg standard procedures” is illustrated in Attachment [23]. Besides a detailed description of the setting, the indications and contraindications for the respective clinical procedure, the required material, the relevant preparatory measures and the practical implementation are listed. The learner is also supported with “tips” and “CAVE” hints in the book. Internal and external distribution The first edition of the “Pocketbook Heidelberg standard examination” went to press in September 2012. This was distributed to all 1,926 students of the Medical Faculty, University of Heidelberg, as well as all 1,584 medical lecturers working in the faculty. The current financial basis allows all new students and medical lecturers to be equipped with a pocketbook in the subsequent three years. The external distribution is ensured through the homepage http://www.heidelbergerklinischestandards.de. We expressly chose not to cooperate with a textbook publisher in order to retain the rights for the contents and graphical material. In addition to sales of individual copies, 13 German and Austrian faculties were provided with large package orders. Besides the local provision of the Heidelberg students and lecturers, these external individual copy sales led to a turnover of approx. 500 books, and approx. 10,800 books were sold to other German-speaking faculties. The “Heidelberg standard procedures” pocketbook went to press in December 2015 [23]. Film hosting The films corresponding to the physical examination techniques and clinical procedures are presented on the homepages http://www.heidelbergerklinischestandards.de. All purchasers of the pocketbooks “Heidelberg standard examination” and “Heidelberg standard procedures” can freely access the film material. Sample sequences can be found at http://www.heidelbergerklinischestandards.de/projekte-video/untersuchungen/ and http://www.heidelbergerklinischestandards.de/projekte-video/prozeduren (see figure 3 (Fig. 3)). Integration into the curriculum Table 1 (Tab. 1) shows the time points in the curriculum of the University of Heidelberg Medical Faculty at which a student receives learning units on physical examination. Furthermore, the time points of the relevant clinical-technical examinations are given. Feedback Through the correspondence address provided in the pocketbook, abundant constructive criticism was given on the part both of the students and the lecturers of the Medical Faculty of Heidelberg University. These were consistently positive and referred to the didactic design (“[…]. The presentation…of this book is great”, You’ve created a super layout”), the quality of the content (“[…]. A brilliant work! Until now I had laboriously compiled the info on examinations myself. Now it’s been done for me...”) and the film realization (“In combination with the […] teaching videos it’s simply great!”) as well as points of criticism with regard to medical content. Discussion In the current project description, we presented the process of developing faculty-wide standards for clinical physical examination and clinical procedures at the University of Heidelberg Medical Faculty. We presented a description of the objectives, the logistical coordination of learning goals and development of the teaching standards, as well as the integration into the curriculum and distribution in the form of the pocketbooks “Heidelberg standard examination” [22] and “Heidelberg standard procedures” [23] and of the online film material, which is automatically accessible upon acquisition of the pocketbooks. Based on the local resonance and the re-evaluation of the project, the initiative can be deemed a success. The feedback which the project team received through the email addresses provided in the pocketbook is consistently positive. Constructive points of criticism regarding content and suggestions for improvement were gratefully accepted and incorporated in the 2nd edition of the pocketbook “Heidelberg standard examination”. Substantial orders from German and international German-speaking medical faculties, and strong demand for the pocketbook “Heidelberg standard examination” from private individuals also illustrate that the pocketbook and the corresponding film material close a gap in demand. A 2nd, improved edition of the pocket book “Heidelberg standard examination” has since been printed. Currently, the corresponding films are being produced and completed in the accompanying online offer. The pocketbook “Heidelberg standard procedures” went to press in December 2015. The film production for the presented procedures has been completed for this subject area and will be fully available when the book is distributed, and in turn made available to all students and medical lecturers of the University of Heidelberg Medical Faculty. Sample videos can be viewed at http://www.heidelbergerklinischestandards.de/projekte-video/prozeduren/. Further development of contents is planned in the coming years for the areas of complex clinical procedures (e.g. anesthesia induction, conducting ward rounds etc.). Besides the distribution of the multimedia implementation of the teaching standards for physical examination and clinical procedures, the integration into the curriculum and into assessment constitutes a challenge. This is implemented in the form of a longitudinal networking of learning contents on physical examination with clinical procedures, which spans from the first semester to the PY. The continuous needs assessment, adjustment of learning goals, evaluation and curriculum adaptation and quality assurance should certainly be seen in this regard as a long-term process in the sense of a Kern cycle [1]. Limitations In terms of limitations, it should be noted that the introduced university standards are based upon expert consensus as an evidence base is currently lacking. Moreover, although the feedback on the project is very positive, it was only requested unsystematically via email within the pocketbook. However, the focus and a potential strength of the current project report can be found in the detailed description of the creation process. This could enable other faculties with a similar intention to gather stimuli and orientation. A particular strength of the project itself is, in our view, the central incorporation of highly motivated students, who helped shape the project to a very high degree, and who were rewarded for this with the GMA prize for teaching students in 2010. Conclusions Overall, the whole process of developing uniform teaching standards for physical examination and clinical procedures has proven to be laborious but worthwhile. It has provoked an important curriculum-based stimulus, an interdisciplinary discourse regarding subject and content, and provides the students with greater security in terms of behavior and examination. Competing interests The authors declare that they have no competing interests. Supplementary Material Example excerpt from the pocketbook “Heidelberg standard procedures” Table 1 Integration into curriculum using the example of teaching of competences for physical examination Figure 1 Iterative coordination process on learning goals and teaching standards (simplified schematic representation) Figure 2 Example excerpt from the pocketbook “Heidelberg standard examination” Figure 3 Sample sequences of the film material for “Heidelberg standard procedures” presented on http://www.heidelbergerklinischestandards.de/projekte-video/prozeduren/ ==== Refs 1 Kern DE Curriculum Development for Medical Education: A Six-step Approach 1998 Baltimore Johns Hopkins University Press 2 Reilly BM Physical examination in the care of medical inpatients: an observational study Lancet 2003 362 9390 1100 1105 10.1016/S0140-6736(03)14464-9 Available from: http://dx.doi.org/10.1016/S0140-6736(03)14464-9 14550696 3 Ende J Fosnocht KM Clinical examination: still a tool for our times? Trans Am Clin Climatol Assoc 2002 113 137 150 12053706 4 Nardone DA Lucas LM Palac DM Physical examination: a revered skill under scrutiny South Med J 1988 81 6 770 773 10.1097/00007611-198806000-00019 Available from: http://dx.doi.org/10.1097/00007611-198806000-00019 3375881 5 Lund F Schultz JH Maatouk I Krautter M Moltner A Werner A Effectiveness of IV cannulation skills laboratory training and its transfer into clinical practice: a randomized, controlled trial PLoS One 2012 7 3 e32831 10.1371/journal.pone.0032831 Available from: http://dx.doi.org/10.1371/journal.pone.0032831 22427895 6 Ortiz-Neu C Walters CA Tenenbaum J Colliver JA Schmidt HJ Error patterns of 3rd-year medical students on the cardiovascular physical examination Teach Learn Med 2001 13 3 161 166 10.1207/S15328015TLM1303_5 Available from: http://dx.doi.org/10.1207/S15328015TLM1303_5 11475659 7 Nikendei C Kraus B Schrauth M Briem S Junger J Ward rounds: how prepared are future doctors? Med Teach 2008 30 1 88 91 10.1080/01421590701753468 Available from: http://dx.doi.org/10.1080/01421590701753468 18278658 8 Berberat PO Harendza S Kadmon M Gesellschaft für Medizinische Ausbildung GMAAfW Entrustable professional activities - visualization of competencies in postgraduate training. Position paper of the Committee on Postgraduate Medical Training of the German Society for Medical Education (GMA) GMS Z Med Ausbild 2013 30 4 Doc47 10.3205/zma000890 Available from: http://dx.doi.org/10.3205/zma000890 24282450 9 Koechel A Scarpone R Stiepak J Gröner J Herzog W Nikendei C Bugaj T CanMEDS-Tätigkeitsprofile von PJ-Studierenden der Inneren Medizin – eine prospektive Längsschnittstudie 2014 Jahrestagung der Gesellschaft für Medizinische Ausbildung (GMA) 25.-27.09.2014 Hamburg Düsseldorf German Medical Science GMS Publishing House DocP462 10.3205/14gma181 Available from: http://dx.doi.org/10.3205/14gma181 10 Krautter M Diefenbacher K Koehl-Hackert N Buss B Nagelmann L Herzog W Short communication: final year students' deficits in physical examination skills performance in Germany Z Evid Fortbild Qual Gesundhwes 2015 109 1 59 61 10.1016/j.zefq.2015.01.003 Available from: http://dx.doi.org/10.1016/j.zefq.2015.01.003 25839370 11 Mangione S Nieman LZ Pulmonary auscultatory skills during training in internal medicine and family practice Am J Respir Crit Care Med 1999 159 4 Pt 1 1119 1124 10.1164/ajrccm.159.4.9806083 Available from: http://dx.doi.org/10.1164/ajrccm.159.4.9806083 10194155 12 Dull P Haines DJ Methods for teaching physical examination skills to medical students Fam Med 2003 35 5 343 348 12772936 13 Dent JA Harden RM A Practical Guide for Medical Teachers 2005 Edinburgh Elsevier Churchill Livingstone 14 Pierre MS Breuer G Simulation in der Medizin: Grundlegende Konzepte - Klinische Anwendung 2013 Berlin, Heidelberg Springer 15 Ziv A Ben-David S Ziv M Simulation based medical education: an opportunity to learn from errors Med Teach 2005 27 3 193 199 10.1080/01421590500126718 Available from: http://dx.doi.org/10.1080/01421590500126718 16011941 16 Lynagh M Burton R Sanson-Fisher R A systematic review of medical skills laboratory training: where to from here? Med Educ 2007 41 9 879 887 10.1111/j.1365-2923.2007.02821.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2007.02821.x 17696985 17 Zendejas B Brydges R Wang AT Cook DA Patient outcomes in simulation-based medical education: a systematic review J Gen Intern Med 2013 28 8 1078 1089 10.1007/s11606-012-2264-5 Available from: http://dx.doi.org/10.1007/s11606-012-2264-5 23595919 18 Biggs J Tang C Teaching for Quality Learning at University: What the students does 2007 Maidenhead : Open University Press/ McGraw Hill 19 Buss B Krautter M Moltner A Weyrich P Werner A Jünger J Nikendei C Can the 'assessment drives learning' effect be detected in clinical skills training? - Implications for curriculum design and resource planning GMS Z Med Ausbild 2012 29 5 Doc70 10.3205/zma000840 Available from: http://dx.doi.org/10.3205/zma000840 23255965 20 Nikendei C Jünger J OSCE – praktische Tipps zur Implementierung einer klinisch-praktischen Prüfung GMS Z Med Ausbild 2006 23 3 Doc47 Available from: http://www.egms.de/static/de/journals/zma/2006-23/zma000266.shtml 21 Wass V Van der Vleuten C Shatzer J Jones R Assessment of clinical competence Lancet 2001 357 9260 945 949 10.1016/S0140-6736(00)04221-5 Available from: http://dx.doi.org/10.1016/S0140-6736(00)04221-5 11289364 22 Pjontek R Scheibe F Tabatabai J Heidelberger Standarduntersuchung: Handlungsanweisung zur Durchführung der körperlichen Untersuchung 2013 Heidelberg HeiCuMed 23 Nikendei C Kadmon M Heidelberger Standardprozeduren 2015 Heidelberg Medizinische Fakultät Heidelberg

PMC005xxxxxx/PMC5003137.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00105110.3205/zma001051Doc52urn:nbn:de:0183-zma0010514Article"Peer2Peer" – A university program for knowledge transfer and consultation in dealing with psychosocial crises in med-school and medical career „Peer2Peer“ – Ein universitäres Programm zur Wissensvermittlung und Beratung im Umgang mit psychosozialen Krisen im Medizinstudium wie im ärztlichen Beruf Vajda Christian *11 Medical University of Graz, Department of Medical Psychology and Psychotherapy, Graz, Austria*To whom correspondence should be addressed: Christian Vajda, Medical University of Graz, Department of Medical Psychology and Psychotherapy, Auenbruggerplatz 2/8, A-8010 Graz, Austria, E-mail: christian.vajda@medunigraz.at15 8 2016 2016 33 4 Clinical skillsDoc5216 10 2015 09 5 2016 15 2 2016 Copyright © 2016 Vajda2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001051.shtmlObjective: Medical students are exposed to various psychosocial problems and challenges. Specific consultations services and programs can support them. “Peer2Peer” is such a consultation program and was implemented at the Medical University of Graz. It focusses on crisis intervention, psychosocial stress management, junior mentoring as well as student education in this field. Besides, it also offers student tutors of the program practical skills trainings. The program was restructured in winter term 2014/15. Methods: On the one hand, “Peer2Peer” gives insights into topics such as the current state of research concerning the students’ psychological strain and psychosocial crises in acutely stressful situations and preventive approaches for coping with these kinds of situations on the other hand. These aspects are taught by means of elective courses, lectures and workshops. Furthermore, “Peer2Peer” provides consultation services by student tutors who give face-to-face advice if required. These tutors receive ongoing training in organizational and professional issues. Results: Since the summer term of 2015, 119 students have been trained (via lectures and elective courses), while 61 contacts (short consultation) and 33 contacts (full consultation) have been supervisied. In total, two psychotherapeutic and one psychosocial follow ups were recommended. There are seven students who participate as tutors in the program. Conclusions: The “Peer2Peer” program is intended to enable a low-threshold access for medical students facing psychosocial crises situations and to help them in dealing with stress and learning problems. An increase in support contacts from the summer term of 2015 to the winter term of 2015/16 can be considered a success. A first evaluation of the different components of the program started in the winter semester of 2015/16. The student tutors have not only acquired practical skills in dealing with students in crises situations but also various organizational skills. Zusammenfassung Zielsetzung: Medizinstudierende sind vielfältigen psychosozialen Belastungen und Herausforderungen ausgesetzt. Spezifische Beratungs- und Betreuungsprogramme können eine Unterstützungsfunktion innehaben. An der Medizinischen Universität Graz nimmt das „Peer2Peer“-Programm diese Aufgabe im Zuge eines Angebotes zur Krisenintervention, psychosozialen Stressbewältigung, Junior-Mentoring, Fort- und Weiterbildung für Studierende als auch praktischer Kompetenzsteigerung der betreuenden Tutor/-innen wahr. Mit Wintersemester 2014/15 erfolgte eine Neustrukturierung. Methodik: Die Vermittlung von Themen wie der Evidenz zur Belastung von Studierenden, Umgang mit psychosozialen Krisen im akuten Fall sowie präventive Ansätze erfolgt über ein Wahlfach, einen Beratungsdienst, besetzt durch studentische Tutor/-innen, sowie spezifische Fort- und Weiterbildungen. Die Tutor/-innen werden laufend in organisatorischen und fachlichen Belangen geschult. Ergebnisse: Seit dem Sommersemester 2015 wurden 119 Studierende in themenspezifischen Veranstaltungen (Vorträge und Wahlfach) weitergebildet und 61 kurze sowie 33 ausführliche Beratungskontakte durchgeführt. Zwei psychotherapeutische und eine psychosoziale Weitervermittlungen wurden veranlasst. Sieben Studierende nehmen als Tutor/-innen am Programm teil. Schlussfolgerungen: Das Peer2Peer-Programm soll einen niederschwelligen Zugang für Medizinstudierende in psychosozialen Krisensituationen sowie bei Fragen zur Stressbewältigung und Lernproblemen ermöglichen. Eine Steigerung der Betreuungskontakte von Sommersemester 2015 auf Wintersemester 2015/16 kann als Erfolg gewertet werden. Eine erste Teilevaluierung der unterschiedlichen Programmbestandteile ist seit dem Wintersemester 2015/16 in Durchführung. Für die studentischen Tutor/-innen ist ein Wissenszuwachs bezüglich des praktischen Umganges mit Personen in Krisensituationen sowie in diversen organisatorischen Kompetenzen zu vermerken. psychosocial support systemsmedical students ==== Body Introduction Medical students are facing many challenges during medical school. Among others, these include dealing with topics like life and death, illness and suffering, the perception one’s physical and emotional limitations, problems in private life and financial matters. Consequences could result in a higher burden of anxiety and depression [1], [2] and a lower quality of life in comparison to students of other disciplines and the general population [3], respectively. Substantive and structural conditions of the curricula can favor the occurrence of these burdens [4], [5] and even intensify them over time [6]. Experiences made at medical school generate an important impact for the students’ future work-life balance. Thus, positive and negative stress management is highly important [7], [8], [9], [10], [11]. The effects of inadequate regulation skills can result in health risks for patients, such as a higher rate of prescription errors of depressed and burnout physicians [12]. An international comparison indicates a large diversity of support services for students [13], [14].These include lectures (including behavior change strategies and evidence-based data on psychosocial strain) and tutorials (exercises for mindfulness and group discussions) in the core curriculum like the Health Enhancement Program (HEP) at Monash University in Australia [15] or consultation programs at the Eastern Virginia Medical School. In this school, medical students can discuss experiences of psychosocial stress and receive medical care from a medical consultant of the department of “Family and Community Medicine” at the beginning and at the end of the academic year [16]. Some universities even involve their students in the development and design of such programs [17]. By implementing the “Peer2Peer” program, the Medical University of Graz strives to increase the students’ awareness for matters of psychosocial stress during their studies and their future professional life. Furthermore, the program supports students in coping with challenges in their academic and private life and aims at increasing knowledge in dealing with patients who are experiencing extreme psychosocial situations. Description of the program The “Peer2Peer” program focuses on topics such as crisis intervention, psychosocial stress management, knowledge transfer and junior mentoring free of charge for all students (n>4000). The program was restructured in the winter term of 2014/15. Consultation services and public relation activities started in the summer term of 2015. The four aforementioned main topics are realized by three organizational components of the program: An elective course open to all students Support in crises and psychosocial stress situations as well as junior mentoring by student tutors (peers) through consultation services Practical and theoretical training (lectures and workshops) Elective courses and training of the tutors An elective course (2 terms) was implemented for training the tutors and transferring knowledge to the students in general. Students of all courses of study at the Medical University of Graz are entitled to participate (“Psychosocial crisis intervention and stress management I and II”, each 1 ECTS). The interdisciplinary staff (n=4, psychiatry, psychology, psychotherapy) is highly experiences in teaching, research and patient care. The five course units for each term (5x135 minutes) include the following topics: Winter term Introduction to psychosocial stress and resources for students, current state of research, general information about the “Peer2Peer” program New in 2015/16: relaxation techniques (overview and practical exercises) Basics in psychotherapeutic topics Basics in psychiatric topics (i.e.: depression, anxiety) Detailed information about “Peer2Peer”, case reports of “Students in psychosocial crises”, reflection on the term (feedback in group discussions/written work) and preparation for the following term Summer term Challenges in the supervision of students (case studies and current evidence) Relaxation techniques/coping mechanisms/resources strengthening Case studies from psychotherapeutic practice, role plays Specific psychiatric topics (handling suicidality) Summary of the acquired knowledge, Q&A on the “Peer2Peer” program, information about support networks, reflection Pre-course requirement for the participation in this elective is the completion of the fourth academic term. The performance assessment is carried out by a. a compulsory attendance of 80%, b. active participation and c. a written reflection at the end of the course. Structure of the consultation service After successfully completing both parts of the elective, students are allowed to work as tutors (1 to 3 hours per week; October to January and March to June) as part of the consultation service of the university. The tutors offer at least two consultation time slots per week (Monday and Wednesday, 3pm to 4:30pm) and they can be contacted by the students via email, Facebook and telephone. Consultations can also be arranged in addition to the two fixed time slots. A room is constantly blocked for the tutors. Tutors mostly deal with the following topics: developing a schedule for revising study contents, psychosocial crises in the private life, introduction to relaxation techniques, and / or coping with exam stress. An advisory board for more complex issues (i.e., psychotherapeutic follow-up care) was set up at the Department of Medical Psychology and Psychotherapy. Students can use the consultation service free of charge and anonymously. The quality management (including: tutor training, team meetings, appraisal interview with tutors, standardized documentation, supervisions) is shown in figure 1 (Fig. 1). Training In addition to consultation in psychosocial, study-related crises cases and junior mentoring, a focus is put on prevention and health promotion since winter term 2014/15. For this purpose, the tutors and the program coordinator organize lectures and workshops, information sessions and analyze structure-related problems (i.e., lack of recreation options) of the study program. Results Report of performance In summer term 2015, five students worked in the program. In winter term 2015/16, seven students participated as tutors in the consultation service. Two new tutors who had previously completed the elective helped for purposes of quality management only in organizational activities (see figure 1 (Fig. 1)). The consultation (see table 1 (Tab. 1)) and knowledge transfer (see table 2 (Tab. 2)) increased from the summer term (starting after restructuring in winter 2014/15) to the winter term of 2015/2016. Exam stress, distress because of board exam, fears of failure, family conflicts and difficulties in learning are the main problems addressed in the consultation service. Lectures and workshop, focused on the topics "relaxation" and "learning strategies", were open to all students of the university. Elective After restructuring the program in the winter term of 2014/15, the number of participants remained at a comparable level in winter term 2015/16 (see table 2 (Tab. 2)). Due the lack of written reflections (third performance criteria), the final result for the winter term of 2015/16 is not available yet. Thus, no final summary can be drawn at the time of submitting this article. However, 15 students meet the necessary requirements (compulsory attendance and active participation). The drop in the numbers of students from winter to summer term and within the two terms can have various reasons. An essential aspect could be the often simultaneously held obligatory core curriculum courses. That makes it impossible for several students to achieve the necessary attendance of at least 80%. The fact that the first part is obligatory for the second part of the elective usually leads to a smaller number of participants compared to the start of the winter term. Practical skills and knowledge transfer for tutors In addition to the acquisition of theoretical knowledge, insights into backgrounds and practical strategies for dealing with psychosocial crises (assessment, different approaches of psychotherapy and psychiatry, helping networks, correct referrals for future work, dealing with suicidal tendencies, accompaniment of students, professional interviewing) there is a focus in organizational skills (team coordination, rostering, meeting planning and execution etc.), event management (training for students) and public relations activities (representation, social media, etc.). Apart from that, tutors participate in at least two professional trainings (for example: relaxation techniques, behavioral approaches in the care of students, case discussions) during the academic term. As part of the quality management, they attend appraisal interviews as well as individual and group supervisions. Discussion Implementing the “Peer2Peer” program is an attempt to create a low-threshold consultation service focused on the needs of students. Due to the fact that there was no program at the Medical University of Graz before, a greater awareness for the topic “distress in academic studies and working life” was achieved on a local university level. A complete restructuring of the program, including new content and organizational aspects (elective, consulting service, training), was carried out in the winter term of 2014/15. As a second step, the consultation service for students and reinforced public relations activities were launched in summer term 2015. The increase in the number of performances of consultation services from summer term 2015 to winter term 2015/16 can be considered as a first success of the efforts to raise the students’ awareness for the program. In a third step, the evaluation of the program components was launched in the winter term of 2015/16. At the time of editing this article, the evaluation of the first part of the elective had not been completed yet. A publication is planned for the GMA conference 2016th. An evaluation of the tutors’ activities, an acceptance analysis of the program and a survey focusing on the needs of students is planned for 2016. Based on past experiences, intense public relations activities (among others: university news feed, posters and flyers, students manuals) are necessary in order to reinforce the awareness of students for the program and to implement it as a permanent consultation service. Further topics (practical year, board exams, internship, etc.) and more consultation time slots are planned for the academic year of 2015/16. Contact details of the program Website: http://www.medunigraz.at/peer2peer Facebook: http://www.facebook.com/peer2peer E-Mail: peer2peer@medunigraz.at Acknowledgements Hereby the author wants to express his heartfelt thanks to the dedicated teachers and student tutors of the program and to the institutional services of the university. He would also like to thank the stuff and departments of the study organization for providing the necessary financial support (employment of the tutors). Without them the implementation of the program would not be possible. Competing interests The author declare that he has no competing interests. Table 1 Table1: Overview of the number of consultation contacts of the „Peer2Peer“ program (February 2016) Table 2 Overview of the number of participating students in educational activities of the program (without tutor trainings) Figure 1 Overview of the three pillars of quality management of the “Peer2Peer” program at the Medical University of Graz ==== Refs 1 Aketin M Karaman T Senol YY Erdem S Erengin H Akaydin M Anxiety, depression and stressful life events among medical students: A prospective study in Antalya, Turkey Med Educ 2001 35 12–17 11123589 2 Miller NM McGowen RK The painful truth: physicians are not invincible South Med J 2000 93 10 966 973 10.1097/00007611-200093100-00004 Available from: http://dx.doi.org/10.1097/00007611-200093100-00004 11147478 3 Henning MA Krägeloh CU Hawken SJ Zhao Y Doherty I The quality of life of medical students studying in new Zealand: a comparison with nonmedical students and a general population reference group Teach Learn Med 2012 24 4 334–340 10.1080/10401334.2012.715261 Available from: http://dx.doi.org/10.1080/10401334.2012.715261 23036001 4 Bernhardt V Rothkötter Kasten E Psychische Belastung durch die Dissektion am Leichnam im anatomischen Präparierkurs bei Erstsemestern des Studienfaches Medizin GMS Z Med Ausbild 2012 29 1 Doc12 10.3205/zma000782? Available from: http://dx.doi.org/10.3205/zma000782? 22403597 5 Reed DA Shanafelt TD Satele DW Power DV Eacker A Harper W Mountier C Durning S Massie FS Jr Thomas MR Sloan JA Dyrbye LN Relationship of pass/fail grading and curriculum structure with well-being among preclinical medical students: a multi-institutional study Acad Med 2011 86 11 1367 1373 10.1097/ACM.0b013e3182305d81 Available from: http://dx.doi.org/10.1097/ACM.0b013e3182305d81 21952063 6 Voltmer E Rosta J Aasland OG Spahn C Study-related health and behavior patterns of medical students: A longitudinal study Med Teach 2010 32 10 422 428 10.3109/0142159X.2010.496008 Available from: http://dx.doi.org/10.3109/0142159X.2010.496008 20423262 7 Frank E Rothenberg R Lewis C Belodoff BF Correlates of physicians´prevention-related practices. Findings from the Woman Physicians´ Health Study Arch Fam Med 2000 9 4 359 367 10.1001/archfami.9.4.359 Available from: http://dx.doi.org/10.1001/archfami.9.4.359 10776365 8 Afzal H Afzal S Siddique SA Naqvi SA Measures used by medical students to reduce test anxiety J Pak Med Assoc 2012 62 9 982 986 23139995 9 Newbury-Birch D White M Kamali F Factors influencing alcohol and illicit drug use amongst medical students Drug Alcohol Depend 2000 59 2 125 130 10.1016/S0376-8716(99)00108-8 Available from: http://dx.doi.org/10.1016/S0376-8716(99)00108-8 10891625 10 Webb E Ashton CH Kelly P Kamah F An update on British medical students ´lifestyles Med Educ 1998 32 3 325 331 10.1046/j.1365-2923.1998.00204.x Available from: http://dx.doi.org/10.1046/j.1365-2923.1998.00204.x 9743790 11 Sreeramareddy C Shankar PR Binu VS Mukhopadhyay C Ray B Menezes RG Psychological morbidity, sources of stress and coping strategies among undergraduate medical students of Nepal BMC Med Educ 2007 7 26 10.1186/1472-6920-7-26 Available from: http://dx.doi.org/10.1186/1472-6920-7-26 17678553 12 Fahrenkopf AM Sectish TC Barger LK Sharek PJ Lewin D Chiang VW Rates of medication errors among depressed and burnt out residents: Prospective cohort study Br Med J 2008 366 388 391 10.1136/bmj.39469.763218.be Available from: http://dx.doi.org/10.1136/bmj.39469.763218.be 13 Estabrook K Medical Student Health Promotion: The Increasing Role of Medical Schools Acad Psych 2008 32 65 68 10.1176/appi.ap.32.1.65 Available from: http://dx.doi.org/10.1176/appi.ap.32.1.65 14 Lee J Graham AV Students perspective of medical school stress and their evaluation of a wellness elective Med Educ 2001 35 652 659 10.1046/j.1365-2923.2001.00956.x Available from: http://dx.doi.org/10.1046/j.1365-2923.2001.00956.x 11437967 15 Hassed C de Lisle S Sullivan G Pier C Enhancing the health of medical students: outcomes of an integrated mindfulness and lifestyle program Adv Health Sci Educ Theory Pract 2009 14 3 387 398 10.1007/s10459-008-9125-3 Available from: http://dx.doi.org/10.1007/s10459-008-9125-3 18516694 16 Davies SM Rutledge CM Davies TC A family physician-counselor program for medical students Fam Med 1993 25 5 327 330 8514003 17 Drolet CD Rodgers S A Comprehensive Medical Student Wellness Program_Design and Implementation at Vanderbilt School of Medicine Acad Med 2010 85 1 103 110 10.1097/ACM.0b013e3181c46963 Available from: http://dx.doi.org/10.1097/ACM.0b013e3181c46963 20042835

PMC005xxxxxx/PMC5003138.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00105510.3205/zma001055Doc56urn:nbn:de:0183-zma0010558ArticleMastery learning improves students skills in inserting intravenous access: a pre-post-study Mastery Learning verbessert die Fertigkeiten von Studierenden im Legen von peripheren Venenverweilkanülen: eine Prä-Post-Studie Friederichs Hendrik 1Brouwer Britta 1Marschall Bernhard 2Weissenstein Anne *11 University of Muenster, Studienhospital, Muenster, Germany2 University of Muenster, Institute of Medical Education – IfAS, Muenster, Germany*To whom correspondence should be addressed: Anne Weissenstein, University of Muenster, Studienhospital, Malmedyweg 17-19, D-48149 Muenster, Germany, E-mail: anne.weissenstein@gmail.com15 8 2016 2016 33 4 Clinical skillsDoc5619 8 2015 04 7 2016 01 6 2016 Copyright © 2016 Friederichs et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001055.shtmlObjective: Inserting peripheral venous catheters (PVCs) has been identified as a core competency for medical students. Because the performance – even of hygienic standards – of both students and novice physicians is frequently inadequate, medical faculties must focus on competence-based learning objectives and deliberate practice, features that are combined in mastery learning. Our aim was to determine the competency of students in inserting PVCs before and after an educational intervention. Design: This study comprised a skills assessment with pre- and post-tests of a group of third-year students who received a simulation-based intervention. A newly established curriculum involved one hour of practice at inserting PVCs on simulators. Students were required to pass a test (total 21 points, pass mark 20 points) developed on the concept of mastery learning. An unannounced follow-up test was performed one week (8 days) after the intervention. Setting: The simulation center of the medical faculty in Muenster. Participants: Third-year students who received the intervention. Results: One hundred and nine complete data sets were obtained from 133 students (82.5%). Most students (97.2%) passed the test after the intervention (mean score increase from 15.56 to 20.50, P<0.001). There was a significant decrease in students’ performance after one week (8 days): only 74.5% of participants passed this retest (mean score reduction from 20.50 to 20.06, P<0.001). Conclusion: Mastery learning is an effective form of teaching practical skills to medical students, allowing a thorough preparation for the challenges of daily clinical practice. Zusammenfassung Hintergrund: Das Legen von peripheren Venenverweilkanülen ist als Kernkompetenz von Medizinstudierenden bestimmt worden. Da die diesbezügliche Performance – auch im Hinblick auf hygienische Standards – häufig inadäquat ist, müssen sich medizinische Fakultäten mehr auf kompetenz-basierte Lernziele und “deliberate practice” fokussieren. Diese Aspekte sin im Mastery Learning kombiniert. Unser Ziel war es, die Kompetenz im Legen von peripheren Venenverweilkanülen nach einer Lehrintervention zu messen. Studiendesign: Diese Studie umfasst eine Vorher-Nachher- Kompetenzmessung von Studierenden im 3. Studienjahr, die eine simulations-basierte Lehrintervention erhielten. Die Studenten mussten einem Test bestehen (insgesamt 21 Punkte, Mindestpunktzahl 20 Punkte), welcher basierend auf dem Konzept der Mastery Learning entwickelt wurde. Ein unangekündigter Follow-up- Test wurde eine Woche (8 Tage) nach dem Eingriff durchgeführt. Rahmen: Das Studienhospital der Medizinischen Fakultät Münster. Teilnehmer: Teilnehmer waren Studierende im 3. Studienjahr. Ergebnisse: Es konnten 109 vollständige Datensätze von 133 Studierenden (82,5%) in die Analyse einbezogen werden. Die meisten Studierenden (97,2 %) haben den Test nach der Lehrintervention bestanden (durchschnittliche Erhöhung der Gesamtpunktzahl von 15,56 auf 20,50, P<0,001). Es gab eine signifikante Abnahme der Performance der Studierenden nach einer Woche (8 Tage): nur 74,5% der Teilnehmer bestanden diesen erneuten Test (mittlere Reduktion der Gesamtpunktzahl von 20,50 auf 20,06, P<0,001). Schlussfolgerung: Mastery Learning ist ein effektives Mittel um Medizinstudierenden praktische Fertigkeiten beizubringen und ermöglicht damit eine umfassende Vorbereitung auf die Herausforderungen des klinischen Alltags. mastery learningundergraduate medical educationperipheral venous cathetermedical students ==== Body Introduction At present, the requirements of physicians are that they are scientifically and practically trained, able to practice medicine, and undertake training and continuing education independently and responsibly (http://www.gesetze-im-internet.de/_appro_2002/BJNR240500002.html cited 4. Dezember 2012). Accordingly, the goal of undergraduate medical education is to enable future physicians to act in a competent and responsible manner. Therefore, medical education should be performed on a scientific basis and be related to medical practice as well as to patients. Despite these requirements, considerable deficits in the competence and performance of students as well as novice physicians are often noted [1], [2]. In addition, a repeated criticism is that education by medical schools is too theoretical, emphasis being primarily on the transmission of knowledge-based content [3]. In response to this criticism, a shift from purely cognitive to more competence-based learning objectives in medical education is occurring internationally [4]. A methodology in providing competence-based education in medicine is mastery learning. First described by Bloom in 1968, this concept means that learners acquire knowledge and skills, which are measured against fixed achieved standards, the time taken to reach the required outcomes being irrelevant [5], [6]. The aim of mastery learning is to ensure that all learners reach teaching objectives with little or no outcome variance [7]. It describes a process of mastering specific learning objectives and requires clearly defined teaching aims organized into small, successively organized units; students are not only required to understand the material before moving to the next lesson [8], but to achieve a defined proficiency before proceding [9]. Various studies in postgraduate medical education have examined the acquisition of skills involved in performing invasive procedures and proven the effectiveness of this method. Among others, there is such evidence for learning how to perform extended resuscitation [10], central venous catheter insertion [11], lumbar puncture in infants [12] and laparoscopic techniques [13]. So far, only one study has described implementation of mastery learning in undergraduate medical education [14]. To our knowledge, the mastery learning method has never been used to teach an invasive procedure to undergraduate medical students. We have therefore conducted a study to integrate and evaluate this type of educational intervention in the medical curriculum. In this study, we implemented the concept of mastery learning in the course “Medical Skills Lab”, which took place in the simulation center of the medical faculty (Studienhospital Muenster®). The framework was insertion of peripheral venous catheters (PVCs), a core clinical skill that students need to acquire. Given that 5-7% of all PVCs have reportedly become colonized by bacteria by the time catheter material [15], [16] and PVC-associated septicemia occurs in up to 2% of all patients with PVCs [16], [17], it is not surprising that insertion and care by specially trained personnel is associated with a significant reduction in occurrence of phlebitis and infection [18], [19]. Thus, in regard to patient safety, it is desirable that medical students are able to perform this task as well as possible. The aims of our study were as follows: Primary Outcomes To assess whether mastery learning increases third-year (fifth semester) students’ skills in inserting PVCs Secondary Outcomes To evaluate third-year students’ skills in inserting PVCs one week (8 days) after the intervention To evaluate the feasibility of integrating mastery learning permanently into the medical curriculum Methods Study Design A skills assessment with pre- and post-tests of a group of third-year students was performed using a simulation-based intervention. Participants’ skills in placing PVCs were assessed before and immediately after the course (on the same day) in the form of pre- and post-tests, which were part of the intervention. Follow-up tests were performed eight days after the intervention. Further, relevant data (age, number of clinical internships, and number of previously placed PVCs) were collected from participants. Participants From December 2012 to January 2013, the proficiency of 133 third year students at inserting PVCs was assessed before and after the intervention. The students had not received any formal training. In Germany, medical students are required to complete six years of study (12 semesters); thus, the students had already completed their preclinical stage (first two years) and entered the clinical phase of their studies. All participants gave their informed consent before the pre-test. The Ethics Committee of the Chamber of Physicians, Westfalen-Lippe, and the Medical Faculty of the Westfalian Wilhelms University, Muenster, waived requirements for an ethical approval procedure. Outcome Measures and Measuring Instrument The primary outcome measure was comparison of the students’ test scores. All three tests (pre-test, post-test, follow-up test) had the same structure and were part of the intervention. They further served as an assessment tool in the form of a checklist, thus providing a guide for the correct insertion of PVCs. The checklist was divided into four units: preparing materials, preparation of the puncture, execution of the puncture, and finally fixation of the venous catheter. Each unit was further subcategorized into smaller, successively organized subunits (seven, four, five and five subunits, respectively). For example, subunits included retrieving gloves, disinfection of the puncture site and announcing the puncture to the patient. The structure and sequence of all units were developed in accordance with the objective structured clinical examination checklist of Juenger and Nikendei [20]. Each successfully performed sub-unit scored one point. Thus, 21 points could be scored; students passed the test once they had scored 20 points. For reasons of patient safety the pass mark in this study was defined as 20 points; thus, it was not possible to pass without correctly disinfecting the skin. After the course, the absolute passing score for the performance examination was defined by a multidisciplinary panel of eight clinical experts, who determined the minimum passing score by the Hofstee (group-based) standard setting method [21]. The panel consisted of faculty members who were certified in family medicine (n=2), surgery (n=1), orthopedics (n=1), internal medicine (n=1) and three faculty members involved in medical education. Each panelist received instructions in a standard setting. As a result, a Hofstee score of 20 points was used as the final minimum passing score. Educational Intervention The newly developed course consisted of one-hour practice at inserting PVCs on simulators (IV-arm, Carl Gustav Carus Management GmbH, Dresden, Germany). Students were asked to imagine caring for a real patient and to provide the imaginary patient while commenting each step of their actions. A pre-test was performed for baseline assessment. After the pre-test, the students practiced inserting PVCs in groups of two on simulators for a total of 1 hour (training period of 30 minutes per student) and the deficits noted in the pre-test were discussed. The training sessions were supported by peer-teachers (trained student tutors from higher semesters). Each tutor had received thorough instructions for about one hour and schooling on the checklist prior to the study. A post-test was performed at the end of the educational intervention. Eight days after the course, an unannounced voluntary follow-up test was offered, allowing students to assess their skills again. Data Processing and Analysis Only students with complete data sets (i.e., who had participated in both pre- and post-tests) were included in the analysis. Participants who passed the pre-test on their first attempt were excluded from the study because the desired outcome measure was an increase in students’ skills after the intervention, and participants who had passed the pre-test could not further improve their performance. Our primary analysis was a comparison of the mean total test scores (pre-test, post-test) and comparison of follow-up test scores. Data were analyzed with IBM Statistics SPSS 19. Descriptive means and standard deviations were calculated and other parametric tests used. Student’s paired t-test was performed to compare the results of pre- and post-tests and of post- and follow-up tests. Pearson's χ2 test was used to compare pass rates. Further, one-way ANOVA was performed in order to evaluate differences between the groups (significance level P≤0.05). We used the standard alpha-level of 0.05 for significance and a power level of 0.8. Thus, we needed a sample size of at least 27 participants to detect a medium effect size (d=0.5) (G*Power 3.1). Results Participants In all, 133 third-year students performed the pre-test. After exclusion of 16 students who passed the pre-test on their first attempt and a further 8 whose records were incomplete, the final sample comprised 109 third-year students. Baseline characteristics of partcipants are described in Table 1 (Tab. 1). Of the 109 third-year students enrolled in the study, 106 completed the follow-up test. Primary Outcomes Following the educational intervention, 106/109 students (97.2%) passed the test with a mean score (standard deviation) of 20.50 (0.56), whereas the mean score at baseline was 15.56 (2.44). This improvement in the students’ performances is highly significant (P<0.001; effect size Cohens' d 2.79). The remaining three students scored 19 points and had the opportunity to repeat the course until they passed the test after the study. Secondary Outcomes About one week (8 days) after the educational intervention, 106 students performed a follow-up test, in which they scored significantly less (20.06 [0.94] points) than in the previous test (post-test) (20.50 [0.56] points) (P < 0.001). The effect size (Cohen's d) in comparison to the mean score at baseline is 2.43. However, with the pass mark being 20 of a possible 21, three-quarters of the students (74.5%) still passed the test (20.8% of students scored 19 points, 2.8% scored 18 and 1.9% scored 17). The performance of students according to sub-units of the task is depicted in Figure 1 (Fig. 1). Discussion Our results show that our educational intervention significantly increased students’ practical skills in inserting PVCs. After the course, the majority of students (97.2%) passed the test. However, there was a significant decrease in students’ performances after one week (8 days), only 74.5% of participants passing the test after this interval (mean score reduction from 20.50 to 20.06 points). Inserting PVCs is a core clinical skill students need to possess and physicians expect them to perform this task during their clinical internships. Most patients are willing to let medical students perform even invasive procedures after simulator-training [22]. thus medical schools have the responsibility of preparing students in the best possible way. In accordance with the concept of mastery learning, our intervention focused on clearly defined goals, deliberate practice and precise outcome measures. Another advantage of this method is that, because the outcome measures are standardized, students’ learning is no longer dependent on the teaching skills of tutors. Further, self-reflection is a key aspect of mastery learning, thus value was set on the active articulation of performed steps. Mastery learning allows students to learn and practice invasive diagnostic and therapeutic procedures with unprecedented frequency. This concept has become an important component of continuing education because it has been shown to improve the quality of patient care [11], [13], [23], [24]. In our opinion the key element of mastery learning of clinical skills is that learners receive individual, immediate feedback according to transparent standards, allowing them to focus on their own learning needs. It seems logical that this concept be applied to medical education for the benefit of students and, of course, the patients treated by them. The results of our study should be compared with the results from a review of mastery simulation-based medical education (SBME) of Cook who states that “In comparison with no intervention, mastery SBME what associated with large effects on skills (41 studies; effect size [ES] 1:29 [95% confidence interval, 1:08 to 1:50])” [9]. Our effect size showed with a value of 2.79 in the post-test and 2.43, eight days later even higher results. Due to the quality of data and analysis the evidence of a meta-analysis is of course significantly higher than the results of our study. But in a subgroup analysis of the described review only 6 trials with a total of 109 medical students were eligible. For them, the effect size was 1.2 (95% CI 0.44 to 1.95) [Attachment Digital Table 2 in this publication ]. We could already include 106 participants in our study. Our study has several limitations. First, the pre- and post-tests were performed on the same day and, because of the course structure, it was not possible to blind the tutors according to pre-, post- and follow-up tests. Second, three third-year students did not pass the post-test; however, the concept of mastery learning dictates that all learners reach the learning objectives. Therefore, after the study these students had the opportunity to repeat the course until they succeeded. How D.A. Cook in his work “If you teach them, they will learn” [25] discussed, studies describing the impact of a single specific training have only a small impact for educational research and the actual teaching. Accordingly, in future a controlled study, ideally even a RCT should be conducted to determine the effectiveness of the teaching intervention in more detail. Despite these limitations, our study has shown that mastery learning is an effective form of teaching practical skills during the education of medical students. We hope to further demonstrate that this teaching format could be introduced to continuing graduate education as well as undergraduate medical education. In light of the promising results of our study, both regarding the efficiency of the teaching units and the favorable feedback from students, we have integrated the concept of mastery learning into the medical curriculum on a long-term basis. Mastery learning is a valuable model that helps to prepare medical students for the challenges of daily clinical practice and thus facilitate their entry into professional life. Authors’ contributions HF conceived of the study and participated in data acquisition, data analysis and the drafting of the manuscript. BB participated in data acquisition and data analysis. BM participated in coordination. AW participated in data analysis and helped to draft the manuscript. All authors have read and approved the final manuscript. Author information HF is the head of the study hospital of the medical school in Muenster and has a master’s degree in Medical Education (MME). BM is the Dean of the medical faculty and head of the Institute of Medical Education and Student Affairs. AW and BB have medical degrees and work as research assistants in the study hospital. Competing interests The authors declare that they have no competing interests. Supplementary Material Digital Table 2 Table 1 Participants’ characteristics at baseline Figure 1 Performance of students according to the sub-units of the task in the pre- and post-tests (n=109) and one week (8 days) after the educational intervention (follow-up test, n=106). *Significant differences between the pre- and post-test results for all four subunits (P<0.05). †Significant difference between post- and follow-up test for “fixation of the PVC” only (P<0.05). ==== Refs 1 Mangione S Nieman LZ Cardiac auscultatory skills of internal medicine and family practice trainees. A comparison of diagnostic proficiency JAMA 1997 278 9 717 722 10.1001/jama.1997.03550090041030 Available from: http://dx.doi.org/10.1001/jama.1997.03550090041030 9286830 2 March SK Bedynek JL Jr Chizner MA Teaching cardiac auscultation: effectiveness of a patient-centered teaching conference on improving cardiac auscultatory skills Mayo Clin Proc 2005 80 11 1443 1448 10.4065/80.11.1443 Available from: http://dx.doi.org/10.4065/80.11.1443 16295024 3 Cooke M Irby DM Sullivan W Ludmerer KM American medical education 100 years after the Flexner report N Engl J Med 2006 355 13 1339 1344 10.1056/NEJMra055445 Available from: http://dx.doi.org/10.1056/NEJMra055445 17005951 4 Bradley P The history of simulation in medical education and possible future directions Med Educ 2006 40 3 254 262 10.1111/j.1365-2929.2006.02394.x Available from: http://dx.doi.org/10.1111/j.1365-2929.2006.02394.x 16483328 5 Bloom BS Learning for mastery Eval Comm 1968 1 2 1 12 6 Wayne DB Butter J Siddall VJ Fudala MJ Wade LD Feinglass J McGaghie WC Mastery Learning of Advanced Cardiac Life Support Skills by Internal Medicine Residents Using Simulation Technology and Deliberate Practice J Gen Intern Med 2006 21 3 251 256 10.1111/j.1525-1497.2006.00341.x Available from: http://dx.doi.org/10.1111/j.1525-1497.2006.00341.x 16637824 7 McGaghie WC Siddall VJ Mazmanian PE Myers J Lessons for continuing medical education from simulation research in undergraduate and graduate medical education: effectiveness of continuing medical education: American College of Chest Physicians Evidence-Based Educational Guidelines Chest 2009 135 3 Suppl 62S 68S 10.1378/chest.08-2521 Available from: http://dx.doi.org/10.1378/chest.08-2521 19265078 8 Anderson JR Learning and memory: An integrated approach 2000 2nd ed New York John Wiley and Sons, Inc 9 Cook DA Brydges R Zendejas B Hamstra SJ Hatala R Mastery learning for health professionals using technology-enhanced simulation: a systematic review and meta-analysis Acad Med 2013 88 8 1178 1186 10.1097/ACM.0b013e31829a365d Available from: http://dx.doi.org/10.1097/ACM.0b013e31829a365d 23807104 10 Wayne DB Butter J Siddall VJ Fudala MJ Wade LD Feinglass J McGaghie WC Mastery learning of advanced cardiac life support skills by internal medicine residents using simulation technology and deliberate practice J Gen Intern Med 2006 21 3 251 256 10.1111/j.1525-1497.2006.00341.x Available from: http://dx.doi.org/10.1111/j.1525-1497.2006.00341.x 16637824 11 Barsuk JH McGaghie WC Cohen ER Balachandran JS Wayne DB Use of simulation-based mastery learning to improve the quality of central venous catheter placement in a medical intensive care unit J Hosp Med 2009 4 7 397 403 10.1002/jhm.468 Available from: http://dx.doi.org/10.1002/jhm.468 19753568 12 Kessler DO Auerbach M Pusic M Tunik MG Foltin JC A randomized trial of simulation-based deliberate practice for infant lumbar puncture skills Simul Healthc 2011 6 4 197 203 10.1097/SIH.0b013e318216bfc1 Available from: http://dx.doi.org/10.1097/SIH.0b013e318216bfc1 21527870 13 Zendejas B Cook DA Bingener J Huebner M Dunn WF Sarr MG Farley DR Simulation-based mastery learning improves patient outcomes in laparoscopic inguinal hernia repair: a randomized controlled trial Ann Surg 2011 254 3 502 209 10.1097/SLA.0b013e31822c6994 Available from: http://dx.doi.org/10.1097/SLA.0b013e31822c6994 21865947 14 Butter J McGaghie WC Cohen ER Kaye M Wayne DB Simulation-based mastery learning improves cardiac auscultation skills in medical students J Gen Intern Med 2010 25 8 780 785 10.1007/s11606-010-1309-x Available from: http://dx.doi.org/10.1007/s11606-010-1309-x 20339952 15 Bregenzer T Conen D Sakmann P Widmer AF Is routine replacement of peripheral intravenous catheters necessary? Arch Intern Med 1998 158 2 151 156 10.1001/archinte.158.2.151 Available from: http://dx.doi.org/10.1001/archinte.158.2.151 9448553 16 Fuchs PC Indwelling intravenous polyethylene catheters: Factors influencing the risk of microbial colonization and sepsis JAMA 1971 216 9 1447 1450 10.1001/jama.1971.03180350025005 Available from: http://dx.doi.org/10.1001/jama.1971.03180350025005 4931311 17 Ena J Cercenado E Martinez D Bouza E Cross-sectional epidemiology of phlebitis and catheter-related infections Infect Control Hosp Epidemiol 1992 13 1 15 20 10.2307/30146963 Available from: http://dx.doi.org/10.2307/30146963 1545109 18 Miller JM Goetz AM Squier C Muder RR Reduction in nosocomial intravenous device-related bacteremias after institution of an intravenous therapy team J Intraven Nurs 1996 19 2 103 106 8852171 19 Soifer NE Borzak S Edlin BR Weinstein RA Prevention of peripheral venous catheter complications with an intravenous therapy team: a randomized controlled trial Arch Intern Med 1998 158 5 473 477 10.1001/archinte.158.5.473 Available from: http://dx.doi.org/10.1001/archinte.158.5.473 9508225 20 Jünger J Nikendei C OSCE Prüfungsvorbereitung Innere Medizin 2005 Stuttgart Thieme 21 Downing SM Tekian A Yudkowsky R Procedures for establishing defensible absolute passing scores on performance examinations in health professions education Teach Learn Med 2006 18 1 50 57 10.1207/s15328015tlm1801_11 Available from: http://dx.doi.org/10.1207/s15328015tlm1801_11 16354141 22 Graber MA Wyatt C Kasparek L Xu Y Does simulator training for medical students change patient opinions and attitudes toward medical student procedures in the emergency department? Acad Emerg Med 2005 12 7 635 639 10.1111/j.1553-2712.2005.tb00920.x Available from: http://dx.doi.org/10.1111/j.1553-2712.2005.tb00920.x 15995096 23 Barsuk JH Cohen ER Feinglass J McGaghie WC Wayne DB Use of simulation-based education to reduce catheter-related bloodstream infections Arch Intern Med 2009 169 15 1420 1423 10.1001/archinternmed.2009.215 Available from: http://dx.doi.org/10.1001/archinternmed.2009.215 19667306 24 Barsuk JH McGaghie WC Cohen ER O'Leary KJ Wayne DB Simulation-based mastery learning reduces complications during central venous catheter insertion in a medical intensive care unit Crit Care Med 2009 37 10 2697 2701 10.1097/CCM.0b013e3181a57bc1 Available from: http://dx.doi.org/10.1097/CCM.0b013e3181a57bc1 19885989 25 Cook DA If you teach them, they will learn: why medical education needs comparative effectiveness research Adv Health Sci Educ Theory Pract 2012(17) 305 310 10.1007/s10459-012-9381-0 Available from: http://dx.doi.org/10.1007/s10459-012-9381-0 22696095

PMC005xxxxxx/PMC5003139.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00104810.3205/zma001048Doc49urn:nbn:de:0183-zma0010484ArticleCommittee on Veterinary Medicine at the Society for Medical Education: Skills Labs in Veterinary Medicine – a brief overview Ausschuss Tiermedizin der GMA: Skills Labs in der Tiermedizin – ein kurzer Einblick Dilly Marc 1Gruber Christian *21 Stiftung Tierärztliche Hochschule Hannover, Clinical Skills Lab, Hannover, Germany2 Stiftung Tierärztliche Hochschule Hannover, BEST-VET, Hannover, Germany*To whom correspondence should be addressed: Christian Gruber, Stiftung Tierärztliche Hochschule Hannover, BEST-VET, Bünteweg 2, D-30559 Hannover, Germany, E-mail: christian.gruber@tiho-hannover.de15 8 2016 2016 33 4 Clinical skillsDoc4930 10 2015 08 7 2016 08 7 2016 Copyright © 2016 Dilly et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001048.shtmlSince 2012, skills labs have been set up to teach practical skills at veterinary training facilities in the German-speaking world. In addition to didactic considerations, ethical points of view in terms of animal protection form the basis of the increasing significance of skills labs in veterinary medicine. Not least because of the quality standards in veterinary medicine training which apply across Europe, the link between veterinary medicine training facilities is particularly significant when it comes to the setting up and development of skills labs. The Committee on Veterinary Medicine is therefore not only interested in exchange and cooperation within veterinary medicine, but also sees an opportunity for mutual gain in the link with the Society for Medical Education Committee “Practical Skills”. Zusammenfassung An den tiermedizinischen Ausbildungsstätten im deutschsprachigen Raum wurden seit 2012 Skills Labs für die Vermittlung praktischer Fertigkeiten eingerichtet. Neben didaktischen Überlegungen sind auch ethische Gesichtspunkte im Hinblick auf den Tierschutz die Grundlage für die zunehmende Bedeutung von Skills Labs in der Veterinärmedizin. Nicht zuletzt durch europaweit geltende Qualitätsnormen der tiermedizinischen Ausbildung ist die Vernetzung zwischen den tiermedizinischen Ausbildungsstätten von großer Bedeutung für die Einrichtung und Weiterentwicklung von Skills Labs. Der Ausschuss Tiermedizin ist daher nicht nur an dem Austausch und Kooperationen innerhalb der Veterinärmedizin interessiert, sondern sieht auch in der Vernetzung mit dem GMA-Ausschuss „Praktische Fertigkeiten“ die Möglichkeit der gegenseitigen Bereicherung. Skills LabSimulatorsSimulation-based TrainingClinical Skills TrainingVeterinary Medical Education ==== Body Introducing and presenting the Society for Medical Education Committee on Veterinary Medicine The Society for Medical Education Committee on Veterinary Medicine includes (“officially” as it were) all members of the Society for Medical Education from veterinary medicine and is therefore an exceptional case (along with the Committee on Dentistry) compared to the other committees, the majority of which focus on specific content, as people are assigned to the committee based on their profession. In this sense, the committee takes on all topics that are of national significance for the seven veterinary medicine training facilities (Berlin, Gießen, Hanover, Leipzig, Munich, Vetsuisse and Vienna) in the German-speaking world. Accordingly, there are links in terms of content to most committees and to the committee “Practical Skills”, because the students’ training in practical skills of course also occurs in veterinary medicine. Until recently, the intensive practical training took place almost exclusively on the animal or on (parts of) the cadaver. This type of training is not only extremely resource-intensive but is also being increasingly questioned from a didactic perspective and from both an ethical and animal protection-related perspective. As a result, in 2012 what are known as “skills labs” were created at various locations in the German-speaking world [3], [4] inspired by the international models which primarily come from the English-speaking world in which skills labs have been part of the day-to-day in (veterinary) medical training for some time. There are currently centrally established skills labs at five veterinary medical training facilities in Germany, Austria and Switzerland: Vienna, Hanover, Leipzig, Gießen and Bern. Skills labs in veterinary medical training – the same and yet different In human medicine in Germany and other countries there are now national catalogues of learning objectives while teaching at veterinary medical training facilities in Europe is organised at a supranational level and is subject to a uniform quality standard set down by the European Commission and is checked at regular intervals by visits carried out on behalf of the European Commission. The core content of the Europe-wide uniform quality standards are what are known as the “day one skills”, in other words skills and abilities that the students must have obtained by the end of their training and therefore by the “first day of their professional lives” [http://www.eaeve.org/fileadmin/downloads/sop/SOP_Annex4to8_Hanover09.pdf checked on 30 October 2015]. From a conceptual perspective they therefore correspond to what are known as the Entrusted Professional Activities (EPAs) in human medical training [2]. At around 50% of all of the skills listed in the day-one skills, practical skills make up the largest group by far and are therefore the focus of the training, leading to an increasing significance of the skills labs. Like in human medicine, complex processes in everyday clinic work are a major challenge for students, requiring the use of simulators and models with varying degrees of complexity. This not only facilitates students’ ability to gain practical skills, but concerns and anxiety about making mistakes with serious consequences for the animal while gaining practical skills are able to be alleviated [8]. The low number of training facilities compared to those for human medicine and the even lower number of skills labs meant that the development and sourcing of models was often a hurdle when setting up skills labs, as models for various species of animal were needed and therefore it was understandably not possible to rely on human medicine models. The relatively small market in veterinary medicine meant that animal models of this kind, however, were difficult to obtain and/or expensive. The rise in the number of skills labs, however, means that an increasing number of models and simulators are available commercially or are being developed in-house [1], [5], [6], [7], [8], [9], [10], [11], [12]. The links between veterinary medical training facilities, such as the online group “Veterinary Clinical Skills & Simulation” in the network NOVICE (Network Of Veterinarians In Continuing Education; http://www.noviceproject.eu/), founded in 2010 and dedicated exclusively to the teaching of practical clinical skills, are particularly helpful. The group is made up of more than 300 members from more than 30 countries. In addition to links on online platforms, new meetings and conferences have also been established for exchange and the promotion of collaboration focusing on simulation-based teaching in veterinary medicine. The organisation “International Veterinary Simulation in Teaching” (InVeST, http://www.vetedsimulation.com), for example, was founded in 2011 and primarily looks at the development and validation of simulators and methods of communication in veterinary medical training. The first skills lab symposium in the Germany/Austria/Switzerland region took place in January 2014 in Hanover. Collaboration of the committees In both of the activities mentioned above – links via platforms and at thematically relevant conferences – the greatest opportunity for collaboration for the Veterinary Medicine Committee is with the Committee on Practical Skills. Veterinary medicine has already been part of the simulator network established by the Committee on Practical Skills for two years. While there were hardly any contributions in the initial phase, participation has increased significantly in the past year and 29 models and simulators from veterinary medicine are now included in the network. More are to follow. There is also an increasing presence of veterinary medicine at physical meetings in the form of conferences, so in the future the “International Skills Lab Symposium (iSLS)” organised by the Society for Medical Education Committee on Practical Skills will include contributions from veterinary medicine. Competing interests The authors declare that they have no competing interests. ==== Refs 1 Baillie S Utilization of simulators in veterinary training Cattle Pract 2007 15 3 244 248 2 Berberat PO Harendza S Kadmon M Entrustable Professional Activities – Visualization of Competencies in Postgraduate Training. Position Paper of the Committee on Postgraduate Medical Training of the German Society for Medical Education (GMA) GMS Z Med Ausbild 2013 30 4 Doc47 10.3205/zma000890 Available from: http://dx.doi.org/10.3205/zma000890 24282450 3 Crowther E Booth N Coombes N Baillie S Veterinary Clinical Skills Labs: Online Collaboration and Moving Forward Health Soc Care Educ 2013 2 1 39 43 4 Dilly M Tipold A Schaper E Ehlers JP Setting Up a Veterinary Medicine Skills Lab in Germany GMS Z Med Ausbild 2014 31 2 Doc20 10.3205/zma000912 Available from: http://dx.doi.org/10.3205/zma000912 24872855 5 Engelskirchen S Rosenthal J Hungerbuehler S Dilly M Development of a dog simulator for ultrasonic based puncture of the urinary bladder 2015 InVeST 2015: International Veterinary Simulation in Teachin Conference 14.-16.09.2015 Hannover Düsseldorf German Medical Science GMS Publishing House Doc15invest05 10.3205/15invest05 Available from: http://dx.doi.org/10.3205/15invest05 6 Fox V Sinclair C Bolt DM Lowe J Weller R Design and Validation of a Simulator for Equine Joint Injections J Vet Med Educ 2013 40 2 152 157 10.3138/jvme.0912-083R1 Available from: http://dx.doi.org/10.3138/jvme.0912-083R1 23709111 7 Gerke L Barrett DC Arnold C Hale-Mitchell L Baillie S Synthetic Models for Teaching Farm Animal Technical and Clinical Skills to Veterinary Undergraduates Cattle Pract 2015 23 1 20 26 8 Giese H Hilke J Gundelach Y Dilly M Validation of a bovine vascular access model for teaching students a technique for placing catheter in the auricular vein of cattle 2015 InVeST 2015: International Veterinary Simulation in Teachin Conference 14.-16.09.2015 Hannover Düsseldorf German Medical Science GMS Publishing House Doc15invest36 10.3205/15invest36 Available from: http://dx.doi.org/10.3205/15invest36 9 Langebæk R Eika B Tanggaard L Jensen AL Berendt M Emotions in Veterinary Surgical Students: A Qualitative Study J Vet Med Educ 2012 39 4 312 321 10.3138/jvme.0611.068R1 Available from: http://dx.doi.org/10.3138/jvme.0611.068R1 23187024 10 Lin YW Lüpke M Tipold A Ehlers JP Dilly M Development of a Dog Manikin-bases Simulator for Epidural Puncture and Atlantooccipital CSF Collection 2013 26th Annual Symposium of the ESVN-ECVN 26.-28.09.2013 Paris (F) Paris ESVN-ECVN 11 McGaghie WC Issenberg SB Petrusa ER Scalese RJ A critical review of simulation-based medical education research: 2003–2009 Med Educ 2010 44 1 50–63 10.1111/j.1365-2923.2009.03547.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2009.03547.x 20078756 12 Scalese RJ Issenberg B Effective Use of Simulations for the Teaching and Acquisition of Veterinary Professional and Clinical Skills J Vet Med Educ 2005 32 4 461 467 10.3138/jvme.32.4.461 Available from: http://dx.doi.org/10.3138/jvme.32.4.461 16421829

PMC005xxxxxx/PMC5003140.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00104710.3205/zma001047Doc48urn:nbn:de:0183-zma0010473Article"Practical skills" – Positioning of the GMA committee for dentistry "Praktische Fertigkeiten" – Standortbestimmung des GMA-Ausschusses Zahnmedizin Scheutzel Petra *1Gerhard-Szép Susanne 21 Universitätsklinikum Münster, Poliklinik für Prothetische Zahnmedizin & Biomaterialien, Münster, Germany2 Goethe-Universität, Carolinum Zahnärztliches Universitäts-Institut gGmbH, Poliklinik Zahnerhaltungskunde, Frankfurt am Main, Germany*To whom correspondence should be addressed: Petra Scheutzel, Universitätsklinikum Münster, Poliklinik für Prothetische Zahnmedizin & Biomaterialien, Waldeyerstr. 30, D-48149 Münster, Germany, E-mail: scheutz@uni-muenster.de15 8 2016 2016 33 4 Clinical skillsDoc4824 11 2015 28 6 2016 24 11 2015 Copyright © 2016 Scheutzel et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001047.shtmlThe GMA committee for dentistry of the German Society for Medical Education (GMA) considers its’ main purpose the representation and interconnection of all aspects of dentistry with and within the GMA. Teaching and assessing practical skills during training is traditionally of great importance in dental education. This is also reflected in the National Competence Based Catalogue of Learning Objectives for Dental Education (NKLZ). Practical skills are not comprised in a separate chapter as they are in the National Competence Based Catalogue of Learning Objectives for Medical Education (NKLM), but are considered in all sections of the NKLZ for the purpose of interdisciplinary patient- or disease-specific application, targeting the educational level of acting competency. The implementation of the associated joined interdisciplinary integrated educational concept has undoubtedly been a challenge for dental curriculum development against the backdrop of German Dental Licensure Act dating back to 1955. Zusammenfassung Der Ausschuss Zahnmedizin der Gesellschaft für Medizinische Ausbildung (GMA) sieht seine Hauptaufgabe in der themenübergreifenden Vertretung und Vernetzung des Faches Zahnmedizin mit und innerhalb der GMA. In der Zahnmedizin ist die Vermittlung und Überprüfung praktischer Fertigkeiten traditionell von großer Bedeutung in der Ausbildung. Dies spiegelt sich auch im Nationalen Lernzielkatalog Zahnmedizin (NKLZ) wieder. Die praktischen Fertigkeiten sind nicht wie im Nationalen Lernzielkatalog Medizin (NKLM) in einem eigenen Kapitel zusammengefasst, sondern finden im Sinne einer fachübergreifenden patienten- bzw. erkrankungsspezifischen Anwendung in allen Abschnitten des NKLZ Berücksichtigung, und es wird häufig der Ausbildungslevel der Handlungskompetenz angestrebt. Die Umsetzung des damit verbundenen integrierten fachübergreifenden Ausbildungskonzeptes stellt für die zahnärztliche Curriculumsentwicklung vor dem Hintergrund einer seit 1955 gültigen Approbationsordnung zweifellos eine Herausforderung dar. skillspractical skillsdental educationdental simulatorDental Licensure Act ==== Body 1. Introduction Duties and Objectives of the GMA Committee for Dentistry The GMA committee for dentistry holds, as well as the committee for veterinary medicine, a special status in comparison to the other committees insofar as it represents a discipline and not a single specific topic. Objective of the committee for dentistry is the linking of both dental teachers and dental learners to the German Society for Medical Education (GMA) and notably the interconnection between GMA and the Working Group for the Advancement of Dental Education (AKWLZ), which was institutionalized by the German Society for Dental and Oral Medicine (DGZMK) and the German Society for University Teachers of Dentistry and Oral Medicine (VHZMK). The GMA committee for dentistry, in close cooperation with the AKWLZ, deals with all matters concerning dental training, e.g. curriculum development, advancement and implementation of teaching, learning and assessment methods, the promotion of educational research and development and implementation of catalogues of learning objectives. The main focus lies with introducing and broadening knowledge gained in dental education into the interdisciplinary discourse with the other medical disciplines. 2. Points of Contact with/Similarities to/Distinctions from the GMA Committee for Practical Skills The GMA Committee for Practical Skills has developed from an initiative of German skills labs in medical education and consists of a network of these training facilities for practical medical skills in Germany. Consequently, the exchange of experience between the existing training facilities, support for the construction of new Skill Labs within the framework of medical training and research concerning the sharing of practical skills as part of Skill Labs form the basis of the committee’s work [https://gesellschaft-medizinische-ausbildung.org/ausschuesse/praktische-fertigkeiten.html]. Based on this, the GMA Committee for practical skills has participated significantly in the development of NKLM work package 14b, “Clinical and Practical Skills” [http://www.nklm.de/kataloge/nklm/lernziel/uebersichthttp://www.nklm.de/kataloge/nklm/lernziel/uebersicht]. The importance of imparting and verifying practical skills in medical education, which is reflected in this chapter of the NKLM, has only been initiated by the amendment of the German Medical Licensure Act for Physicians in 2002, increasing the importance of professional competencies in undergraduate medical education [1]. Contrary to this, practical skills always have been and still are accorded greater importance in undergraduate dental education. According to §1 of the still valid German Dental Licensure Act (Zahnärztliche Approbationsordnung - ZÄApprO) of 1955 [http://www.gesetze-im-internet.de/z_pro/BJNR000370955.html] which states:” For his profession the dentist is trained scientifically and practically”, both preclinical and clinical part of undergraduate dental curriculum contain several practical courses with training exercises on the dental simulator as well as laboratory courses and patient treatment in accordance with the ZÄApprO. Practical skills imparted during undergraduate dental education are in addition to cognitive skills also part of the oral-practical state exams. This is also represented in the NKLZ [http://www.nklz.de/kataloge/nklz/lernziel/uebersicht] that mostly aims for competency level 3b (independent and adequate treatment with awareness of consequences) in both basic dental competency as well as professional competency. 3. Future Joint Fields of Activity and Research Despite the fact that the relaying of practical medical skills within the scope of skills labs has so far formed the primary basis for the GMA practical skills committee’s work and thus provides fewer direct connection points to the specific situation regarding dentistry, there nevertheless are entry points for a future cooperation against the background of NKLM and NKLZ and advancement in dental simulators from the GMA dental committee’s point of view: Interdisciplinary and disease- or patient-related taught practical skills on competence level 3a and b require the establishment of workplace based exams in an actual clinical treatment environment – this lends itself to an exchange and collaboration with the GMA committees for practical skills and assessment. Furthermore, in addition to traditional dental simulators, virtual 3-D-patients based on haptic technology become more and more available, providing the means for practising practical skills (e.g. preparation of teeth) in the context of individual patient history, including aspects of diagnostic and differential therapy [2]. Here dentistry might possibly benefit from insights gained in medical skill labs when it comes to the question of how far practical skills and diagnostic abilities practised through self-study or under the instruction of peers could increase competence concerning treatment of actual patients. Notes Authors have written this report in their capacity as vice chairperson1 und chairperson2 of the GMA committee for dentistry. Competing interests The authors declare that they have no competing interests. ==== Refs 1 Schnabel KP Boldt PD Breuer G Fichtner A Karsten G Kujumdshiev S Schmidts M Stosch C Konsensusstatement "Praktische Fertigkeiten im Medizinstudium" - ein Positionspapier des GMA-Ausschusses für praktische Fertigkeiten GMS Z Med Ausbild 2011 28 4 Doc58 10.3205/zma000770 Available from: http://dx.doi.org/10.3205/zma000770 22205916 2 de Boer IR Wesselink PR Vervoorn JM Student performance and appreciation using 3D vs. 2D vision in a virtual learning invironment Eur J Dent Educ 2016 20 3 142 147 10.1111/eje.12152 Available from: http://dx.doi.org/10.1111/eje.12152 26072997

PMC005xxxxxx/PMC5003141.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00105810.3205/zma001058Doc59urn:nbn:de:0183-zma0010588ArticleImproved self- and external assessment of the clinical abilities of medical students through structured improvement measures in an internal medicine bedside course Verbesserte Selbst- und Fremdeinschätzung klinischer Fähigkeiten bei Medizinstudierenden durch strukturierte Verbesserungsmaßnahmen in einem internistischen Bedside-Kurs Fünger S. M. 1Lesevic H. 1Rosner S. 1Ott I. 1Berberat P. 2Nikendei C. 3Sonne C. *141 Technical University Munich, German Heart Centre, Munich, Germany2 Technical University Munich, Klinikum Rechts der Isar, TUM MeDiCAL, Centre of Medical Education, Munich, Germany3 Heidelberg University Hospital, Department of General Internal Medicine & Psychosomatic, Heidelberg, Germany4 Praxis, Maroussi, Greece*To whom correspondence should be addressed: C. Sonne, Praxis, Leof. Kifissias 195/I. Doussi 23, GR-15124 Maroussi, Greece15 8 2016 2016 33 4 Clinical skillsDoc5928 9 2015 30 6 2016 15 6 2016 Copyright © 2016 Fünger et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001058.shtmlBackground: Bedside courses are of outstanding importance when training medical students. The fact that less and less teaching is taking place nowadays at the patient's bedside makes it all the more important that the available time be put to effective use. The aim of this study was to check whether structured improvement measures in the course (scripts, lecturer briefing, e-learning cases) would improve the abilities of the students on the basis of a subjective self-assessment as well as an external assessment by the lecturers with respect to clinical abilities. Methods: Bedside teaching takes place in the fourth study year in the Medical Clinics of the TU Munich. Both students and lecturers had the chance to hand in an anonymous, quantitative self- and external assessment of the clinical abilities of the students (German grading system) after every course date. This assessment took place online in the three categories "Medical history & examination", "Diagnosis" and "Therapy". An overall period of four semesters, each with 6 course dates, was investigated. After two of the total of four semesters in the study, the course was changed by introducing scripts, lecturer briefing as well as interactive e-learning cases. The self- and external assessment was compared both within the semester (date 1-3: A; date 4-6: B), during the course as well as before and after introducing the improvement measures ("before" (T0): SS 2012, SS 2013, "after" (T1): WS 2013/2014, SS 2014). Results: There was a significant improvement in one's own abilities on the basis of the self-assessment within each semester when comparing the first (A) and the last (B) course dates. Moreover, there was a significant improvement in the performances in all three categories when T0 was compared with T1, from both the point of view of the students ("Medical history & examination": T0 =2.5±0.9, T1=2.2±0.7, pp<0.001; "Diagnosis" T0=3.1±1.0, T1=2.8 ±0.9, pp<0.001; "Therapy": T0=3.8±1.3, T1=3.5±1.2, pp<0.018) and in two of the three categories from the point of view of the lecturers ("Diagnosis": T0=3.0±1.0, T1=2.7±0.7, p.=0.028; "Therapy": T0=3.8±1.1, T1=3.1±1.0, p<0.001). Summary: The structured measures to improve the course including the interactive e-learning cases could have contributed to improved practical abilities with respect to the medical history and examination techniques as well as diagnostic and therapeutic thinking. The external evaluation by lecturers confirmed the improvement with respect to the diagnostic and therapeutic abilities. They only saw no dynamic change in the student's taking histories and clinical examinations. Zusammenfassung Hintergrund: Bedside-Kurse sind in der Ausbildung von Medizinstudierenden von herausragender Bedeutung. Die Tatsache, dass heutzutage immer weniger Unterricht am Krankenbett stattfindet, lässt es umso wichtiger erscheinen, die vorhandene Zeit effektiv zu nutzen. Ziel der Studie war es zu überprüfen, ob sich durch strukturierte Verbesserungsmaßnahmen des Kurses (Skripte, Dozentenbriefing, E-Learning-Fälle) die Fähigkeiten der Studenten auf Basis einer subjektiven Selbsteinschätzung sowie einer Fremdbeurteilung durch die Dozenten im Hinblick auf klinische Fähigkeiten verbessern. Methoden: Im vierten Studienjahr findet in den Medizinischen Kliniken der TU München ein Bedside-Teaching statt. Sowohl Studenten als auch Dozenten hatten nach jedem Kurstermin die Möglichkeit eine anonyme, quantitative Selbst- und Fremdeinschätzung der klinischen Fähigkeiten der Studenten abzugeben (deutsches Notensystem). Diese Einschätzung erfolgte online in den drei Kategorien „Anamnese & Untersuchung“, „Diagnose“ und „Therapie“. Insgesamt wurde ein Zeitraum von vier Semestern mit je 6 Kursterminen untersucht. Nach zwei der insgesamt vier untersuchten Semester wurde der Kurs durch die Einführung von Skripten, einem Dozentenbriefing sowie interaktiven E-Learning Fällen verändert. Die Selbst- und Fremdeinschätzung wurde sowohl innerhalb der Semester (Termin 1-3: A; Termin 4-6: B), im Laufe des Kurses, als auch vor und nach Einführung der Verbesserungsmaßnahmen („vorher“ (T0): SS 2012, SS 2013, „nachher“ (T1): WS 2013/2014, SS 2014) verglichen. Ergebnisse: Innerhalb eines jeden Semesters zeigte sich auf Basis der Selbsteinschätzung eine signifikante Verbesserung der eigenen Fähigkeiten beim Vergleich der ersten (A) mit den letzten (B) Kursterminen. Außerdem zeigte sich eine signifikante Verbesserung der Leistungen in allen drei Kategorien beim Vergleich von T0 zu T1 sowohl aus der Perspektive der Studenten („Anamnese & Untersuchung“: T0 =2,5±0,9, T1=2,2±0,7, p<0,001; „Diagnose“: T0=3,1±1,0, T1=2,8 ±0,9, p<0,001; „Therapie“: T0=3,8±1,3, T1=3,5±1,2, p<0,018) sowie bei zwei der drei Kategorien aus der Sicht der Dozenten („Diagnose“: T0=3,0±1,0, T1=2,7±0,7, p=0,028; „Therapie“: T0=3,8±1,1, T1=3,1±1,0, p<0,001). Zusammenfassung: Die strukturierten Maßnahmen zur Verbesserung des Kurses inklusive der interaktiven E-Learning Fälle könnten zu verbesserten praktischen Fähigkeiten im Hinblick auf Anamnese- und Untersuchungstechniken sowie diagnostisches und therapeutisches Denken beigetragen haben. In der Fremdevaluation durch Dozenten konnte die Verbesserung im Hinblick auf diagnostische und therapeutische Fähigkeiten bestätigt werden. Lediglich in der Anamneseerhebung und klinischen Untersuchung der Studenten sahen sie keine Dynamik. medical historyphysical examinationmedical educatione-learningevaluationself-assessmentmedical examination ==== Body Notes The authors Lesevic and Nikendei = shared initial authorship and shared final authorship. Author Sonne: Master's thesis for the Master of Medical Education, Ruprecht-Karls-University Heidelberg University Hospital Masculine names used grammatically in this article apply equally for both male and female persons. Introduction Teaching at the patient's bedside, so-called bedside teaching (BST), has been an important part of training medical students since ancient times [1]. The major part of a human medicine degree course nowadays is made up of lectures and seminars at many places. This teaching is being increasingly supplemented by practical exercises (skill courses, simulation training). The share of teaching that takes place directly at the patient's bedside and that focuses on the patient continues to decrease [2], [3], [4]. Nevertheless, bedside courses are eminently important to learn about the clinical pictures, to practice history-taking and examination techniques, a structured clinical way of thinking, though also to preserve basic "soft skills" [3], [5], [6]. What's more, up to 70% of the correct suspected diagnoses can made by history-taking and physical examinations [7]. The students also expressed a clear wish for more practical exercises [8], [9] and supportive medical assistance when learning the most important medical abilities [10]. Apart from learning clinical abilities, BST offers some decisive advantages. For example, it could be shown amongst other things that bedside courses (BSC) have a positive effect on the doctor-patient relations and promote the patient's understanding of their illness, and thus have significant effects on the attitude of the prospective doctors [3]. Despite the numerous advantages of bedside teaching, its integration in everyday hospital routines often poses problems [3]. Staff shortages as well as tightly scheduled procedures on the wards make a student-based learning environment a real challenge [11], [12]. This make a constant analysis of and improvement in the quality of BST based on feedback from both students and lecturers all the more important. A study by Gonzalo et al. has already described various strategies on the preparation and performance of bedside courses by experienced BSC lecturers, such as a good choice and preparation of patients, explicit assignment of the students' roles, practical instruction of the students on patients and the encouragement of feedback [13]. In our own study, it could be shown that following the introduction of simple, structured improvement measures, the students rated their clinical abilities much higher than beforehand. The measures included materials for a better preparation of the students such as examination scripts, lecturer briefings and the definition of global learning objectives by the lecturers [14]. In the fourth study year of the human medicine degree course at the Technical University Munich there is an extensive BSC in internal medicine. The goal of this course is the further development of the history-taking and physical examination techniques that have already been learnt in the third semester. Moreover, the student's diagnostic and therapeutic way of thinking should be improved so that the students are able to draw up a "diagnosis" and therapy plan for the frequent clinical pictures that are presented on their own. Based on the feedback from the online questionnaire, suggestions for concrete improvements to the BST were drawn up and implemented. These include the introduction of scripts, lecturer briefing as well as the possibility of preparation for and follow-up to the course with the aid of e-learning cases. The latter has already been described as an effective method to supplement the direct teaching [15], [16], [17], [18], [19]. The aim of the study was to investigate whether and to what extent the introduction of improvement measures enhances the clinical abilities of the students on the basis of the self-assessment as well as an external assessment by the lecturers. In addition to the quantitative parameters, the qualitative free text answers of the students can indicate desired improvements and express their satisfaction with respect to the performance of the course. Methods Bedside teaching The course was held for students in their fourth study year at the Technical University Munich once a week for three hours during the semester (a total of 6 dates per semester). The students visited internal medicine wards of the nephrology, pneumology, cardiology, rheumatology and gastroenterology specialties in small groups of up to eight students. The hospitals involved were the I and II Medical Clinics of the Klinikum Rechts der Isar, the German Heart Centre Munich, the Klinikum Bogenhausen as well as the Barmherzige Brüder Hospital in Munich. Sample group The students were supervised by one lecturer on each ward during the course. In the summer semester 2012 (SS 2012) there were 165, in the summer semester 2013 (SS 13) 200, in the winter semester 2013/2014 (WS 13/14) 205 and in the summer semester 2014 (SS 14) 152 students in the course. 60 lecturers were assigned to the course each semester. Evaluation All of the students and lecturers were contacted by e-mail after every day of the course and asked to complete an online questionnaire. The evaluation took place anonymously by means of an online questionnaire after every day of the course. This procedure has already been used successfully in previous publications [14], [20]. Both the students and lecturers were reminded about handing in the assessment each week. The evaluation took place over four semesters (SS 12, SS 13, WS 13/14, SS 14). The students were asked for a self-assessment of their performance with respect to the history-taking and examination as well as diagnostic testing and therapy planning as follows: Do you now feel able to perform history-taking and a physical examination? Do you now feel able to draw up a plan for the relevant diagnostic steps for the most common illnesses in internal medicine? Do you now feel able to draw up a plan for the relevant therapeutic steps for the most common illnesses in internal medicine? The lecturers were asked to assess the students' performance in the same categories: Are the students now able to perform history-taking and a physical examination? Are the students now able to draw up a plan for the relevant diagnostic steps for the most common illnesses in internal medicine? Are the students now able to draw up a plan for the relevant therapeutic steps for the most common illnesses in internal medicine? The evaluation was carried out using the German school grading system. "1" indicated the best, "6" the poorest grade (1=very good, 2=good, 3=satisfactory, 4=adequate, 5=poor, 6=inadequate). The respondents were also able to make suggestions for improvements ("In general, what would you change?") and say which parts of the course they though were particularly good ("What did you think was particularly good?") in a free text field. Feedback from the evaluations by students and lecturers The number of feedbacks from the evaluations by students and lecturers are shown in Table 1 (Tab. 1) and Table 2 (Tab. 2). Structured improvement of the course content and forms of teaching So as to achieve an improvement in the BSC, the course content was revised and global as well as specific learning objectives were defined. In order to cater for the wishes and suggestions for improvements of the students and lecturers involved, resulting from the evaluations of this course as well as the feedback from students in other studies on the optimisation of practical courses [14], structured improvement measures were integrated in the course structure (see Figure 1 (Fig. 1)). This led to the introduction of scripts in WS 13/14. The scripts contained medical history sheets as well as instructions on how to carry out patient interviews. Structured instructions and tips were also provided on different internal medicine examinations, e.g. a thorax examination. Apart from the physiological findings, pathologies were also often explained along with their clinical evaluation. A targeted preparation for and follow-up to the course by the students was possible thanks to the prior division over the various internal medicine wards and the associated specialist fields. The lecturers were also able to prepare for the individual days of the course with the help of the scripts. In addition to these scripts, a lecturer briefing was held in the morning meetings in each of the hospitals during which all of the lecturers were instructed in the content of the course on the basis of the scripts and the desired evaluation by the students, and during which the importance of supervising the students during history-taking and the physical examination was stressed. Furthermore, a contact person was appointed for the courses whose work included regular reminders of the course and who also helped supervise these. E-learning cases were introduced simultaneously in the winter semester 2013/14. The lecturers were instructed to recommend these to the students as a voluntary form of preparation for and follow-up to the course. These were patient cases that had been created with the program Articulate®, an e-learning software. These cases were taken from the internal medicine disciplines of cardiology, pneumology, gastroenterology, haematology, oncology as well as nephrology and endocrinology. These could be accessed via the Moodle platform of the Technical University Munich, which is open to all students of the TU, after they had registered for the course. The 14 cases constituted an interactive elaboration of a clinical picture on the basis of a concrete patient with his medical history. A quiz in the form of multiple choice questions was linked to information about the respective illness and the associated diagnostic testing and therapy (see Figure 2 (Fig. 2) and Figure 3 (Fig. 3)). Statistical analysis The grades (man values, MV, and standard deviations, SD) of the first half (date 1-3, A) of the evaluations via the self-/external assessment were tested against those of the second half (B) of the evaluations in each case within each semester. This should allow an assessment of the students' performance in the course of each semester. Two semesters were summarised in each case for an analysis of the development of the students' performance compared to before ("before" (T0): summer semester 2012 and 2013) and after the introduction of the improvement measures ("after" (T1): winter semester 2013/2014 and summer semester 2014). These analyses were also carried out separately for the students and lecturers. Since the questionnaires were completely anonymised, the results could not be evaluated jointly. An explorative statistical analysis of the evaluation share was therefore performed with a Mann-Whitney-U-Test for independent random samples. A p-value of p<0.05 is rated as being statistically significant (SPSS 23, SPSS Inc. Chicago, IL, USA). Free text answers The students also had the possibility of providing feedback in the form of a free text answer to the questions "In general, what would you change?" and "What did you think was particularly good?". All free text answers to a question were screened and the tips, suggestions and comments that were provided were then classified in categories. The individual free text answers of the students and lecturers were then assigned to these categories separately for the periods before (T0) and after the introduction of the improvement measures (T1) and their frequency evaluated. The three most common answers in a period were worked out, provided these had been given by at least 10% of the respondents. Results Comparison within each semester (A and B) The MV (±SD) of the self-assessments with respect to the categories "Medical history and examination", "Diagnosis" and "Therapy" of the four semesters are shown in Figure 4. The performances of the students improve significantly in the summer semester 2012 according to their self-assessment in the second half of the course (B) in the categories "Medical history & examination" (MV A=3.0±1.1, MV B=2.7±1.0, p.=0.005), and "Diagnosis” (MV A=3.5±1.2, MV B=3.1±1.1, p.=0.004) and "Therapy" (MV A=4.5±1.1, MV B=4.0±1.4, p<0.001) to the performances in the first half of the course (A). A significant improvement can also be identified with respect to the students' self-assessment in the summer semester 2013 ("Medical history & examination": MV A=2.7±0.8, MV B=2.4±0.7, p<0.001, "Diagnosis": A=30.5±0.9, MV B=3.1±0.9, p<0.001, "Therapy": MV A=4.2±1.1, MV B=3.6±1.1, p<0.001). In WS 13/14 the assessments also improve between the first and second half of the course for all three categories ("Medical history & examination“: MV A= 2.7±1.0, MV B=2.1±0.5, p<0.001; "Diagnosis" MV A=3,6±1.1, MV B=2.9±0.9, <0.001; "Therapy": MV A=4.8±1.1, MV B=3.9±1.1, <0.001). A similarly significant effect can be seen for the summer semester 2014 in the three categories ("Medical history & examination“: MV A=2,0.7±0.9, MV B=2.2±0.7, <0.001; "Diagnosis" MV A=3.4±1.0, MV B=2.8±0.9, p<0.001; "Therapy": MV A=4,0.0±1.2, MV B=3.4±1.3, <0.001). The external assessments of the students' performance by the lecturers are also shown in Figure 4 (Fig. 4). The external assessment of the students' performance by the lecturers also improves visually over the semester (see Figure 4 (Fig. 4)). This altered evaluation is not significant. Comparison before and after the improvement measures (T0 and T1) The trend of the students' evaluation as a comparison of T0 and T1 is shown in Figure 5 (Fig. 5). An improvement in all three categories can be seen in SS12 to SS14 in the self-assessment. The difference in the self-assessment grades between T0 and T1 is significant for the categories "Medical history and examination" (MV T0=2.5±0.9, MV T1=2.2±0.7, p<0.001) as well as "Diagnosis" (MV T0=3.1±1.0, MV T1=2.8 ±0.9, p<0.001) and "Therapy" (MV T0=3.8±1.3, MV T1=3.5±1.2, p.=0.018). The trend for the evaluations the lecturers over the years is also shown in Figure 5 (Fig. 5). The students' improvements in the assessment of the lecturers is also visible. The difference between SS 12 and SS 13 on the one hand and between WS 13/14 and SS 14 on the other is significant for the category "Diagnosis" (MV T0=3.0±1.0, MV T1=2.7±0.7, p.=0.038) and "Therapy" (MV T0=3.8±1.1, MV T1=3.1±1.0, p<0.001). The difference in the "Medical history & examination“ is not significant (MV T0=2.4±0.8, MV T1=2.3±0.8, p.=0.798). The free text answers of the students and lecturers The free text answers of the students and lecturers for the periods before (T0) and after introduction of the improvement measures (T1) are summarised in table 3 (Tab. 3). Discussion Bedside courses are essential for teaching fundamental medical skills, such as examination and structured clinical assessment of a patient. We investigated the effect of introducing structured improvement measures with respect to self- and external assessment, with respect to medical history, examination techniques, diagnostic testing and therapy. The study was based on the bedside courses in Munich Technical University. The study groups were also asked about suggested improvements and aspects of the courses which they had particularly liked. It was found that the students considered that the students considered that their performance had significantly improved during a bedside course. This self-observed learning effect was less marked from the perspective of the lecturers. The improvements in the students' self-assessment of their clinical abilities was only found during individual semesters. Moreover, from summer semester 2012 to summer semester 2014, the students' clinical abilities at the patients' bed significantly improved after the improvement measures had been introduced - from the points of view of both the students and lecturers. Aside from the results of the quantitative analysis, the students and lecturers praised and criticized largely similar points - both before and after the introduction of improvement measures. Thus, the students stated that the groups were too large and that they wished more support and feedback from the doctors for the physical examination. However, this was stated by many fewer students after the improvement measures had been introduced. Moreover, many more students praised lecturers' motivation and commitment and the good case discussion after the introduction of E-learning and scripts. Both before and afterwards, the lecturers praised the students' motivation, but criticised their poor prior knowledge and the rooms. The results of the qualitative and quantitative analysis were presumably influenced by the systematic and structured improvement in the bedside course by the introduction of scripts, a lecturer briefing and the preparation of the students with E-learning cases. Another factor might be the lecturers' growing experience from the semesters, as these also use the scripts for their preparation of their own students. In our 2013 study, we showed that the self-assessment of the students in practical course significantly improved with the introduction of structured learning goals, examination course scripts and the provision of online material [11]. The results of this study are also consistent with an American study. This study showed that medical students' ability to perform physical examinations was significantly improved by introducing structured instructions, in comparison with unstructured learning [21]. Structured teaching sessions, e.g. in accordance with the sandwich principle, enhance the students' attentiveness and concentration, and thus support more efficient learning processes than frontal mediation, and contribute to the students' independence and a positive learning climate [22]. Moreover, the possible improvement in the students' self-assessment may be linked to their motivation. Previous studies have shown that blended learning – which describes the integration of E-learning and direct teaching - is accompanied in the students' self-assessment by significantly greater motivation, satisfaction and increase in knowledge, in comparison with a conventional course without E-learning [23]. In addition, in a randomised study in students, mediation of medical knowledge by E-learning led to significantly better test results than with conventional lectures [18]. However, the possibility should also be considered that the self-assessed clinical abilities may also have been influenced by parallel curricular elements, such as lectures in internal medicine in the fourth study year, as well as the special subjects and optional courses. In the present study, self- and external assessment were recorded with an anonymous online questionnaire. Data collection by self- and external assessment is a recognised approach and has already been used in previous studies [14], [20]. Nevertheless, this procedure has the disadvantage that the subject has to assess his own performance. For this reason, lecturers and students were instructed to use the school marking system in the evaluation, in order to reduce differences in the evaluation criteria for abilities and performance to as great an extent as possible. Moreover, the students' self-assessment could be influenced by the introduction of blended learning [23]. In addition, the present study showed that self-assessment for often more marked for students with poorer performance than for students with better performance [24]. The same study also showed that there was only poor agreement between the students' self-assessment and that of their lecturers. Both these and other studies show that there is low agreement between the students' self-evaluation and external controls of performance [24], [25]. Heidelberg University performed a case control study to optimise control. This used BST with the existing curriculum with a structured course containing three different learning units, such as communication training, examination courses and bedside teaching. The number of learning units was the same in both groups. There was no difference between the two groups in a multiple choice test, but there was a significant difference between the two groups in the results of an “objective structured clinical examination” (OSCE) [26]. OSCEs may possibly be a more valid form of checking practical clinical abilities [14]. This test format is regularly used in university medical course [27] and consists of an obstacle course, in which different practical abilities – including taking histories and physical examinations - are checked and examined. Limitations After the students had been rotated between the individual courses and were therefore taught by different lecturers, the lecturers can only assess the students' momentary status of knowledge and ability, but not their actual individual development or changes in this. Together with the lack of an objective measure of performance, e.g. in the form of an OSCE, can make it more difficult to evaluate the students throughout the course or semester. Moreover, most of the students and lecturers did not report back, so that a representative evaluation is difficult. Because of the anonymous character of the questionnaire, it is impossible to make out whether predominantly high performing or weak students report back. Moreover, there are differences between the semesters with respect to the numbers of reports back for self- and external assessment. While in the winter semester 13/14, 9% of students reported back, the corresponding figure in the summer semester 12 was 35% for the quantitative questions. For lecturers too, the assessments varied from 25% in the summer semester 13 to 55% in summer semester 12. The differences in the numbers of reports back make it more difficult to compare the semesters directly. Another limiting point is the lack of missing data for winter semester 2012/2013; these could not be integrated in this study for technical reasons and which would have been helpful for a complete comparison between different periods. Summary After introduction of structured measures, such as lecturer briefing, introduction of scripts and provision of E-learning courses for a bedside course, the students considered that their clinical abilities improved significantly. This assessment was partially confirmed by the lecturers. It cannot be conclusively decided whether the students' improved performance in medical history, diagnostic testing and therapy was only due to improvement measures. Competing interests The authors declare that they have no competing interests. Table 1 Number of feedbacks from evaluations by students Table 2 Number of feedbacks from evaluations by lecturers Table 3 Free text answers from students and lecturers for the periods before (T0) and after introduction of improvement measures (T1) Figure 1 The course design Figure 2 Patients and their symptoms were introduced in the interactive cases. Figure 3 After identifying the symptoms, the illnesses and their characteristic features are dealt with interactively and questions on these are answered. Figure 4 Self-assessment and external assessment by students and lecturers in all four semesters. The results masked with X are significant. Figure 5 Self-assessment and external assessment by students and lecturers compared before (T0) and after introduction of improvement measures (T1). The results masked with X are significant. ==== Refs 1 Porter R The Greatest Benefit to Mankind: A Medical History of Humanity (The Norton History of Science) 1999 New York WW Norton & Company 2 El Bagir K Ahmed M What is happening to bedside clinical teaching? Med Educ 2002 36 12 1185 1188 10.1046/j.1365-2923.2002.01372.x Available from: http://dx.doi.org/10.1046/j.1365-2923.2002.01372.x 12472754 3 Aldeen AZ Gisondi MA Bedside teaching in the emergency department Acad Emerg Med 2006 13 8 860 866 16766739 4 Khan IA Bedside Teaching-Making it an effective Instrucitional tool J Ayub Med Coll Abbottabad 2014 26 3 286 289 25671928 5 Goldstein EA Maclaren CF Smith S Mengert TJ Maestas RR Fay HM Wenrich MD Ramsey PG Promoting fundamental clinical skills: a competency-based college approach at the University of Washington Acad Med 2005 80 5 423 433 10.1097/00001888-200505000-00003 Available from: http://dx.doi.org/10.1097/00001888-200505000-00003 15851451 6 Spencer J Learning and teaching in the clinical environment BMJ 2003 326 7389 591 594 12637408 7 Kirch W Schafii C Misdiagnosis at a University Hospital in 4 Medical Eras Report on 400 Cases Medicine (Baltimore) 1996 75 1 29 40 10.1097/00005792-199601000-00004 Available from: http://dx.doi.org/10.1097/00005792-199601000-00004 8569468 8 Nair BR Coughlan JL Hensley MJ Student and patient perspectives on bedside teaching Med Educ 1997 31 5 341 346 10.1046/j.1365-2923.1997.00673.x Available from: http://dx.doi.org/10.1046/j.1365-2923.1997.00673.x 9488854 9 Williams KN Ramani S Fraser B Orlander JD Improving Bedside Teaching: Findings from a Focus Group Study of Learners Acad Med 2008 83 3 257 264 10.1097/ACM.0b013e3181637f3e Available from: http://dx.doi.org/10.1097/ACM.0b013e3181637f3e 18316871 10 Köhl-Hackert N Krautter M Andreesen S Hoffmann K Herzog W Jünger J Nikendei N Workplace Learning: An analysis of students' expectations of learning on the ward in the Department of Internal Medicine GMS Z Med Ausbild 2014 31 4 Doc43 10.3205/zma000935 Available from: http://dx.doi.org/10.3205/zma000935 25489343 11 Olson LG Hill SR Newby DA Barriers to student access to patients in a group of teaching hospitals Med J Aust 2005 183 9 461 16274346 12 Gonzalo JD Chuang CH Huang S Smith C The return of bedside rounds: an educational intervention J Gen Intern Med 2010 25 8 792 798 10.1007/s11606-010-1344-7 Available from: http://dx.doi.org/10.1007/s11606-010-1344-7 20386997 13 Gonzalo JD Heist BS Duffy BL Dyrbye L Fagan MJ Ferenchick G Harrell H Hemmer PA Kernan WN Kogan JR Rafferty C Wong R Elnicki DM The art of bedside rounds: a multi-center qualitative study of strategies used by experienced bedside teachers J Gen Intern Med 2013 28 3 412 420 10.1007/s11606-012-2259-2 Available from: http://dx.doi.org/10.1007/s11606-012-2259-2 23129164 14 Sonne C Vogelmann R Lesevic H Bott-Flügel L Ott I Seyfarth M Significant improvement of a clinical training course in physical examination after basic structural changes in the teaching content and methods GMS Z Med Ausbild 2013 30 2 Doc21 10.3205/zma000864 Available from: http://dx.doi.org/10.3205/zma000864 23737918 15 Ruiz JG Mintzer MJ Leipzig RM The impact of e-learning in medical education Acad Med 2006 81 3 207 212 10.1097/00001888-200603000-00002 Available from: http://dx.doi.org/10.1097/00001888-200603000-00002 16501260 16 Hörnlein A Mandel A Ifland M Lüneberg E Deckert J Puppe F Akzeptanz medizinischer Trainingsfälle als Ergänzung zu Vorlesungen GMS Z Med Ausbild 2011 28 3 Doc42 10.3205/zma000754 Available from: http://dx.doi.org/10.3205/zma000754 21866244 17 Salajegheh A Jahangiri A Dolan-Evans E Pakneshan S A combination of traditional learning and e-learning can be more effective on radiological interpretation skills in medical students: a pre-and post-intervention study BMC Med Educ 2016 16 1 46 10.1186/s12909-016-0569-5 Available from: http://dx.doi.org/10.1186/s12909-016-0569-5 26842495 18 Subramanian A Timberlake M MIttakanti H Lara M Brandt ML Novel educational approach for medical students: improved retention rates using interactive medical software compared with traditional lecture-based format J Surg Educ 2012 69 2 253 256 10.1016/j.jsurg.2011.12.007 Available from: http://dx.doi.org/10.1016/j.jsurg.2011.12.007 22365876 19 da Costa Vieira RA Lopes AH Sarri AJ Benedetti ZC de Oliveira CZ Oncology E-Learning for Undergraduate. A Prospective Randomized Controlled Trial J Cancer Educ 2016 10.1007/s13187-015-0979-9 Available from: http://dx.doi.org/10.1007/s13187-015-0979-9 20 Celenza A Rogers I Qualitative evaluation of a formal bedside clinical teaching programme in an emergency department Emerg Med J 2006 23 10 769 773 10.1136/emj.2006.037796 Available from: http://dx.doi.org/10.1136/emj.2006.037796 16988303 21 Schwind CJ Boehler ML Folse R Dunnington G Markwell SJ Development of physical examination skills in a third-year surgical clerkship Am J Surg 2001 181 4 338 340 10.1016/S0002-9610(01)00573-6 Available from: http://dx.doi.org/10.1016/S0002-9610(01)00573-6 11438268 22 Kadmon M Strittmatter-Haubold V Greifeneder R Ehlail F Lammerding-Köppel M The sandwich principle--introduction to learner-centred teaching/learning methods in medicine Z Evid Fortbild Qual Gesundheitswes 2007 102 10 628 633 10.1016/j.zefq.2008.11.018 Available from: http://dx.doi.org/10.1016/j.zefq.2008.11.018 23 Woltering V Herrler A Spitzer K Spreckelsen C Blended learning positively affects students' satisfaction and the role of the tutor in the problem-based learning process: results of a mixed-method evaluation Adv Health Sci Educ Theory Pract 2009 14 5 725 738 10.1007/s10459-009-9154-6 Available from: http://dx.doi.org/10.1007/s10459-009-9154-6 19184497 24 Woolliscroft JO TenHaken J Smith J Calhoun JG Medical students' clinical self-assessments: comparisons with external measures of performance and the students' self-assessments of overall performance and effort Acad Med 1993 68 4 285 294 10.1097/00001888-199304000-00016 Available from: http://dx.doi.org/10.1097/00001888-199304000-00016 8466612 25 Gordon MJ A review of the validity and accuracy of self-assessments in health professions training Acad Med 1991 66 12 762 769 10.1097/00001888-199112000-00012 Available from: http://dx.doi.org/10.1097/00001888-199112000-00012 1750956 26 Nikendei C Schilling T Nawroth P Hensel M Ho AD Schwenger V Zeier M Herzog W Schellberg D Katus HA Dengler T Tremmel W Müller M Jünger J Integriertes Skills-Lab-Konzept für die studentische Ausbildung in der inneren Medizin Dtsch Med Wochenschr 2005 130 18 1133 1138 10.1055/s-2005-866799 Available from: http://dx.doi.org/10.1055/s-2005-866799 15856395 27 Jünger J Schäfer S Roth C Schellberg D Friedman Ben-David M Nikendei C Effects of basic clinical skills training on objective structured clinical examination performance Med Educ 2005 39 10 1015 1020 10.1111/j.1365-2929.2005.02266.x Available from: http://dx.doi.org/10.1111/j.1365-2929.2005.02266.x 16178828

PMC005xxxxxx/PMC5003142.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00106410.3205/zma001064Doc65urn:nbn:de:0183-zma0010641ArticleAssessment formats in dental medicine: An overview Prüfungsformate für die Zahnmedizin: eine Übersicht Gerhard-Szep Susanne *1Güntsch Arndt 2Pospiech Peter 3Söhnel Andreas 4Scheutzel Petra 5Wassmann Torsten 6Zahn Tugba 71 Goethe-Universität, Carolinum Zahnärztliches Universitäts-Institut gGmbH, Poliklinik Zahnerhaltungskunde, Frankfurt am Main, Deutschland2 Marquette University School of Dentistry, Department of Surgical Sciences, Milwaukee, USA und Universitätsklinikum Jena, Zentrum für Zahn-, Mund- und Kieferheilkunde, Jena, Deutschland3 Universität Würzburg, Poliklinik für Zahnärztliche Prothetik, Würzburg, Deutschland4 Universitätsmedizin Greifswald, Poliklinik für Zahnärztliche Prothetik, Alterszahnheilkunde und medizinischer Werkstoffkunde, Greifswald, Deutschland5 Universitätsklinikum Münster, Poliklinik für Prothetische Zahnmedizin & Biomaterialien, Münster, Deutschland6 Universitätsmedizin Göttingen, Poliklinik für Zahnärztliche Prothetik, Göttingen, Deutschland7 Goethe-Universität, Carolinum Zahnärztliches Universitäts-Institut gGmbH, Poliklinik für Zahnärztliche Prothetik, Frankfurt am Main, Deutschland*To whom correspondence should be addressed: Susanne Gerhard-Szep, Goethe-Universität, Carolinum Zahnärztliches Universitäts-Institut gGmbH, Poliklinik Zahnerhaltungskunde, Frankfurt am Main, Deutschland, E-mail: s.szep@em.uni-frankfurt.de15 8 2016 2016 33 4 Clinical skillsDoc6523 10 2015 09 5 2016 24 3 2016 Copyright © 2016 Gerhard-Szep et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001064.shtmlAim: At the annual meeting of German dentists in Frankfurt am Main in 2013, the Working Group for the Advancement of Dental Education (AKWLZ) initiated an interdisciplinary working group to address assessments in dental education. This paper presents an overview of the current work being done by this working group, some of whose members are also actively involved in the German Association for Medical Education's (GMA) working group for dental education. The aim is to present a summary of the current state of research on this topic for all those who participate in the design, administration and evaluation of university-specific assessments in dentistry. Method: Based on systematic literature research, the testing scenarios listed in the National Competency-based Catalogue of Learning Objectives (NKLZ) have been compiled and presented in tables according to assessment value. Results: Different assessment scenarios are described briefly in table form addressing validity (V), reliability (R), acceptance (A), cost (C), feasibility (F), and the influence on teaching and learning (EI) as presented in the current literature. Infoboxes were deliberately chosen to allow readers quick access to the information and to facilitate comparisons between the various assessment formats. Following each description is a list summarizing the uses in dental and medical education. Conclusion: This overview provides a summary of competency-based testing formats. It is meant to have a formative effect on dental and medical schools and provide support for developing workplace-based strategies in dental education for learning, teaching and testing in the future. Zusammenfassung Ziele: Auf Initiative des Arbeitskreises für die Weiterentwicklung der Lehre in der Zahnmedizin (AKWLZ) wurde 2013 auf dem Deutschen Zahnärztetag in Frankfurt am Main eine interdisziplinäre Arbeitsgruppe zum Thema „Prüfungen in der Zahnmedizin“ gegründet. Diese Übersicht stellt die Zusammenfassung der aktuellen Arbeit dieser AG dar, deren Mitglieder zum Teil auch in der Arbeitsgruppe Zahnmedizin der Gesellschaft für Medizinische Ausbildung (GMA) aktiv mitwirken. Ziel der vorliegenden Übersichtsarbeit ist es, allen Interessierten, die sich mit der Planung, Durchführung und Auswertung von fakultätsinternen Prüfungen im Fach Zahnmedizin beschäftigen, den aktuellen Forschungsstand zum Thema darzustellen und zusammenzufassen. Methoden: Basierend auf einer systematischen Literaturrecherche wurde anlehnend am sogenannten Nutzwertindex von Prüfungen eine tabellarische Darstellung der im NKLZ aufgeführten Szenarien realisiert. Ergebnisse: Unterschiedliche Prüfungsszenarien wurden nach einer kurzen Beschreibung bezüglich ihrer Validität / Gültigkeit (V), der Reliabilität / Zuverlässigkeit (R), der Akzeptanz (A), der Kosten (C), der Durchführbarkeit (F) und des Einflusses auf Lernen und Lehren (EI) nach aktuellem Stand der Literatur tabellarisch dargestellt. Die Darstellungsform der Infoboxen wurde bewusst gewählt, um den Interessierten einen schnellen Einstieg in und Vergleich der einzelnen Prüfungsszenarien zu ermöglichen. Am Ende jeder Prüfungsbeschreibung befinden sich eine stichwortartige Sammlung der jeweiligen Anwendung in der Zahn(Medizin) und ein Fazit für die Anwendung. Folgerungen: Die vorliegende Übersichtsarbeit beinhaltet eine Zusammenfassung zum Thema von kompetenzbasierten Prüfungsszenarien. Sie soll einen formativen Effekt auf die (zahn)medizinischen Fakultäten haben und sie darin unterstützen, die für die Zahnmedizin typischen Schwerpunkte des arbeitsplatzbasierten Lernens, Lehrens und Prüfens in der Zukunft weiterzuentwickeln. ==== Body 1. Starting point A concerted connection between teaching and testing (constructive alignment) is crucial to impart dental competencies during university study [1]. Also connected with the definition of competency-based learning objectives are the appropriate testing formats that measure the requisite combination of knowledge, practical skills and ability to engage in professional decision-making for each particular task (see Figure 1 (Fig. 1)). 2. Method A survey of the literature was undertaken between January 17 and December 17, 2014, in the databases of the German National Library (DNB), MEDLINE using the PubMed interface, Excerpta Medica Database (EMBASE), Education Resource Information Centre (ERIC), Cochrane Library, Science Citation Index, and Google Scholar. The search was conducted automatically and supplemented manually. In addition, available dissertations, open-access publications by German Medical Science (GMS), BEME (Best Evidence Medical and Health Professional Education), including German-language conference proceedings, such as the AKWLZ and GMA, were evaluated. The search terms included the German equivalents for “MCQ”; “MEQ”; “multiple choice”; “MC”; “multiple-choice questionnaire”; “SMP”; “structured oral examination”; “SOE”; “key feature”; “OSCE”; “OSPE”; “standardized patient”; “CEX”; “miniCEX”; “entrustable professional activities”; “DOPS”; “portfolio”, “multi-source feedback” in combination with “AND” and “dental”; “medicine”; “education”; “assessment.” In an initial step, literature was selected based on title and abstract in accordance with pre-defined inclusion and exclusion criteria (inclusion criteria: published 1966-2013 in German or English with topical relevance; exclusion criteria: failure to meet inclusion criteria, full text not available in English or German, lack of topical relevance). The selected publications were then evaluated in terms of their relevance to the issue at hand and excluded, as required. The articles were analyzed, and the results were described according to categories based on the value of assessments [2]. These categories cover the parameters of validity (V), reliability (R), acceptance (A), cost (C), feasibility (F) and influence on teaching and learning (EI). The results were then organized according to the evaluation parameters above. These criteria were further developed in 2011 by the working group headed by Norcini, and the parameters of “equivalence” (assessments conducted at different sites) and “catalytic effect” (consequences for the medical school) were added [3]. In this overview, both of these additional parameters were included and discussed in text form. The focus of this analysis has been carried out using the six criteria listed above (V, R, A, C, F, EI) in the form of tables to allow for clarity and comparisons. 3. Results In total, n=223 publications were identified using the search strategy outlined above and drawn upon as the basis for the following analyses. 4. Discussion Structured oral examinations and the multiple-choice questionnaire (MCQ) are suited for testing theoretical knowledge, meaning descriptive knowledge of competency level 1 [4], [5], [6], [7]. The MCQ is a written assessment with several response options (closed questions), of which a single choice or multiple ones (multiple choice, multiple select) can be the right answer. After a brief introduction of the content and question come the response options that include the correct answer(s) and distracters (wrong answers). Multiple-choice exams can be paper-based, combined with computer-assisted grading, or even administered entirely at computer workstations [4]. Use in medical and dental education MCQs are presently used in both medical and dental study programs [6], [8]. The most important preliminary and final examinations include multiple-choice questions: the preliminary exam in natural science (NVP), preliminary dental exam (ZVP), dental exam (ZP), first and second state medical exams (ÄP). Moreover, MCQs are specifically found in all pre-clinical and clinical subjects in both study programs; this type of question represents one of the most traditional, predominating assessment formats [4], [8]. To assess factual knowledge, the MCQ offers a cost-efficient testing format with high reliability and validity if the questions correspond to the quality criteria. With MCQs it is possible to objectively test a large amount of content in a short period of time. However, this type of assessment can lead to superficial learning of facts. Multiple-choice Questionnaire Validity Classified as high [9] Quality criteria for questions must be met to have sufficient validity [10]. A high construct validity can be achieved if questions are subjected to a review process (e.g. via Item Management System [IMS]) [11]. Reliability Classified as high [9], [11] A minimum of 40 high-quality questions are needed to yield α Cronbach’s α of 0.8 [6]. Acceptance Scoring is objective [4]. MCQs are considered fair if what has been taught corresponds with what is tested [12]. The possibility of passing by giving “strategic” responses, guessing, or picking up on cues is viewed critically by teachers [13], [14], [15], [16]. Cost In light of the numbers and frequency of tests, it is an effective assessment format [9], [17]. A broad range of content can be assessed on one test [5]. Proportionally low costs [18] Positive cost-benefit ratio An existing question pool can be kept current at relatively little cost [19]. Feasibility Effort is primarily involved in generating questions; administering and grading tests require much less time and resources. Creating the question pool is associated with not insignificant costs [4]. Online assessment with digital scoring is possible [5]. A question pool shared by multiple universities increases efficiency via synergies (e.g. IMS) [20]. Influence on teaching and learning Can lead to superficial learning [21] Theoretical knowledge is more important than practical skills [4]. Correct responses are already given making passive recognition possible [14]. Structured oral examinations (SOE) are oral assessments which are conducted by an individual examiner or a panel of examiners. Structured Oral Examination Validity Direct dependence on the degree of structuredness [22] Validity increases with planning, design and conditions of testing [23], [24]. Validity is more dependent on the examiners than the method. Reliability Increases with the number of questions, length of assessment and decreases in the face of strongly differentiated scoring [10] Reliability and objectivity increase with several examiners [10], [17]. Absolute verification of reliability is practically impossible [10]. With a Cronbach’s α of 0.65-0.88 [25], [26], [27], SOEs come out ahead of conventional exams (Cronbach’s α of 0.24-0.50) [25], [28], [29]. Acceptance Performance-inhibiting stress, anxiety and other disruptive factors play a larger role compared to MCQs [12]. Acceptance by teachers and students is reduced by: Intensive supervision by examiner Justification of scores Limited information during limited time Questions or objections possible on the part of the student with no written test to refer to; the difference between content and type of response can lead to this [12]. Cost More cost-intensive than MCQ exams [10] Relativizes itself on high-stakes exams: emphasis is on reliability and validity, not on cost-effectiveness. [30], [31] Feasibility More effort is required compared with MCQ, high financial burden resulting from need for staff and rooms / logistics [10]. Influence on teaching and learning Alongside facts, clinical reasoning, professional thinking, self-confidence and self-assurance can be assessed [12], [22]. Since students adapt their behavior to fit a test [4], [5], [18], extensive preparation can be assumed. If an examination is taken before a panel, the examiners consult and agree on their evaluation of the examinee’s performance. Ideally, the final grades are assigned according to a blueprint governing exam content [7]. The SOE is a testing format that enables assessment of competency level 1 (NKLZ) and beyond within the scope of usual interactions in dental care. However, the higher expenses connected with the greater need for time and personnel should be noted, as well as the potential for performance-inhibiting stress in examinees. Use in medical and dental education Oral examinations with different degrees of structuredness are used in dental and medical study programs [8]. The most important preliminary and final assessments (high-stakes exams) in both study programs (NVP, ZVP, ZP, first and second ÄP) include SOEs in various settings. Furthermore, the SOE is represented in all pre-clinical and clinical subjects in both study programs; it represents one of the traditional, predominating assessment formats [4], [32]. Assessments that do not just measure factual knowledge (=descriptive knowledge: knows) [33], [34]), but also capture the ability to apply theoretical knowledge in a specific context to solve a problem or reach a clinical decision (= procedural knowledge: knows how), require a special testing format that is indeed capable of representing this skill. It must be noted that the ability to solve problems or reason is highly specific to context and always depends on the particular context-related factual knowledge [2], [35]. In addition to the SOE, other assessment formats for evaluating procedural knowledge are the written modified essay question (MEQ) and key features exam. These involve case-based, written assessments that evaluate active knowledge recall, problem-solving and higher order cognitive skills while simulating clinical situations in which decisions are made in the course of a physical examination, diagnosis and therapy. A patient’s history is presented in stages, after each of which several questions are responded to in writing or by selecting the best of several possible responses. Previous questions are partially explained in the following sections making it impermissible to flip back and forth between pages. Use in medical and dental education Developed in Great Britain in the 1970s for the membership examination of the Royal College of General Practitioners [36], [37], [38], [39], [40], [41]. Used internationally in the field of medicine, from undergraduate education to post-graduate training [42], [43], [44], [45], [46], [47], [48], [49], [50]. Used in Germany as an undergraduate testing format and as a written exam that replaces the state examination [51], [52] in model study programs (Witten/Herdecke, Cologne, Bochum, etc.). Hardly any examples of use in dental education; potential areas of application include assessing problem-solving skills within POL and independent learning using case-based, problem-based learning [53], practical, case-based testing with virtual patient cases (e.g. in connection with procedures for handling acute toothache in endodontics) [54]. The MEQ represents a reliable instrument to assess context-specific, procedural knowledge in clinical situations if several basic rules are adhered to: 1. inclusion of the largest number of cases possible; 2. quality control of the pre-defined grading criteria for the write-in (WI) format by several evaluators; 3. computer-based short-menu (SM) or long-menu (LM) response format. Through the simulation of decision making in a clinical setting with questions that build off of each other, learning paired with feedback becomes part of the test experience. The MEQ format represents a significant addition to the written tests commonly used at present in dental education, but it is connected with distinctly higher costs than simply running down a list of MCQs to measure purely factual knowledge. Modified Essay Question Validity Higher validity than for the MCQ format through case-based, context-rich question format [48], [55], [56] Contradictory results for correlation (γ) between MEQs and the results of the final exam (NBME) and post-graduate performance in the first year of professional medical practice: γ 0 0.3/0.3–0.26 [57], γ=0.51 [56]. Reliability Reliability (Cronbach’s α)=0.57–0.91 [38] depends on multiple factors [38], [39], [40], [47], [48], [58], [59]: Quality of the predetermined performance scale Response format (open-ended responses poorer than selecting from a given list) Number of cases and questions Number of graders → e.g. increase of Cronbach’s α from 0.7 to 0.8 by increasing the number of questions from 7 to 12 or increasing the number of graders from 1 to 4 [40]. Acceptance Students generally rate the MEQ positively [41], [51] since the MEQ format reflects practice more closely than the MCQ [60]. Teachers/examiners: greater effort involved in creating tests, coordination challenges [51] Cost Drafting and grading an MEQ are very time consuming and requires personnel [36], [41], [51]. Efforts can be minimized in terms of grading by using a computer-based testing format [61]. Feasibility Generating and grading MEQs is distinctly more involved than for MCQs; difficult to design questions that actually measure the ability to solve problems or make clinical decisions and do not simply test factual knowledge [37], [41], [42], [43], [44], [45], [46], [50], [52], [53], [54], [62]. Influence on teaching and learning MEQs simulate clinical reasoning processes enabling feedback and learning during the test [39], [51], [60]. In the key features exam (KFE) a case unfolds in a specific clinical situation about which multiple questions are asked focusing very closely on only those critical actions or decisions (key features) that are central to the key feature problem or those that are often done incorrectly [34], [63]. Key feature cases are developed in eight defined steps [34], [64], [65]]: identification of the domain or context; selection of a clinical situation; identification of the critical elements of the situation (key features [KF] of the problem); selection and description of the clinical scenario (case vignette); drafting of the questions about the key features of the problem (1-3 question per KF); determination of the response format (open-ended text = write-in, selection = short menu or long menu); generation of the evaluation scale; and content validation. Use in medical and dental education The KF assessment format proposed by Bordage und Page was developed to replace the commonly used written assessment of procedural knowledge using patient management problems (PMP) in medical specialty examinations [64], [65]. Transfer to undergraduate education by Hatala & Norman [66], used worldwide since in medical education as a written assessment format to evaluate context-specific procedural knowledge during the study phase and post-graduate education [67], [68]. Recognized testing format in the German-speaking countries in the field of medicine (see the detailed information on the design and implementation of assessments published by the medical schools at the Universities of Bern and Graz [34], [60], [69]. Studies and reports on the use of the KFE as a written assessment at German medical schools, including internal medicine (Universities of Freiburg, Heidelberg, and Munich [70], Universities of Heidelberg, Tübingen [71]), hematology and oncology (University of Düsseldorf [72]), communication skills (University of Witten-Herdecke [73]). Extensive pilot project in veterinary medicine at the school of veterinary medicine at the University of Hanover [74]. Only a few reports of KF problems used as a written assessment format in dental education [75], [76]. The key feature exam is a valid and reliable instrument for assessing context-specific, procedural knowledge in connection with solving a clinical problem and represents a meaningful addition to the written testing formats currently used in dental education. KFEs can also be used in independent learning with virtual patient cases. For practical reasons, the computer-based format with the long menu response format is preferable to the paper-based version. It is also easier to hinder examinees from returning to previous pages or turning the pages out of order. To increase reliability, it is better to use many short KF cases (at least 15) with a maximum of three questions each than to use fewer, more in-depth cases with four or more questions. Key Features Exam Validity High content validity (92-94%) when graded by teachers/examiners [63], [65], [67]. Piloting and regular review of the key features by students, teachers/examiners is a pre-requisite for high content validity [34], [63], [65]. When a LM format is intended, a WI format is recommended for the pilot to improve the quality of the LMs (supplement missing answers and distracters) [34]. Correlation between KFE scores and other assessment scores (e.g. MCQ) is only moderate (γ=0.35-0.54, [66], [70] which can be explained by the reference to different competency levels. Reliability Reliability of the KF format is higher than for the PMP format [65]. Due to greater case specificity [48], reliability is directly dependent on the number of KF problems (KFP=cases) → number of cases should be as high as possible; number of questions on each case should not exceed three items, since four or more reduces reliability [77]. The selected response format appears to influence reliability, when the same number of KF cases are used: 15 KFPs with 1-4 questions, 2h length, WI format: Cronbach’s α=0.49 [66] 15 KFPs with 3-5 questions, 1.5h length, computer-based LM format: Cronbach’s α=0.65 [70] → α=0.75 is possible with 25 KFs! Acceptance Students: relatively high acceptance [74], [78]: evaluated as realistic and supportive of practical learning. Cost Generation and validation of a KFE involves great amounts of time and staff [67]. Feasibility Generating KFEs is more difficult and requires more time than an MCQ [60], [69]. Necessary testing time depends on the selected response format: LM>WI>SM>MC [79]. The advantages of LM response format (lower cueing effect than MCQ/SM, higher inter-rater reliability than WI) can be realized by using computer-based testing with a moderate testing time [70], [72], [79]. Testing time for 15 KFPs with 3-5 questions is 90 minutes for a computer-based exam [70] and 120 minutes for a paper-based test with a WI response format [66]. Practical examples exist [68], [71], [75], [76], [80]. Influence on teaching and learning KFE format is closer to a real patient situation, promotes the learning of clinically relevant material and practical case-based learning [81][. While study programs in dental medicine do impart advanced theoretical knowledge, they also require students to develop manual skills. Consequently, suitable assessment formats are needed to measure not only factual and procedural knowledge but also to give students an opportunity to demonstrate their practical abilities (shows how, [33]) and to evaluate this objectively. Simply “knows how” is raised a level to “shows how”. When creating such assessments, the learning objectives should be selected in advance and only those which represent a practical competency level should be employed. Standardization of test and examiner allows for an objective assessment of student performance. Suitable assessment formats for this are objective structured clinical examinations (OSCE), objective structured practical examinations (OSPE) and the use of simulated, or standardized, patients (SP). An OSCE is appropriate for evaluating practical skills and the ability to communicate [14]. Students pass through different stations where particular practical skills are demonstrated (including partial treatments) or mock medical consultations are conducted. Evaluations are documented using a checklist created by a group of experts according to how the exam content is weighted. Test time per station is around five minutes; two minutes need to be planned for the examinee to change stations and for the examiner to make final notes or give feedback. Use in medical and dental education Widely used internationally in all clinical subjects since its introduction. Can be used in undergraduate and post-graduate programs [82], [83], [84], [85]. There are many examples of use in dental disciplines: pre-clinical phase [86], [87], [88], orthodontics [89], [90], oro-maxillofacial surgery [91], [92], [93], restorative dentistry [87], [94], [95], [96], parodontology [97], clinical prosthetics [86], pediatric dentistry [98], radiology [99], microbiology [94], [97]. Interdisciplinary OSCE [87], [94], [100], [101]. Integration of an OSCE in the preliminary dental exam [102]. Also used in dental education to evaluate communication skills [103], [104], problem-solving skills, and critical thinking [105]. If possible, feedback should be included as part of the exam. The OCSE is a reliable and valid testing format to assess individual competencies; it enjoys a high level of acceptance by students and teachers. Objective Structured Clinical Examination Validity Predictive validity Significant correlation between OSCE and performance on practical tests and scores on preliminary practical medical exams p<0.01 [87] No correlation between OSCE and MCQ [105] High content and construct validities [4], [106] High face validity [107] Acceptable predictive validity [108] Caution required if students have a language problem or suffer from high levels of stress [109] Attention must be paid to the blueprint [106], [110] Determine content areas early [110] Define questions within the content areas [110] Reliability Cronbach’s α between 0.11-0.97 [4] High reliability among OSCEs, with fewer than n=10 stations it being approximately 0.56, with more than n=10 stations, 0.74 [111] Varying recommendations on station number: at least 19 [4] 14-18 for 5-10 minutes each [106], [112] Stations with an SP should be assessed for at least 15 minutes [110] The more examiners, the higher the values. [111], [113] Method of evaluation critical: high values for global assessments, combinations of global assessments and checklists are good, only checklists alone are least suitable Post-OSCE tests increase the reliability [110]. Acceptance Students: high acceptance, appropriate testing format for functional skills [96] Teachers/examiners: high acceptance [112], [114], [115], [116] Cost 181 Euros/examinee 86-130 Euros/examinee [4], [117] 2.5h/examinee [118] $15-200/examinee [117], [119], [120] Higher costs when including SPs [121] $21-200/examinee [122] Feasibility Testing format demands great amounts of time and resources [106], [119] Thorough preparation needed: Establishing shared structures helps on interdisciplinary OSCEs [100]. Evaluation by external examiners is recommended. Ensure the quality of SPs Station content should be selected to match the OSCE scenario. Peer reviews pre- and post-OSCE (psychometric analysis with difficulty, discrimination, etc. is recommended) Take the extent of the examiner’s experience, field of expertise, sex, and level of fatigue into consideration [3], [106], [112]. Practical examples exist [95], [102]. Influence on teaching and learning Positive influence on learning [106], [108], [123] Stimulates learning [112] Learning at the stations has little to do with the reality of patients [112]. Allot time for feedback [110] Due to the extensive preparation involved before and after its administration, interdisciplinary cooperation is recommended to minimize this disadvantage. OSCEs can be substituted for previously used assessment formats or supplement them in meaningful ways. A sufficient number of stations (n>10), a blueprint, peer review of station content and the scoring criteria, as well as a balance among the modes of evaluation (global, checklist, combination), training the examiners and, if needed, conducting a pilot OSCE should be taken into account when designing an OSCE. A special type of OSCE is embodied in the objective structured practical examination (OSPE) during which practical skills, knowledge and/or interpretation of data are demonstrated in a non-clinical situation [124]. These assessments can be conducted in labs or simulated stations in SimLab. In contrast to the OSCE, an entire process can be evaluated through to the end result (for instance, a dental filling). It is possible to confidently assess practical skills and/or the interpretation of clinical data with the OSPE. This format involves a reliable and valid assessment method to evaluate individual competencies; The OSPE enjoys a high level of acceptance by students and teachers. Objective Structured Practical Examination Validity High validity, γ>7 High construct validity Reliability High reliability among the stations, Cronbach’s α=0.8 [125] Inter-rater reliability ICC>0.7 High inter-rater reliability with equivalent levels of experience and knowledge among examiners, γ=0.79-0.93; p<0.001 Acceptance Students: high acceptance [126], [127] felt to be a “fair test” [128] preferred over traditional exam formats [126] Teachers: relevant, fair, objective and reliable testing format Cost No information available Feasibility Requires extensive planning and teamwork [128] Influence on teaching and learning Individual competencies can be assessed, the need to demonstrate factual and procedural knowledge influences learning behavior [128]. Makes strengths and weaknesses in practical skills discernible [129] Stimulates learning [129] Positive learning experience [130] Defined grading criteria for each step within a process are necessary. Use in medical and dental education OSPEs are administered around the world in medicine, including pharmacology [128], physiology, forensic medicine [130], and dentistry [131], [132]. In Germany they are primarily used in the pre-clinical phase of dental education [133]. Simulated, or standardized, patients in dental education are specially trained (lay) actors who are capable of acting out common clinical pictures or typical occasions for dental consultations. They are used for both practicing and assessing doctor-patient consultations and examination techniques; the use of an SP also provides opportunities to learn how to conduct physical examinations and acquire better communication skills. It is also possible to incorporate SPs into assessments, most frequently in OSCE scenarios. Standardized patients can be used to assess doctor-patient interactions and examination techniques. They are especially suited for evaluation of clinical competencies and communication skills within the scope of an OSCE. When implementing this, the complexity of the case should be tailored to match the testing scenario. Standardized Patients Validity Assesses clinical competencies [134] Reliability Consistent examination (No significant differences between exam cohorts and time points) [135] Acceptance Use of standardized patients (SP) within the scope of an OSCE station [136] Cost 10-18 Euro/examinee [136] Feasibility Case complexity can be controlled and adjusted to reflect educational level [137] Faculty members can determine relevant learning objectives and coordinate role creation. Greater need for time and staff to select and train SPs and to monitor for quality [137] Checklists to record all SP observations of the doctor-patient consultation [138] Practical examples exist [139]. Influences on teaching and learning Improves students’ clinical skills [140] Use in medical and dental education This method has been used in clinical education since the 1960s [138]. Patient contact can be simulated under standardized conditions [139]. SPs can also provide feedback and critique the examinee’s abilities [139]. The term “workplace-based assessment” (WBA) encompasses a wide variety of testing scenarios meant to assess practical skills associated with treating patients in complex situations. The clinical evaluation exercise (CEX) involves a workplace-based assessment in the clinical setting that stretches over a longer period of time (several hours to days) and covering treatment processes during which an examinee conducts a consultation with a single patient recording a patient health history and carrying out a physical examination. A maximum of two assessors should participate, but do not generally have to be present the entire time. Often the data is collected from the patient without the assessor being present. This assessment format, also known as the tCEX (traditional CEX), represents a single event measure. Use in medical and dental education Originally developed in the 1960s as an assessment in internal medicine by the American Board of Internal Medicine (ABIM), it replaced the oral examination as the standard method in 1972 [141], [142]. Replaced by the mini-CEX around 1995 [143], [144]. No documented examples of use in dental education are found in the literature. This assessment format is an instrument of low validity and poor reliability for testing practical skills in complex situations. It is possible to improve the assessment by using the greatest number of patients possible (cases), the greatest number of assessors possible, and the most structured evaluation instruments possible. In addition, providing feedback as part of this testing format should be mandatory. Overall, it can be asserted that in dental education the CEX is a reasonable assessment format for measuring practical competencies in complex situations only if the previously mentioned attempts at improvement have been made. Clinical Evaluation Exercise Validity Insufficient content validity; does not completely cover curricular learning objectives [145] Simulated situation, does not correspond with the reality of medical practice since it is too long and detailed [144] Reliability Questionable reliability since only few exercises can be done due to the great amount of time needed [146] Low inter-rater reliability [147] Cronbach’s α is 0.24 for one case and even for two cases only 0.39 [141]. Acceptance Low level of acceptance since it is very dependent on the assessor [148] Cost Less costly than the OSCE because real patients are used who do not need to be trained [145] Feasibility Relatively simple since no special preparation is necessary [141] Practical examples exist [142], [143]. Influence on teaching and learning Patient-oriented, real-life situations [141] The mini-clinical evaluation exercise (mCEX) is a patient-centered assessment format in the clinical setting that, in contrast to the CEX, requires a shorter amount of time and always includes feedback (approximately 15 minutes of assessment and 10 minutes of feedback). This testing format can be described as having three phases: observation, documentation and feedback. Over the course of the assessment, several assessors observe the examinee and evaluate what they see according to pre-defined criteria. Medical care is given to more than one patient under normal circumstances with a focus on communication and clinical examination [144]. Evaluations are generally formulated according to defined criteria valid for each examinee. These criteria can consist of a rating scale and/or short written comments. The difficulty remains in terms of the different patients undergoing physical examination. Viewed according to Miller’s pyramid, a high level of practical skill is attained. Strictly speaking, it is a structured clinical observation. Mini-Clinical Evaluation Exercise Validity Higher validity than CEX [149] Acceptable validity and reliability have been demonstrated [146], [150]. Able to validly differentiate between competency levels (first year, second year, etc.) [151] Reliability Low inter-rater reliability [149] A minimum of 10 evaluations are necessary to yield reliable results; a larger number is better [151] At least 12-14 evaluations are recommended per year if there are different assessors to increase inter-rater reliability [152]. Reliability of G=0.4 for 10 evaluations; G=0.8 for 50 evaluations [151] Dependent on number of assessors: if there is one examiner, a minimum of eight observations of different patients are necessary for a reliability of 0.8, in the case of two, four are necessary, and for three examiners, three observations [153]. Nine items are better than five to cover differences in competencies [154]. Acceptance High level of satisfaction for students and teachers [151], [155], [156] Implementation is at present slow, since it involves something new [156]. Partially problematic due to discrepancies between self-assessment and assessment by another [157]. Cost Substantial expense as a consequence of the amount of time needed [158], [159]. Feasibility Observations of authentic doctor-patient interactions by different educators in different situations; feedback on different clinical pictures at different locations each with a different focus [155] Thorough planning is necessary because giving feedback takes 8-17 minutes [155], [160]. Relatively simple to implement with enough flexibility in the dental setting [161] Practical examples exist [162]. Influence on teaching and learning Improvement in competency through regular feedback from experts [163] Examiner/examinee receive feedback or a clear impression of clinical work making targeted mentoring possible [156]. Giving constructive feedback must be learned and practiced; teaching skills are needed [164]. No new discoveries or knowledge in comparison with traditional evaluation procedures [158] No influence in comparison with control groups [153] Learning objectives must reflect teaching content [165]. Predictive validity between OSCE and mCEX cannot be demonstrated [165]. This assessment format is frequently referred to as the mCEX (mini-CEX) and represents a single event measure. Use in medical and dental education Developed in 1995 by Norcini [144]; replaced the tCEX in the 1990s. Reliability depends heavily on the number of assessors and cases [151], [153]. Several documented instances in the literature of use in dental medicine (Dental Foundation Training in Great Britain), however, often without any precise information on the evaluation instruments [161], [162]. The mCEX is a valid and reliable instrument to assess practical skills in complex situations. Options for improvement include 1. increasing the number of response items (nine are better than five) or increasing the number of observations (a minimum of 10 observations are needed) and 2. offering train-the-teacher programs (for instance in the form of video demonstrations and role playing). Longitudinal use is recommended with implementation conceivable in a wide variety of different settings (including high-stakes exams). The mCEX format is a good testing format for use in dental education to measure practical competencies in dental medicine. Entrustable professional activities close the gap between the theory of competency-based education and patient-centered practice in a clinical context [166]. This method first became known for its use in the area of post-graduate education; since 2013 it has also appeared in undergraduate medical education [167], [168]. The integration of theoretical and practical knowledge to solve complex problems is assessed (e.g. anamnesis, clinical examination of a patient in connection with different reasons for seeking medical advice) using existing competency-based roles, such as those defined by CanMeds or ACGME. During the assessment it is determined whether the examinee is able to perform the activity while receiving directions, under supervision, with occasional assistance, or independently [169], [170]. As a result, different performance levels can be identified [171]. It is not individual learning objectives that are assessed, but rather an overall activity centering on a patient [172]. In order to differentiate EPAs from general learning objectives, it is recommended that following sentence be completed: One day, the doctor/dentist will be expected to do (insert particular activity) without direct supervision [166]. According to its definition, an EPA should include activities that are important to daily practice, very often are subject to error when being performed, and integrate multiple competencies [172], [173]. Consequently, an EPA consists of diverse roles, each role, in turn, of multiple learning objectives, and each learning objective of different performance levels. The assessment can be a direct or indirect observation and include feedback. It is crucial that the observed performance of the examinee is combined with the performance evaluation over a defined period of time. Entrustable Professional Activities Validity High face validity [174] Reliability Low inter-rater reliability [175] Acceptance Potential for wide acceptance [166] Helps those learning to develop their own study schedule [176] Helps the entire faculty to maintain transparency in education [176] Costs No information available Feasibility Initially requires intensive, well thought-out preparation while EPAs are being designed [177] 20-30 EPAs are recommended for a degree program [177] Practical examples exist [178], [179] Influence on teaching and learning EPAs require numerous competencies in an integrated, holistic manner [177]. Methods of evaluation that focus on the required degree of supervision [180] Feedback is vital [174]. Support from the faculty is necessary [175]. Enables a broad (panoramic) view of the educational program [174]. A commonly reported combination is that of the mCEX with MSF (Multi-source feedback). Strictly speaking, this involves a multiple event measure. Use in medical and dental education Introduced in the Netherlands by ten Cate in 2005; since then it has been used in the fields of surgery, family medicine, internal medicine, neurology, emergency medicine, pediatrics, urology, and is used widely by the Royal Australian and New Zealand College of Psychiatrists [178], [179]. Initially in the pilot phase in German medical education [165]. No documented instances of use in dental medicine EPAs are a relatively new, little researched instrument for assessing practical skills in complex situations. The implementation of EPAs requires extensive and well thought-out preparation when determining the focus. To the extent possible, a maximum of 30 interdisciplinary EPAs per curricular unit should be defined drawing upon input from university instructors and practicing physicians or dentists. EPAs create a realistic link between competency-based learning objectives and higher level activities. Train-the-teacher programs (with practice giving feedback) should improve implementation. Longitudinal use is recommended. Implementation is conceivable in a wide variety of settings, including high-stakes exams. The EPA format represents an innovative approach with great future potential in terms of assessing practical skills in complex situations in dental education. Similar to the mCEX, Directly Observed Procedural Skills (DOPS) entail a short workplace-based assessment in a clinical setting that includes feedback (approximately 15 minutes of assessment and 10 minutes of feedback). This also involves a three-phase assessment in which observation, documentation and feedback occur. Treatment given to (multiple) patients under conditions typical to a medical practice, as with the mCEX, but with a focus on manual skills and interventions observed by several assessors and evaluated according to defined criteria. This assessment format also represents a single event measure. Use in medical and dental education Originally introduced in the United Kingdom by the General Medical Council in 2002 [144]. Use reported in the fields of general medicine, surgery, and internal medicine [181]. International reports of use in dentistry in Iran (universities of Shiraz and Mashad) and at Kings College in London [182], [183]. DOPS is a valid and reliable instrument to evaluate practical skills in complex situations. It is possible to improve this format by having three assessors intervene during two observations, conducting at least two observations, and by holding train-the-teacher sessions. Overall, longitudinal use is recommended. Implementation is conceivable in diverse settings, including high-stakes exams. The DOPS format is a very reasonable testing format to capture practical skills in complex situations during dental education. Directly Observed Procedural Skills Validity High face validity [181] Formative assessment tool [182] Significantly different from MCQ; provides different assessments of student performance [182] Separate assessment tool that does not enable an overall evaluation; a system with different possibilities is needed [184]. DOPS efficiently evaluates practical skills [182]. Reliability To achieve a high reliability, at least three assessors should observe a student during two different case scenarios [181]. G=0.81 [185] Internal consistency is 0.94 and inter-rater reliability is 0.81 Students do not view it as suitable for improving inter-rater reliability [186]. Substantial differences between the assessors can influence the validity of the results if there has not been strict standardization [187]. Good reliability and consensus among assessors is possible [188]. Fewer assessors are needed in comparison with the mCEX [160]. Fewer assessors and cases are needed in comparison with the mCEX [181]. Higher item correlation values than for the mCEX: 0.7-0.8 versus 0.5-0.8 [150], [189] Reliability depends on the case [181]. Reliability independent of process [160] Acceptance High acceptance by students [186] Examinees find the scenarios to be stressful, but appreciate the feedback [190]. Cost Substantial expense is to be expected [159], [191]. Feasibility Great amount of time needed [163], [182] Great amount of time needed for preparing DOPS, including giving feedback [160][ To increase the learning effect, it is necessary to give feedback directly after the assessment and to address strengths and weaknesses [192]. Assessors must be trained in advance [12]. It is feasible to use only one assessor [193]. Influence on teaching and learning Examinees perceive a positive influence on independence and the learning process [186][. DOPS assessment improves practical clinical skills [192]. Positive effect through directly observing the learner [192] Promotes an in-depth approach to learning in the clinical context [21] Positive influence on student reflections [181] Seventy percent of those observed believe that DOPS is helpful for improving practical skills [194]. Compared to control groups there are significantly better results for DOPS regarding practical skills [195]. Can also be used in peer arrangements in the pre-clinical and clinical context [183] The Portfolio as an assessment tool is a pre-defined, objectives-centered collection of student learning activities with assigned self-reflection exercises, as well as feedback [20]. Portfolio contents are developed in alignment with the learning process; the following aspects can be taken into consideration: personal experiences (what was done, seen, written, created?), learning process (awareness that what has been experienced is relevant to future medical or dental practice), documentation (certificates, etc.), future goals regarding learning (looking ahead), and learning environments [196]. Portfolios are a multiple event measure. Use in medical and dental education Portfolio-based learning was introduced in 1993 by the Royal College of General Practitioners, Portfolio assessing described by Shulman in 1998 [197], [198]. Publications in the fields of general medicine, otorhinolaryngology, internal medicine, pediatrics, public health at universities in Maastricht (NL), Nottingham (GB), and Arkansas (USA) [196]. Found in German medical education in Cologne [196]. International reports of use in dentistry [199], [200], [201]. The portfolio entails a highly valid and reliable instrument for evaluating practical skills in complex situations, one that assesses collected, cumulative information about performance and development. Possibilities for optimization exist when more than one neutral grader is used, the student’s mentor is not one of these graders, and train-the-teacher sessions on giving feedback are held. Longitudinal use is recommended. Implementation is conceivable in diverse setting, including high-stakes exams. The portfolio format represents a valuable assessment format to evaluate practical skills in complex situations in dental education. Portfolio Validity Good validity if there is an appropriate selection of all required competency areas [202], [203]. Reliability Cronbach’s α is 0.8 with four graders [204] Cronbach’s α is 0.8 with 15 portfolio entries and two graders [202]. Use of a clear, competency-based master plan, clear grading criteria, inclusion of guidelines and experienced graders for development and evaluation [202], [203] Uniform and consistent grading is difficult [200]. Acceptance Portfolios are viewed as time consuming, a source of anxiety and not very effective [205]. The acceptance of portfolios decreases the longer students spend time on them [205]. Cost No information available Feasibility A portfolio typically includes seven case reports, two presentations, three self-reflections [202]. Typical content includes diagnoses and treatment plans [202]. Problematic since there is a conflict when portfolios are used for both assessment and learning [205]. Difficulties being self-critical and honest [205] Conducting interviews with students about portfolio content improved feasibility [206] Practical examples exist [199], [201]. Influence on teaching and learning Allows the assessment of competencies that could not otherwise be measured [200] Portfolio content must be aligned with the learning objectives [202]. Increases self-knowledge and encourages critical thinking [205] Improves the ability to learn independently and connects theory with practice [205] Time consuming for grader and student [200], [207] Students receive constructive feedback [207]. Calibration and validation are critically important [200]. Provides cumulative information on performance and progress [205] When it is known that the portfolio will be graded, students attempt to fulfill expectations which, in turn, affects the portfolio’s content and educational value [205]. Positive effects are heavily dependent on the support, direction, time commitment and feedback given by the teacher [205]. Multi-source feedback, also known as 360-degree feedback (MSF, multi-rater feedback), involves a workplace-based assessment in a clinical setting incorporating different groups of people associated with that particular work setting and the examinee (peers, dentists, nursing staff, patients, administrators, etc.). The focus of the observations is on professional conduct and teamwork, as well as the examinee taking responsibility as the person in charge [208], [209]. These aspects are observed by several assessors and evaluated according to defined criteria. The “supervisor” is given a special role in this testing scenario: this person collects all the results and gives them to the examinee. As a result, the individuals who have given feedback remain anonymous. The student receives a comprehensive picture based on all the input from different sources. High acceptance is achieved through selection of the assessors. Narrative comments and metric rating scales can be combined. This format entails a multiple event measure. Use in medical and dental education Used in medicine since 1970, widespread in North America (Canada and USA), Europe (England, Holland), and Asia [210], [211]. Reports of use in the fields of general medicine, internal medicine, surgery, gynecology, psychiatry, pathology, and radiology, etc. [210]. Used in dental medicine by the Royal College of Surgeons of England, University of Bristol, UK Committee of Postgraduate Dental Deans. Validated instruments exist for evaluation (PAR: Physicians Achievement Review, SPRAT: Sheffield Peer Assessment Tool). This method consists of a highly valid and reliable instrument for evaluating practical skills in complex situations. Multisource Evaluations Validity Can make it easier to evaluate inter-personal and communicative skills in particular [212] Good validity [213] Reliability Review: to reach a value of 0.9 minimum for Cronbach’s α, eight medical assessors, eight non-medical assessors and 25 patients must participate [210] High internal consistency (=0.8) with five assessors on two observed occasions [214] To reach a value of 0.8 for Cronbach’s α, a minimum of 11 assessors must participate [215]. Value for Cronbach’s α is 0.98 [216]. Problematic due to the number of assessors required [217] Acceptance Rated 4.5 by examinees on a scale of 1-7 [214] Rated 5.3 by assessors on a scale of 1-7 [214] Evaluations are possibly too positive since anonymization is not fully trusted [217] Cost Expense needs to be taken into account before implementation [159]. Feasibility Rated 4.4 by examinees on a scale of 1-7 [214] Rated 5.1 by assessors on a scale of 1-7 [214] Evaluations are generally verified via questionnaires making the process simple [159]. To achieve a valid assessment, a certain number of evaluations are necessary; however, not all are possible to do [217]. Ideally, feedback is gathered over a longer period of time [217]. Can be easily implemented, even in a busy hospital [211], [218] Influence on teaching and learning General improvement in clinical work, communication with co-workers and patients [219] Rated 4.2 by examinees on a scale of 1-7 [214] Rated 4.4 by assessors on a scale of 1-7 [214] Improvement of the evaluation process, advantage of receiving more detailed information and being exposed to different perspectives [217] Varying results: improvement in communication and conduct after receiving 360° feedback [220]. Immensely time consuming and no improvement in assessment as a consequence of the feedback [221] It is possible to identify weak performers at an early stage [218]. Feedback from SPs for students also possible [222]. Belonging to the success factors are a clear definition of the objectives and the sources of feedback. An important role is played by the selection of the assessors, credibility of the assessors and their familiarity with the situation under evaluation, along with the anonymity of the individuals supplying the feedback. This format can be optimized by using approximately five assessors for two observed situations and holding train-the-teacher sessions concerning constructive feedback. The combination of external feedback with self-evaluation by the examinee can be helpful, as can be jointly determining specific learning objectives for the future, including the discussion and documentation of concrete learning opportunities and supports. Longitudinal use is recommended. Implementation is also conceivable in diverse setting, including high-stakes exams. The MSF format represents a valuable assessment format for evaluating practical skills in complex situations in dental education. 5. Conclusion The range of assessment methods presented in this overview significantly broadens the spectrum of already established university-specific exams—mostly MCQs and (structured) oral exams. Each of the methods outlined here meets different requirements and thus covers different competency levels. This must be taken into particular consideration by those who are involved in designing, administering and evaluating assessments in dental medicine. When developing and implementing a curriculum, not only the choice of assessment format is critical but also noting the general functions of an exam, which in turn has an effect on the curriculum [223]: assessments can be summative or formative. Summative assessments usually come at the end of a semester or after a skill has been taught in order to evaluate learning outcomes. Formative assessments are reflective of the learning process itself and do not determine whether a student passes or fails a course or is ultimately successful in displaying the mastery of a particular competency. Such an assessment shows students their current level of proficiency and is supposed to support the learning process through reflection by students on their weaknesses. Purely formative assessments are few in the face of limited staffing resources and time constraints, but are an ideal tool for fostering the learning process. Within the scope of drafting the NKLZ it became clear that in the future other assessment formats will be needed in addition to the established methods such as oral examinations and MC exams; these new formats will need to measure required practical skills in dental medicine, not just in the Skills Lab, but also in patient treatment. Each assessment format should correspond with the targeted competency levels. The presentation of the assessment formats in this overview enables quick orientation within each method and makes reference to relevant literature for those who wish to know more. Including even more detailed information on each of the assessment formats would have compromised the intended character of this article as an overview. Along with theoretical knowledge of an assessment format, it is important to engage in direct exchange with colleagues in higher education who are already following a particular method. For this reason, it is desirable, and perhaps the task of the relevant working groups, to establish a network of professionals who have already gathered experience with special assessment formats and who are willing to make themselves available to those with questions. Depending upon demand, continuing education programs could emerge from such a network providing substantial assistance in implementing new assessment formats. 6. Outlook With the new licensing regulations for dentists (Approbationsordnung), German dental education will be brought up to date and more closely linked to medical education. The assessment methods mentioned as examples in the NKLZ and outlined in this paper demonstrate the various options for assessing at the competency level. After experience has been gathered with university examinations in dental education and following scientific analysis of these testing methods, additional appropriate assessment methods should be included in the licensing requirements for dentistry. These should also be used to improve the quality of the state examinations. Together with the introduction of the NKLZ, compiling experience in organizing, preparing, administering, conducting and evaluating the assessment formats profiled here will be an important task in the coming years, whereby dental medicine can make good use of the competencies under development for medical students since 2002. Dental medicine can also bring to bear its own experience and expertise in the assessment of practical skills. Our shared goal should be to continue developing assessment formats for the different competency levels in dental and medical education in cooperation with the German medical schools. Acknowledgements The authors wish to extend their gratitude to all those who have helped to write, edit and finalize this article. Special thanks to the executive board of AKWLZ, especially Prof. P. Hahn, MME (University of Freiburg) and Prof. H.-J Wenz, MME (University of Kiel), for the detailed feedback and suggestions for improvement. Compting interests The authors declare that they have no competing interests. Authors Authors are listed in alphabetical order. Figure 1 Examples of assessment scenarios depending on competency level according to the requirements in Miller and the National Competency-based Catalogue of Learning Objectives for Undergraduate Dental Education (NKLZ). ==== Refs 1 Biggs J Enhancing teaching through constructive alignment High Educ 1996 32 3 347–364 10.1007/BF00138871 Available from: http://dx.doi.org/10.1007/BF00138871 2 van der Vleuten CP Verwijnen GM Wijnen W Fifteen years of experience with progress testing in a problem-based learning curriculum Med Teach 1996 18 2 103–109 10.3109/01421599609034142 Available from: http://dx.doi.org/10.3109/01421599609034142 3 Norcini J Anderson B Bollela V Burch V Costa MJ Duvivier R Criteria for good assessment: consensus statement and recommendations from the Ottawa 2010 Conference Med Teach 2011 33 3 206–214 10.3109/0142159X.2011.551559 Available from: http://dx.doi.org/10.3109/0142159X.2011.551559 21345060 4 Chenot JF Ehrhardt M Objective structured clinical examination (OSCE) in der medizinischen Ausbildung: Eine Alternative zur Klausur Z Allg Med 2003 79 1 7 5 Examination and Assessments: Academic Integrity [Internet] cited 2015 Jan 17 London Imperial College London Available from: https://workspace.imperial.ac.uk/registry/Public/Procedures%20and%20Regulations/Policies%20and%20Procedures/Examination%20and%20Assessment%20Academic%20Integrity.pdf 6 Jünger J Just I Empfehlungen der Gesellschaft für Medizinische Ausbildung und des Medizinischen Fakultätentags für fakultätsinterne Leistungsnachweise während des Studiums der Human-, Zahn-und Tiermedizin GMS Z Med Ausbild 2014 31 3 Doc34 10.3205/zma000926 Available from: http://dx.doi.org/10.3205/zma000926 25228936 7 Nationaler Kompetenzbasierter Lernzielkatalog Zahnmedizin (NKLZ) [Internet] cited 2016 June14 Available from: http://www.nklz.de/files/nklz_katalog_20150706.pdf 8 Möltner A Schultz JH Briem S Böker T Schellberg D Jünger J Grundlegende testtheoretische Auswertungen medizinischer Prüfungsaufgaben und ihre Verwendung bei der Aufgabenrevision GMS Z Med Ausbild 2005 22 4 Doc138 Available from: http://www.egms.de/static/de/journals/zma/2005-22/zma000138.shtml 9 Norcini JJ Swanson DB Grosso LJ Webster GD Reliability, validity and efficiency of multiple choice question and patient management problem item formats in assessment of clinical competence Med Educ 1985 19 3 238–247 10.1111/j.1365-2923.1985.tb01314.x Available from: http://dx.doi.org/10.1111/j.1365-2923.1985.tb01314.x 4010571 10 Roloff S Mündliche Prüfungen [Internet] cited 2016 June 14 Available from: http://www.hochschuldidaktik.net/documents_public/20121127-Roloff-MuendlPruef.pdf 11 Considine J Botti M Thomas S Design, format, validity and reliability of multiple choice questions for use in nursing research and education Collegian 2005 12 1 19–24 10.1016/S1322-7696(08)60478-3 Available from: http://dx.doi.org/10.1016/S1322-7696(08)60478-3 16619900 12 Memon MA Joughin GR Memon B Oral assessment and postgraduate medical examinations: establishing conditions for validity, reliability and fairness Adv Health Sci Educ 2010 15 2 277–289 10.1007/s10459-008-9111-9 Available from: http://dx.doi.org/10.1007/s10459-008-9111-9 13 Harden RM Lever R Wilson GM Two systems of marking objective examination questions Lancet 1969 293 7584 40–42 10.1016/S0140-6736(69)90999-4 Available from: http://dx.doi.org/10.1016/S0140-6736(69)90999-4 4178778 14 Harden RM Stevenson M Downie WW Wilson GM Assessment of clinical competence using objective structured examination BMJ 1975 1 5955 447–451 10.1136/bmj.1.5955.447 Available from: http://dx.doi.org/10.1136/bmj.1.5955.447 1115966 15 Lennox B Marking multiple-choice examinations Br J Med Educ 1967 1 3 203–211 10.1111/j.1365-2923.1967.tb01698.x Available from: http://dx.doi.org/10.1111/j.1365-2923.1967.tb01698.x 6080736 16 McCarthy WH An assessment of the influence of cueing items in objective examinations J Med Ed 1966 41 3 263–266 10.1097/00001888-196603000-00010 Available from: http://dx.doi.org/10.1097/00001888-196603000-00010 17 Hart IR Competence OCOAC Harden RM Centre RCOPASOCRSMEAR, médecins et chirurgiens du Canada des CR Further Developments in Assessing Clinical Competence 1987 Boston Can-Heal Publications 18 Van der Vleuten CP Schuwirth LW Scheele F Driessen EW Hodges B The assessment of professional competence: building blocks for theory development Best Pract Res Clin Obstet Gynaecol 2010 24 6 703–719 10.1016/j.bpobgyn.2010.04.001 Available from: http://dx.doi.org/10.1016/j.bpobgyn.2010.04.001 20510653 19 Schoonheim-Klein ME Habets LL Aartman IH van der Vleuten CP Hoogstraten J van der Velden U Implementing an Objective Structured Clinical Examination (OSCE) in dental education: effects on students' learning strategies Eur J Dent Educ 2006 10 4 226–235 10.1111/j.1600-0579.2006.00421.x Available from: http://dx.doi.org/10.1111/j.1600-0579.2006.00421.x 17038015 20 Fischer MR Holzer M Jünger J Prüfungen an den medizinischen Fakultäten - Qualität, Verantwortung und Perspektiven GMS Z Med Ausbild 2010 27 5 Doc66 10.3205/zma000703 Available from: http://dx.doi.org/10.3205/zma000703 21818211 21 Cobb KA Brown G Jaarsma DADC Hammond RA The educational impact of assessment: a comparison of DOPS and MCQs Med Teach 2013 35 11 e1598–1607 23808609 22 Elmer A Grifka J Vergleich von Prüfungsmethoden in der klinischen Ausbildung Gesundheitswesen (Suppl Med Ausbild) 1998 15 Suppl1 14 17 Available from: https://gesellschaft-medizinische-ausbildung.org/files/ZMA-Archiv/1998/1/Elmer_A,_Grifka_J.pdf 23 Sadaf S Khan S Ali SK Tips for developing a valid and reliable bank of multiple choice questions (MCQs) Educ Health 2012 25 3 195–197 10.4103/1357-6283.109786 Available from: http://dx.doi.org/10.4103/1357-6283.109786 24 Wenzel A Kirkevang L Students'attitudes to digital radiography and measurement accuracy of two digital systems in connection with root canal treatment Eur J Dent Educ 2004 8 4 167 171 10.1111/j.1600-0579.2004.00347.x Available from: http://dx.doi.org/10.1111/j.1600-0579.2004.00347.x 15469443 25 Yang JC Laube DW Improvement of reliability of an oral examination by a structured evaluation instrument J Med Educ 1983 58 11 864–872 6631925 26 Hottinger U Krebs R Hofer R Feller S Bloch R Strukturierte mündliche Prüfung für die ärztliche Schlussprüfung–Entwicklung und Erprobung im Rahmen eines Pilotprojekts 2004 Bern Universität Bern 27 Wass V Wakeford R Neighbour R van der Vleuten C Royal College of General Practitioners Achieving acceptable reliability in oral examinations: an analysis of the Royal College of General Practitioners membership examination's oral component Med Educ 2003 37 2 126–131 10.1046/j.1365-2923.2003.01417.x Available from: http://dx.doi.org/10.1046/j.1365-2923.2003.01417.x 12558883 28 Schubert A Tetzlaff JE Tan M Ryckman JV Mascha E Consistency, inter-rater reliability, and validity of 441 consecutive mock oral examinations in anesthesiology: implications for use as a tool for assessment of residents Anesthesiology 1999 91 1 288–298 10.1097/00000542-199907000-00037 Available from: http://dx.doi.org/10.1097/00000542-199907000-00037 10422954 29 Kearney RA Puchalski SA Yang HYH Skakun EN The inter-rater and intra-rater reliability of a new Canadian oral examination format in anesthesia is fair to good Can J Anaesth 2002 49 3 232–236 10.1007/BF03020520 Available from: http://dx.doi.org/10.1007/BF03020520 11861339 30 Postgraduate Medical Education and Training Board Developing and Maintaining an Assessment System 2007 London General Medical Council 1 31 van der Vleuten CP Assessment of the Future [Internet] cited 2016 June 14 Available from: http://www.youtube.com/watch?v=bvFbmTRVjlE 32 Möltner A Schellberg D Briem S Böker T Schultz JH Jünger J Wo Cronbachs alpha nicht mehr reicht GMS Z Med Ausbild 2005 22 4 Doc137 Available from: http://www.egms.de/de/journals/zma/2005-22/zma000137.shtml 33 Miller GE The assessment of clinical skills/competence/performance Acad Med 1990 65 9 Suppl S63–67 10.1097/00001888-199009000-00045 Available from: http://dx.doi.org/10.1097/00001888-199009000-00045 2400509 34 Kopp V Möltner A Fischer MR Key-Feature-Probleme zum Prüfen von prozeduralem Wissen: Ein Praxisleitfaden GMS Z Med Ausbild 2006 23 3 Doc50 Available from: http://www.egms.de/static/de/journals/zma/2006-23/zma000269.shtml 35 Wass V van der Vleuten C Shatzer J Jones R Assessment of clinical competence Lancet 2001 357 9260 945–949 10.1016/S0140-6736(00)04221-5 Available from: http://dx.doi.org/10.1016/S0140-6736(00)04221-5 11289364 36 Knox J What is.… a Modified Essay Question? Med Teach 1989 11 1 51–57 10.3109/01421598909146276 Available from: http://dx.doi.org/10.3109/01421598909146276 2747485 37 Knox JD Bouchier IA Communication skills teaching, learning and assessment Med Educ 1985 19 4 285–289 10.1111/j.1365-2923.1985.tb01322.x Available from: http://dx.doi.org/10.1111/j.1365-2923.1985.tb01322.x 4021854 38 Feletti GI Reliability and validity studies on modified essay questions J Med Educ 1980 55 11 933–941 10.1097/00001888-198011000-00006 Available from: http://dx.doi.org/10.1097/00001888-198011000-00006 7441674 39 Rabinowitz HK Hojat M A comparison of the modified essay question and multiple choice question formats: their relationship to clinical performance Fam Med 1989 21 5 364–367 2792608 40 Lockie C McAleer S Mulholland H Neighbour R Tombleson P Modified essay question (MEQ) paper: perestroika Occas Pap R Coll Gen Pract 1990 46 18–22 1670201 41 Feletti GI Smith EK Modified essay questions: are they worth the effort? Med Educ 1986 20 2 126–132 10.1111/j.1365-2923.1986.tb01059.x Available from: http://dx.doi.org/10.1111/j.1365-2923.1986.tb01059.x 3959927 42 van Bruggen L Manrique-van Woudenbergh M Spierenburg E Vos J Preferred question types for computer-based assessment of clinical reasoning: a literature study Perspect Med Educ 2012 1 4 162–171 10.1007/s40037-012-0024-1 Available from: http://dx.doi.org/10.1007/s40037-012-0024-1 23205341 43 Irwin WG Bamber JH The cognitive structure of the modified essay question Med Educ 1982 16 6:326–331.DOI 10.1111/j.1365 2923.1982.tb00945 10.1111/j.1365-2923.1982.tb00945.x Available from: http://dx.doi.org/10.1111/j.1365-2923.1982.tb00945.x 44 Weinman J A modified essay question evaluation of pre-clinical teaching of communication skills Med Educ 1984 18 3 164–167 10.1111/j.1365-2923.1984.tb00998.x Available from: http://dx.doi.org/10.1111/j.1365-2923.1984.tb00998.x 6717309 45 Khan MU Aljarallah BM Evaluation of Modified Essay Questions (MEQ) and Multiple Choice Questions (MCQ) as a tool for Assessing the Cognitive Skills of Undergraduate Medical Students Int J Health Sci 2011 5 1 39–43 46 Bodkha P Effectiveness of MCQ, SAQ and MEQ in assessing cognitive domain among high and low achievers IJRRMS 2012 2 4 25–28 47 Wallerstedt S Erickson G Wallerstedt SM Short Answer Questions or Modified Essay questions–More Than a Technical Issue Int J Clin Med 2012 3 28 10.4236/ijcm.2012.31005 Available from: http://dx.doi.org/10.4236/ijcm.2012.31005 48 Elstein AS Beyond multiple-choice questions and essays: the need for a new way to assess clinical competence Acad Med 1993 68 4 244–249 10.1097/00001888-199304000-00002 Available from: http://dx.doi.org/10.1097/00001888-199304000-00002 8466598 49 Ferguson KJ Beyond multiple-choice questions: Using case-based learning patient questions to assess clinical reasoning Med Educ 2006 40 11 1143 10.1111/j.1365-2929.2006.02592.x Available from: http://dx.doi.org/10.1111/j.1365-2929.2006.02592.x 17054648 50 Palmer EJ Devitt PG Assessment of higher order cognitive skills in undergraduate education: modified essay or multiple choice questions? Research paper BMC Med Educ 2007 7 1 49 10.1186/1472-6920-7-49 Available from: http://dx.doi.org/10.1186/1472-6920-7-49 18045500 51 Wild D Rützler M Haarhaus M Peters K Der Modified Essay Question (MEQ)-Test an der medizinischen Fakultät der Universität Witten/Herdecke Gesundheitswesen (Suppl Med Ausbild) 1998 15 Suppl2 65 69 Available from: https://gesellschaft-medizinische-ausbildung.org/files/ZMA-Archiv/1998/2/Wild_D,_R%C3%BCtzler_M,_Haarhaus_M,_Peters_K.pdf 52 Peters K Scheible CM Rützler M MEQ – angemessen und praktikabel? Jahrestagung der Gesellschaft für Medizinische Ausbildung - GMA 2006 Jahrestagung der Gesellschaft für Medizinische Ausbildung - GMA 10.-12.11.2006 Köln Düsseldorf, Köln German Medical Science Doc06gma085 Available from: http://www.egms.de/en/meetings/gma2006/06gma085.shtml 53 O'Neill PN Assessment of students in a problem-based learning curriculum J Dent Educ 1998 62 9 640–643 9789486 54 Geerlings G van de Poel AC De gestructureerde open Vraag: Een Mogelijkheit tot Patientensimulatie binnen Hetonderwijs in de Endodontologie The modified essay question: a possibility for patient simulation in endodontic education Ned Tijdschr Tandheelkd 1984 91 7-8 305–308 6592468 55 Van der Vleuten CP Schuwirth LW Assessing professional competence: from methods to programmes Med Educ 2005 39 3 309–317 10.1111/j.1365-2929.2005.02094.x Available from: http://dx.doi.org/10.1111/j.1365-2929.2005.02094.x 15733167 56 Schwartz RW Donnelly MB Sloan DA Young B Knowledge gain in a problem-based surgery clerkship Acad Med 1994 69 2 148–151 10.1097/00001888-199402000-00022 Available from: http://dx.doi.org/10.1097/00001888-199402000-00022 8311887 57 Rabinowitz HK The modified essay question: an evaluation of its use in a family medicine clerkship Med Educ 1987 21 2 114–118 10.1111/j.1365-2923.1987.tb00676.x Available from: http://dx.doi.org/10.1111/j.1365-2923.1987.tb00676.x 3574162 58 Stratford P Pierce-Fenn H Modified essay question Phys Ther 1985 65 7 1075–1079 4011687 59 Norman GR Smith EK Powles AC Rooney PJ Henry NL Dodd PE Factors underlying performance on written tests of knowledge Med Educ 1987 21 4 297–304 10.1111/j.1365-2923.1987.tb00367.x Available from: http://dx.doi.org/10.1111/j.1365-2923.1987.tb00367.x 3626897 60 Bloch R Hofer D Krebs R Schläppi P Weis S Westkämper R Kompetent prüfen. Handbuch zur Planung, Durchführung und Auswertung von Facharztprüfungen. 1999 Wien Institut für Aus-, Weiter-und Fortbildung 61 Lim EC Seet RC Oh VM Chia BL Aw M Quak SH Onk BK Computer-based testing of the modified essay question: the Singapore experience Med Teach 2007 29 9 e261–268 10.1080/01421590701691403 Available from: http://dx.doi.org/10.1080/01421590701691403 18158650 62 Palmer EJ Devitt PG A method for creating interactive content for the iPod, and its potential use as a learning tool: Technical Advances BMC Med Educ 2007 7 1 32 10.1186/1472-6920-7-32 Available from: http://dx.doi.org/10.1186/1472-6920-7-32 17888168 63 Bordage G Brailovsky C Carretier H Page G Content validation of key features on a national examination of clinical decision-making skills Acad Med 1995 70 4 276–281 10.1097/00001888-199504000-00010 Available from: http://dx.doi.org/10.1097/00001888-199504000-00010 7718059 64 Bordage G Page G Hart IR Harden RM An alternative approach to PMPs: The "key features" concept Further developments in assessing clinical competence 1987 Montreal Can-Heal 59 75 65 Page G Bordage G Allen T Developing key-feature problems and examinations to assess clinical decision-making skills Acad Med 1995 70 3 194–201 10.1097/00001888-199503000-00009 Available from: http://dx.doi.org/10.1097/00001888-199503000-00009 7873006 66 Hatala R Norman GR Adapting the Key Features Examination for a clinical clerkship Med Educ 2002 36 2 160–165 10.1046/j.1365-2923.2002.01067.x Available from: http://dx.doi.org/10.1046/j.1365-2923.2002.01067.x 11869444 67 Trudel JL Bordage G Downing SM Reliability and validity of key feature cases for the self-assessment of colon and rectal surgeons Ann Surg 2008 248 2 252–258 10.1097/SLA.0b013e31818233d3 Available from: http://dx.doi.org/10.1097/SLA.0b013e31818233d3 18650635 68 Ali SK Bordage G Validity of key features for a family medicine pilot exam at the College of Physicians and Surgeons Pakistan J Coll Phys Surg Pakistan 1995 5 6 256–260 69 Bernhardt J Griesbacher T Ithaler D Kresse A Öttl K Roller-Wirnsberger R Vogl S Kürzübersicht gängiger Prüfungsformate 2012 Graz Medizinische Universität Graz Available from: http://www.medunigraz.at/fileadmin/lehren/planen-organisieren/pdf/Kurzuebersicht_Pruefungsformate_v7.pdf 70 Fischer MR Kopp V Holzer M Ruderich F Jünger J A modified electronic key feature examination for undergraduate medical students: validation threats and opportunities Med Teach 2005 27 5 450–455 10.1080/01421590500078471 Available from: http://dx.doi.org/10.1080/01421590500078471 16147800 71 Nikendei C Mennin S Weyrich P Kraus B Effects of a supplementary final year curriculum on students' clinical reasoning skills as assessed by key-feature examination Med Teach 2009 31 9 e438 e443 10.1080/01421590902845873 Available from: http://dx.doi.org/10.1080/01421590902845873 19811181 72 Rotthoff T Baehring T Dicken HD Fahron U Richter B Fischer MR Scherbaum WA Comparison between Long-Menu and Open-Ended Questions in computerized medical assessments. A randomized controlled trial BMC Med Educ 2006 6 1 50 10.1186/1472-6920-6-50 Available from: http://dx.doi.org/10.1186/1472-6920-6-50 17032439 73 Zupanic M Iblher P Töpper J Gartmeier M Bauer J Prenzel M Möller G Hoppe-Seyler T Karsten G Fischer MR Key Feature-Assessment kommunikativer Leistungen: Weiterent¬wicklung und quantitative Evaluation Jahrestagung der Gesellschaft für Medizinische Ausbildung (GMA) 2011 Jahrestagung der Gesellschaft für Medizinische Ausbildung (GMA) 05.-08.10.2011 München Düsseldorf German Medical Science GMS Publishing House Doc11gma024 10.3205/11gma024 Available from: http://dx.doi.org/10.3205/11gma024 74 Schaper E Tipold A Ehlers JP Use of key feature questions in summative assessment of veterinary medicine students Ir Vet J 2013 66 1 3 10.1186/2046-0481-66-3 Available from: http://dx.doi.org/10.1186/2046-0481-66-3 23497425 75 TU Dresden Studienordnung für den Studiengang Zahnmedizin vom 08.09.2011 [Internet] cited 2015 Feb 24 Dresden Available from: https://tu-dresden.de/die_tu_dresden/fakultaeten/medizinische_fakultaet/studium/zm/recht-zm/index_html#StudOrgZM 76 Gerhardt-Szep S Hahn P Key feature - Fallerstellung (Master of Medical Education, Modul V) 2008 Heidelberg Universität Heidelberg Available from: http://archiv.ub.uni-heidelberg.de/volltextserver/11537/1/Manuscript_KF.pdf 77 Norman G Bordage G Page G Keane D How specific is case specificity? Med Educ 2006 40 7 618–623 10.1111/j.1365-2929.2006.02511.x Available from: http://dx.doi.org/10.1111/j.1365-2929.2006.02511.x 16836533 78 Huwendiek S Mennin SP Nikendei C Medical education after the Flexner report N Engl J Med 2007 356 1 90 10.1056/NEJMc062922 Available from: http://dx.doi.org/10.1056/NEJMc062922 17202463 79 Schuwirth LW van der Vleuten CP de Kock CA Peperkamp AG Donkers HH Computerized case-based testing: A modern method to assess clinical decision making Med Teach 1996 18 4 294–299 10.3109/01421599609034180 Available from: http://dx.doi.org/10.3109/01421599609034180 80 Huwendiek S Reichert F Brass K Bosse H-M Heid J Möltner A Haag M Leven FJ Hoffmann GF Jünger J Tönshoff B Etablierung von fallbasiertem computerunterstütztem Prüfen mit langen Auswahllisten: Ein geeignetes Instrument zur Prüfung von Anwendungswissen GMS Z Med Ausbild 2007 24 1 Doc51 Available from: http://www.egms.de/static/de/journals/zma/2007-24/zma000345.shtml 81 Huwendiek S Heid J Möltner A Haag M Tönshoff B E-Learning und E-Prüfung mit virtuellen Patienten in der Medizin 2008 Heidelberg Universitätsklinikum Heidelberg Available from: http://www.his-he.de/veranstaltung/dokumentation/Workshop_E-Pruefung/pdf/TOP04.pdf 82 Ananthakrishnan N Microteaching as a vehicle of teacher training--its advantages and disadvantages J Postgrad Med 1993 39 3 142–143 8051644 83 Arnold RC Walmsley AD The use of the OSCE in postgraduate education Eur J Dent Educ 2008 12 3 126–130 10.1111/j.1600-0579.2008.00469.x Available from: http://dx.doi.org/10.1111/j.1600-0579.2008.00469.x 18666892 84 Taguchi N Ogawa T OSCEs in Japanese postgraduate clinical training Hiroshima experience 2000-2009 Eur J Dent Educ 2010 14 4 203–209 10.1111/j.1600-0579.2009.00610.x Available from: http://dx.doi.org/10.1111/j.1600-0579.2009.00610.x 20946247 85 Pugh D Touchie C Wood TJ Humphrey-Murto S Progress testing: is there a role for the OSCE? Med Educ 2014 48 6 623–631 10.1111/medu.12423 Available from: http://dx.doi.org/10.1111/medu.12423 24807438 86 Curtis DA Lind SL Brear S Finzen FC The correlation of student performance in preclinical and clinical prosthodontic assessments J Dent Educ 2007 71 3 365–372 17389571 87 Eberhard L Hassel A Bäumer A Becker F Beck-Mußotter J Bömicke W Corcodel N Cosgarea R Eiffler C Giannakopoulos NN Kraus T, Mahabadi J, Rues S, Schmitter M, Wolff D, Wege KC Analysis of quality and feasibility of an objective structured clinical examination (OSCE) in preclinical dental education Eur J Dent Educ 2011 15 3 172 178 10.1111/j.1600-0579.2010.00653.x Available from: http://dx.doi.org/10.1111/j.1600-0579.2010.00653.x 21762322 88 Graham R Bitzer LA Anderson OR Reliability and Predictive Validity of a Comprehensive Preclinical OSCE in Dental Education J Dent Educ 2013 77 2 161–167 23382525 89 Fields H Rowland M Vig K Huja S Objective structured clinical examination use in advanced orthodontic dental education Am J Orthod Dentofacial Orthop 2007 131 5 656–663 10.1016/j.ajodo.2007.01.013 Available from: http://dx.doi.org/10.1016/j.ajodo.2007.01.013 17482087 90 Derringer KA Undergraduate orthodontic assessment and examination in UK dental schools Br Dent J 2006 201 4 225–229 10.1038/sj.bdj.4813884 Available from: http://dx.doi.org/10.1038/sj.bdj.4813884 16902574 91 Macluskey M Durham J Balmer C Bell A Cowpe J Dawson L Dental student suturing skills: a multicentre trial of a checklist-based assessment Eur J Dent Educ 2011 15 4 244–249 10.1111/j.1600-0579.2010.00665.x Available from: http://dx.doi.org/10.1111/j.1600-0579.2010.00665.x 21985209 92 Hoefer SH Schuebel F Sader R Landes C Development and implementation of an objective structured clinical examination (OSCE) in CMF-surgery for dental students J Craniomaxillofac Surg 2013 41 5 412–416 10.1016/j.jcms.2012.11.007 Available from: http://dx.doi.org/10.1016/j.jcms.2012.11.007 23257318 93 Landes CA Hoefer S Schuebel F Ballon A Teiler A Tran A Weber R Walcher F Sader R Long-term prospective teaching effectivity of practical skills training and a first OSCE in Cranio Maxillofacial Surgery for dental students J Craniomaxillofac Surg 2014 42 5 e97–e104 10.1016/j.jcms.2013.07.004 Available from: http://dx.doi.org/10.1016/j.jcms.2013.07.004 24012014 94 Larsen T Jeppe-Jensen D The introduction and perception of an OSCE with an element of self- and peer-assessment Eur J Dent Educ 20081 12 1 2–7 18257758 95 Kupke J Wicht MJ Stützer H Derman SH Lichtenstein NV Noack MJ Does the use of a visualised decision board by undergraduate students during shared decision-making enhance patients' knowledge and satisfaction? - A randomised controlled trial Eur J Dent Educ 2012 17 1 19 25 10.1111/eje.12002 Available from: http://dx.doi.org/10.1111/eje.12002 23279388 96 Hammad M Oweis Y Taha S Hattar S Madarati A Kadim F Students' Opinions and Attitudes After Performing a Dental OSCE for the First Time: A Jordanian Experience J Dent Educ 2013 77 1 99–104 23314473 97 Mossey PA Newton JP Stirrups DR Scope of the OSCE in the assessment of clinical skills in dentistry Br Dent J 2001 190 6 323–326 10.1038/sj.bdj.4800961 Available from: http://dx.doi.org/10.1038/sj.bdj.4800961 11325158 98 Boone WJ McWhorter AG Seale NS Purposeful assessment techniques (PAT) applied to an OSCE-based measurement of competencies in a pediatric dentistry curriculum J Dent Educ 2001 65 11 1232–1237 11765869 99 Lele SM A Mini-OSCE for Formative Assessment of Diagnostic and Radiographic Skills at a Dental College in India J Dent Educ 2011 75 12 1583–1589 22184597 100 Schoonheim-Klein M Walmsley AD Habets L An implementation strategy for introducing an OSCE into a dental school Eur J Dent Educ 2005 9 4 143 149 10.1111/j.1600-0579.2005.00379.x Available from: http://dx.doi.org/10.1111/j.1600-0579.2005.00379.x 16194245 101 Licari FW Knight GW Developing a group practice comprehensive care education curriculum J Dent Educ 2003 67 12 1312–1315 14733262 102 Ratzmann A Wiesmann U Kordaß B Integration of an Objective Structured Clinical Examination (OSCE) into the dental preliminary exams GMS Z Med Ausbild 2012 29 1 Doc09 10.3205/zma000779 Available from: http://dx.doi.org/10.3205/zma000779 22403594 103 Ogawa T Taguchi N Sasahara H Assessing communication skills for medical interviews in a postgraduate clinical training course at Hiroshima University Dental Hospital Eur J Dent Educ 2003 7 2 60 65 10.1034/j.1600-0579.2002.00273.x Available from: http://dx.doi.org/10.1034/j.1600-0579.2002.00273.x 12823519 104 Cannick GF Horowitz AM Garr DR Reed SG Neville BW Day TA Woolson RF Lackland DT Use of the OSCE to evaluate brief communication skills training for dental students J Dent Educ 2007 71 9 1203–1209 17761627 105 Dennehy PC Susarla SM Karimbux NY Relationship between dental students' performance on standardized multiple-choice examinations and OSCEs J Dent Educ 2008 72 5 585–592 18451082 106 Khan KZ Ramachandran S Gaunt K Pushkar P The Objective Structured Clinical Examination (OSCE): AMEE Guide No. 81. Part I: an historical and theoretical perspective Med Teach 2013 35 9 e1437–1446 10.3109/0142159X.2013.818634 Available from: http://dx.doi.org/10.3109/0142159X.2013.818634 23968323 107 Deis N Narciß E Rahe J Schüttpelz-Braun K Objektive standardisierte praktische Prüfungen zur Messung von praktischen Fertigkeiten und berufsrelevanten Kompetenzen Z Gesundheit Sport 2012 2 2 25–33 108 Beard JD Marriott J Purdie H Crossley J Assessing the surgical skills of trainees in the operating theatre: a prospective observational study of the methodology Health Technol Assess 2011 15 1 i–xxi–1–162 109 Brand HS Schoonheim-Klein M Is the OSCE more stressful? Examination anxiety and its consequences in different assessment methods in dental education Eur J Dent Educ 2009 13 3 147–153 10.1111/j.1600-0579.2008.00554.x Available from: http://dx.doi.org/10.1111/j.1600-0579.2008.00554.x 19630933 110 Nikendei C Jünger J OSCE-praktische Tipps zur Implementierung einer klinisch-praktischen Prüfung GMS Z Med Ausbild 2006 23 3 Doc47 Available from: http://www.egms.de/static/de/journals/zma/2006-23/zma000266.shtml 111 Brannick MT Erol-Korkmaz HT Prewett M A systematic review of the reliability of objective structured clinical examination scores Med Educ 2011 45 12 1181–1189 10.1111/j.1365-2923.2011.04075.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2011.04075.x 21988659 112 Schoonheim-Klein M Muijtjens A Muijtens A Habets L Manogue M van der Vleuten C Hoogstraten J Van der Velden U On the reliability of a dental OSCE, using SEM: effect of different days Eur J Dent Educ 2008 12 3 131–137 10.1111/j.1600-0579.2008.00507.x Available from: http://dx.doi.org/10.1111/j.1600-0579.2008.00507.x 18666893 113 Norcini JJ Maihoff NA Day SC Benson JA Trends in medical knowledge as assessed by the certifying examination in internal medicine JAMA 1989 262 17 2402–2404 10.1001/jama.1989.03430170064029 Available from: http://dx.doi.org/10.1001/jama.1989.03430170064029 2795825 114 Hofer M Jansen M Soboll S Potential improvements in medical education as retrospectively evaluated by candidates for specialist examinations Dtsch med Wochenschr 2006 131 8 373–378 10.1055/s-2006-932527 Available from: http://dx.doi.org/10.1055/s-2006-932527 16479467 115 Fischer MR Gesellschaft für Medizinische Ausbildung Kompetenzzentrum Prüfungen Baden-Württemberg Leitlinie für Fakultäts-interne Leistungsnachweise während des Medizinstudiums: Ein Positionspapier des GMA-Ausschusses Prüfungen und des Kompetenzzentrums Prüfungen Baden-Württemberg GMS Z Med Ausbild 2008 25 1 Doc74 Available from: http://www.egms.de/static/de/journals/zma/2008-25/zma000558.shtml 116 Davenport ES Davis JE Cushing AM Holsgrove GJ An innovation in the assessment of future dentists Br Dent J 1998 184 4 192–195 9549918 117 Rau T Fegert J Liebhardt H How high are the personnel costs for OSCE? A financial report on management aspects GMS Z Med Ausbild 2011 28 1 Doc13 10.3205/zma000725 Available from: http://dx.doi.org/10.3205/zma000725 21818228 118 Kropmans TJ O'Donovan BG Cunningham D Murphy AW Flaherty G Nestel D Dunne FPl An Online Management Information System for Objective Structured Clinical Examinations CIS 2011;28 5 1 38 48 119 Barman A Critiques on the Objective Structured Clinical Examination Ann Acad Med Singap 2005 34 8 478–482 16205824 120 Stillman PL Swanson DB Smee S Stillman AE Ebert TH Emmel VS Gaslowitz J Green HL Hamolsky M Hatem C Assessing clinical skills of residents with standardized patients Ann Intern Med 1986 105 5 762–771 10.7326/0003-4819-105-5-762 Available from: http://dx.doi.org/10.7326/0003-4819-105-5-762 3767153 121 Turner JL Dankoski ME Objective structured clinical exams: a critical review Fam Med 2008 40 8 574–578 18988044 122 Carpenter JL Cost analysis of objective structured clinical examinations Acad Med 1995 70 9 828–833 10.1097/00001888-199509000-00022 Available from: http://dx.doi.org/10.1097/00001888-199509000-00022 7669163 123 Duerson MC Romrell LJ Stevens CB Impacting faculty teaching and student performance: nine years' experience with the Objective Structured Clinical Examination Teach Learn Med 2000 12 4 176–182 10.1207/S15328015TLM1204_3 Available from: http://dx.doi.org/10.1207/S15328015TLM1204_3 11273366 124 Harden RM Cairncross RG Self assessment Med Teach 1980 2 3 145–148 10.3109/01421598009081185 Available from: http://dx.doi.org/10.3109/01421598009081185 24483358 125 Kundu D Das HN Sen G Osta M Mandal T Gautam D Objective structured practical examination in biochemistry: An experience in Medical College, Kolkata J Nat Sci Biol Med 2013 4 1 103–107 10.4103/0976-9668.107268 Available from: http://dx.doi.org/10.4103/0976-9668.107268 23633843 126 Abraham RR Raghavendra R Surekha K Asha K A trial of the objective structured practical examination in physiology at Melaka Manipal Medical College, India Adv Physiol Educ 2009 33 1 21–23 10.1152/advan.90108.2008 Available from: http://dx.doi.org/10.1152/advan.90108.2008 19261756 127 Adome RO Kitutu F Creating an OSCE/OSPE in a resource-limited setting Med Educ 2008 42 5 525–526 10.1111/j.1365-2923.2008.03045.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2008.03045.x 128 Nayak V Bairy KL Adiga S Shenoy S Magazine BC Amberkar M Kumari MK OSPE in Pharmacology: Comparison with the conventional Method and Students' Perspective Towards OSPE Br Biomed Bull 2014 2 1 218 222 129 Wani P Dalvi V Objective Structured Practical Examination vs Traditional Clinical Examination in Human Physiology: Student's perception Int J Med Sci Public Health 2013 2 3 522–547 10.5455/ijmsph.2013.080320133 Available from: http://dx.doi.org/10.5455/ijmsph.2013.080320133 130 Menezes RG Nayak VC Binu VS Kanchan T Rao PP Baral P Lobo SW Objective structured practical examination (OSPE) in Forensic Medicine: students' point of view J Forensic Leg Med 2011 18 8 347–349 10.1016/j.jflm.2011.06.011 Available from: http://dx.doi.org/10.1016/j.jflm.2011.06.011 22018165 131 Huth KC Baumann M Kollmuss M Hickel R Fischer MR Paschos E Assessment of practical tasks in the Phantom course of Conservative Dentistry by pre-defined criteria: a comparison between self-assessment by students and assessment by instructors Eur J Dent Educ 2015 10.1111/eje.12176 Available from: http://dx.doi.org/10.1111/eje.12176 [Epub ahead of print] 10.1111/eje.12176 132 Banerjee R Chandak A Radke U Bringing objectivity to assessment in Preclinical Prosthodontics: The student's perspective on OSPE JETHS 2014 1 2 30 33 133 Schmitt I Möltner A Bärmeier J Gärtner K Dopfer S Kuschel B Kunkel F Heidemann D Gerhardt-Szép S Wie viele Prüfer braucht ein OSCE? 2013 Düsseldorf AKWLZ Available from: https://www.akwlz.de/documents/262773/490615/K8+-+Schmitt+-+Pr%E2%80%99er+OSPE.pdf/dc030606-f303-4e81-b85c-601177f11e0c?version=1.0 134 Adamo G Simulated and standardized patients in OSCEs: achievements and challenges 1992-2003 Med Teach 2003 25 3 262–270 10.1080/0142159031000100300 Available from: http://dx.doi.org/10.1080/0142159031000100300 12881047 135 Colliver JA Barrows HS Vu NV Verhulst SJ Mast TA Travis TA Test security in examinations that use standardized-patient cases at one medical school Acad Med 1991 66 5 279–282 10.1097/00001888-199105000-00011 Available from: http://dx.doi.org/10.1097/00001888-199105000-00011 2025360 136 Ortwein H Fröhmel A Burger W Einsatz von Simulationspatienten als Lehr-, Lern-und Prüfungsform Psychother Psychosom Med Psychol 2006 56 01 23–29 10.1055/s-2005-867058 Available from: http://dx.doi.org/10.1055/s-2005-867058 16421779 137 Collins J Harden RM AMEE Medical Education Guide No. 13: real patients, simulated patients and simulators in clinical examinations Med Teach 1998 20 6 508–521 10.1080/01421599880210 Available from: http://dx.doi.org/10.1080/01421599880210 138 Barrows H An overview of the uses of standardized patients for teaching and evaluating clinical skills Acad Med 1993 68 6 443–451 10.1097/00001888-199306000-00002 Available from: http://dx.doi.org/10.1097/00001888-199306000-00002 8507309 139 Cleland JA Abe K Rethans JJ The use of simulated patients in medical education: AMEE Guide No 42 Med Teach 2009 31 6 477–486 10.1080/01421590903002821 Available from: http://dx.doi.org/10.1080/01421590903002821 19811162 140 Hendrickx K De Winter B Tjalma W Avonts D Peeraer G Wyndaele JJ Learning intimate examinations with simulated patients: The evaluation of medical students' performance Med Teach 2009 31 4 E139–147 10.1080/01421590802516715 Available from: http://dx.doi.org/10.1080/01421590802516715 19404886 141 Norcini JJ The death of the long case? BMJ 2002 324 7334 408–409 10.1136/bmj.324.7334.408 Available from: http://dx.doi.org/10.1136/bmj.324.7334.408 11850375 142 Norcini J The validity of long cases Med Educ 2001 35 8 720–721 10.1046/j.1365-2923.2001.01006.x Available from: http://dx.doi.org/10.1046/j.1365-2923.2001.01006.x 11489096 143 Norcini JJ Blank LL Arnold GK Kimball HR The mini-CEX (clinical evaluation exercise): a preliminary investigation Ann Intern Med 1995 123 10 795–799 10.7326/0003-4819-123-10-199511150-00008 Available from: http://dx.doi.org/10.7326/0003-4819-123-10-199511150-00008 7574198 144 Norcini JJ Blank LL Duffy FD Fortna GS The mini-CEX: a method for assessing clinical skills Ann Intern Med 2003 138 6 476–481 10.7326/0003-4819-138-6-200303180-00012 Available from: http://dx.doi.org/10.7326/0003-4819-138-6-200303180-00012 12639081 145 Thornton S A literature review of the long case and its variants as a method of assessment Educ Med J 2012 4 1 5–11 146 Durning SJ Cation LJ Markert RJ Pangaro LN Assessing the reliability and validity of the mini-clinical evaluation exercise for internal medicine residency training Acad Med 2002 77 9 900–904 10.1097/00001888-200209000-00020 Available from: http://dx.doi.org/10.1097/00001888-200209000-00020 12228088 147 Herbers JE Noel GL Cooper GS Harvey J Pangaro LN Weaver MJ How accurate are faculty evaluations of clinical competence? J Gen Intern Med 1989 4 3 202–208 10.1007/BF02599524 Available from: http://dx.doi.org/10.1007/BF02599524 2723833 148 Yousuf N Mini clinical evaluation exercise: validity and feasibility evidences in literature Educ Med J 2012 4 1 e100 e107 149 Hill F Kendall K Galbraith K Crossley J Implementing the undergraduate mini-CEX: a tailored approach at Southampton University Med Educ 2009 43 4 326–334 10.1111/j.1365-2923.2008.03275.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2008.03275.x 19335574 150 Kogan JR Bellini LM Shea JA Feasibility, reliability, and validity of the mini-clinical evaluation exercise (mCEX) in a medicine core clerkship Acad Med 2003 78 10 Suppl S33–35 10.1097/00001888-200310001-00011 Available from: http://dx.doi.org/10.1097/00001888-200310001-00011 14557089 151 Alves De Lima A Barrero C Baratta S Castillo Costa Y Bortman G Carabajales J Conde D Galli A Degrange G Van der Vleuten C Validity, reliability, feasibility and satisfaction of the Mini-Clinical Evaluation Exercise (Mini-CEX) for cardiology residency training Med Teach 2007 29 8 785–790 10.1080/01421590701352261 Available from: http://dx.doi.org/10.1080/01421590701352261 17917984 152 Norcini JJ The mini clinical evaluation exercise (mini-CEX) Clin Teach 2005 2 1 25 30 10.1111/j.1743-498X.2005.00060.x Available from: http://dx.doi.org/10.1111/j.1743-498X.2005.00060.x 153 Alves de Lima A Assessment of clinical competence: Reliability, Validity, Feasibility and Educational Impact of the mini-CEX 2013 Maastricht Universität Maastricht Available from: http://www.icba.com.ar/profesionales/pdf/aal/Thesis_Alberto_Alves_de_Lima_170x240_v10.pdf 154 Cook DA Beckman TJ Does scale length matter? A comparison of nine- versus five-point rating scales for the mini-CEX Adv Health Sci Educ 2009 14 5 655–664 10.1007/s10459-008-9147-x Available from: http://dx.doi.org/10.1007/s10459-008-9147-x 155 Alves de Lima AE Conde D Aldunate L van der Vleuten CP Teachers' experiences of the role and function of the mini clinical evaluation exercise in post-graduate training Int J Med Educ 2010 1 68–73 10.5116/ijme.4c87.8e13 Available from: http://dx.doi.org/10.5116/ijme.4c87.8e13 156 Berendonk C Beyeler C Westkämper R Giger M Strukturiertes Feedback in der ärztlichen Weiterbildung: Mini-CEX und DOPS Schweiz Ärztez 2008 89 32 1337–1340 157 Eva KW Regehr G Self-assessment in the health professions: a reformulation and research agenda Acad Med 2005 80 10 Suppl S46–54 10.1097/00001888-200510001-00015 Available from: http://dx.doi.org/10.1097/00001888-200510001-00015 16199457 158 Brazil V Ratcliffe L Zhang J Davin L Mini-CEX as a workplace-based assessment tool for interns in an emergency department--does cost outweigh value? Med Teach 2012 34 12 1017–1023 10.3109/0142159X.2012.719653 Available from: http://dx.doi.org/10.3109/0142159X.2012.719653 23039836 159 Magnier KM Dale VH Pead MJ Workplace-based assessment instruments in the health sciences J Vet Med Educ 2012 39 4 389–395 10.3138/jvme.1211-118R Available from: http://dx.doi.org/10.3138/jvme.1211-118R 23187032 160 Wilkinson JR Crossley JG Wragg A Mills P Cowan G Wade W Implementing workplace-based assessment across the medical specialties in the United Kingdom Med Educ 2008 42 4 364–373 10.1111/j.1365-2923.2008.03010.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2008.03010.x 18338989 161 Prescott-Clements L van der Vleuten CP Schuwirth LW Hurst Y Rennie JS Evidence for validity within workplace assessment: the Longitudinal Evaluation of Performance (LEP) Med Educ 2008 42 5 488–495 10.1111/j.1365-2923.2007.02965.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2007.02965.x 18298449 162 Deshpande S Chahande J Impact of computer-based treatment planning software on clinical judgment of dental students for planning prosthodontic rehabilitation Adv Med Educ Pract 2014 5 269–274 10.2147/AMEP.S66264 Available from: http://dx.doi.org/10.2147/AMEP.S66264 25170288 163 Veloski J Boex JR Grasberger MJ Evans A Wolfson DB Systematic review of the literature on assessment, feedback and physicians' clinical performance: BEME Guide No. 7 Med Teach 2006 28 2 117–128 10.1080/01421590600622665 Available from: http://dx.doi.org/10.1080/01421590600622665 16707292 164 Holmboe ES Yepes M Williams F Huot SJ Feedback and the mini clinical evaluation exercise J Gen Intern Med 2004 19 5 Pt 2 558–561 10.1111/j.1525-1497.2004.30134.x Available from: http://dx.doi.org/10.1111/j.1525-1497.2004.30134.x 15109324 165 Montagne S Rogausch A Gemperli A Berendonk C Jucker-Kupper P Beyeler C The mini-clinical evaluation exercise during medical clerkships: are learning needs and learning goals aligned? Med Educ 2014 48 10 1008–1019 10.1111/medu.12513 Available from: http://dx.doi.org/10.1111/medu.12513 25200021 166 Berberat PO Harendza S Kadmon M Gesellschaft für Medizinische Ausbildung GMA-Ausschuss für Weiterbildung Entrustable professional activities - visualization of competencies in postgraduate training. Position paper of the Committee on Postgraduate Medical Training of the German Society for Medical Education (GMA) GMS Z Med Ausbild 2013 30 4 Doc47 10.3205/zma000890 Available from: http://dx.doi.org/10.3205/zma000890 24282450 167 Colleges AOAM Core Entrustable Professional Activities for Entering Residency. [Internet] 2014 cited 2016 June 14 Available from: https://members.aamc.org/eweb/upload/Core%20EPA%20Curriculum%20Dev%20Guide.pdf 168 Ten Cate O Trusting graduates to enter residency: what does it take? J Grad Med Educ 2014 6 1 7–10 10.4300/JGME-D-13-00436.1 Available from: http://dx.doi.org/10.4300/JGME-D-13-00436.1 24701302 169 Ten Cate O Snell L Carraccio C Medical competence: the interplay between individual ability and the health care environment Med Teach 2010 32 8 669–675 10.3109/0142159X.2010.500897 Available from: http://dx.doi.org/10.3109/0142159X.2010.500897 20662579 170 Jones MD Rosenberg AA Gilhooly JT Carraccio CL Perspective: Competencies, Outcomes, and Controversy—Linking Professional Activities to Competencies to Improve Resident Education and Practice Acad Med 2011 86 2 161–165 10.1097/ACM.0b013e31820442e9 Available from: http://dx.doi.org/10.1097/ACM.0b013e31820442e9 21169788 171 Mulder H Ten Cate O Daalder R Berkvens J Building a competency-based workplace curriculum around entrustable professional activities: The case of physician assistant training Med Teach 2010 32 10 e453–459 10.3109/0142159X.2010.513719 Available from: http://dx.doi.org/10.3109/0142159X.2010.513719 20854153 172 Chang A Bowen JL Buranosky RA Frankel RM Ghosh N Rosenblum MJ Thompson S Green ML Transforming primary care training--patient-centered medical home entrustable professional activities for internal medicine residents J Gen Intern Med 2013 28 6 801 809 10.1007/s11606-012-2193-3 Available from: http://dx.doi.org/10.1007/s11606-012-2193-3 22997002 173 Scheele F Teunissen P van Luijk S Heineman E Fluit L Mulder H Meininger A Wijnen-Meijer M Glas G Sluiter H Hummel T Introducing competency-based postgraduate medical education in the Netherlands Med Teach 2008 30 3 248–253 10.1080/01421590801993022 Available from: http://dx.doi.org/10.1080/01421590801993022 18484450 174 Englander R Carraccio C From theory to practice: making entrustable professional activities come to life in the context of milestones Acad Med 2014 89 10 1321–1323 10.1097/ACM.0000000000000324 Available from: http://dx.doi.org/10.1097/ACM.0000000000000324 24892405 175 Aylward M Nixon J Gladding S An entrustable professional activity (EPA) for handoffs as a model for EPA assessment development Acad Med 2014 89 10 1335–1340 10.1097/ACM.0000000000000317 Available from: http://dx.doi.org/10.1097/ACM.0000000000000317 24892402 176 Boyce P Spratt C Davies M McEvoy P Using entrustable professional activities to guide curriculum development in Psychiatry training BMC Med Educ 2011 11 1 96 10.1186/1472-6920-11-96 Available from: http://dx.doi.org/10.1186/1472-6920-11-96 22112295 177 Ten Cate O Competency-based education, entrustable professional activities, and the power of language J Grad Med Educ 2013 5 1 6–7 10.4300/JGME-D-12-00381.1 Available from: http://dx.doi.org/10.4300/JGME-D-12-00381.1 24404218 178 Hauer KE Kohlwes J Cornett P Hollander H Ten Cate O Ranji SR Soni K Iobst W O'Sullivan P Identifying entrustable professional activities in internal medicine training J Grad Med Educ 2013 5 1 54–59 10.4300/JGME-D-12-00060.1 Available from: http://dx.doi.org/10.4300/JGME-D-12-00060.1 24404227 179 Shaughnessy AF Sparks J Cohen-Osher M Goodell KH Sawin GL Gravel J Jr Entrustable Professional Activities in Family Medicine J Grad Med Educ 2013 5 1 112–118 10.4300/JGME-D-12-00034.1 Available from: http://dx.doi.org/10.4300/JGME-D-12-00034.1 24404237 180 Ten Cate O Scheele F Competency-based postgraduate training: can we bridge the gap between theory and clinical practice? Acad Med 2007 82 6 542–547 10.1097/ACM.0b013e31805559c7 Available from: http://dx.doi.org/10.1097/ACM.0b013e31805559c7 17525536 181 Naeem N Validity, reliability, feasibility, acceptability and educational impact of direct observation of procedural skills (DOPS) J Coll Physicians Surg Pak 2013 23 1 77–82 23286629 182 Bazrafkan L Comparison of the Assessment of Dental Students' J Med Ed 2009 13 1, 2 3 8 183 Tricio J Woolford M Thomas M Lewis-Greene H Georghiou L Andiappan M Escudier M Dental students' peer assessment: a prospective pilot study Eur J Dent Educ 2014 19 3 140 148 10.1111/eje.12114 Available from: http://dx.doi.org/10.1111/eje.12114 25168409 184 Abraham RR Upadhya S Torke S Ramnarayan K Student perspectives of assessment by TEMM model in physiology Adv Physiol Educ 2005 29 2 94–97 10.1152/advan.00051.2004 Available from: http://dx.doi.org/10.1152/advan.00051.2004 15905152 185 Barton JR Corbett S van der Vleuten CP English Bowel Cancer Screening Programme UK Joint Advisory Group for Gastrointestinal Endoscopy The validity and reliability of a Direct Observation of Procedural Skills assessment tool: assessing colonoscopic skills of senior endoscopists Gastrointest Endosc 2012 75 3 591–597 10.1016/j.gie.2011.09.053 Available from: http://dx.doi.org/10.1016/j.gie.2011.09.053 22227035 186 Akbari M Shamsabadi RM Direct Observation of Procedural Skills (DOPS) in Restorative Dentistry: Advantages and Disadvantages in Student's Point of View Iran J Med Educ 2013 13 3 212 220 187 Andersen RM Davidson PL Atchison KA Hewlett E Freed JR Friedman JA Pipeline, profession, and practice program: evaluating change in dental education J Dent Educ 2005 69 2 239–248 15689608 188 Hamdy H Prasad K Williams R Salih FA Reliability and validity of the direct observation clinical encounter examination (DOCEE) Med Educ 2003 37 3 205–212 10.1046/j.1365-2923.2003.01438.x Available from: http://dx.doi.org/10.1046/j.1365-2923.2003.01438.x 12603759 189 Torre DM Simpson DE Elnicki DM Sebastian JL Holmboe ES Feasibility, reliability and user satisfaction with a PDA-based mini-CEX to evaluate the clinical skills of third-year medical students Teach Learn Med 2007 19 3 271–277 10.1080/10401330701366622 Available from: http://dx.doi.org/10.1080/10401330701366622 17594223 190 Cohen SN Farrant PBJ Taibjee SM Assessing the assessments: U.K. dermatology trainees' views of the workplace assessment tools Br J Dermatol 2009 161 1 34–39 10.1111/j.1365-2133.2009.09097.x Available from: http://dx.doi.org/10.1111/j.1365-2133.2009.09097.x 19416267 191 Center of Innovation in Professional Health Education and Research (CIPHER) Review of work-based assessment methods 2007 Syndey University of Sydney 192 Roghieh N Fateme H Hamid S Hamid H The effect of formative evaluation using "direct observation of procedural skills" (DOPS) method on the extent of learning practical skills among nursing students in the ICU Iran J Nurs Midwifery Res 2013 18 4 290–293 24403924 193 Hamilton KES Coates V Kelly B Boore JRP Cundell JH Gracey J Performance assessment in health care providers: a critical review of evidence and current practice J Nurs Manag 2007 15 8 773–791 10.1111/j.1365-2934.2007.00780.x Available from: http://dx.doi.org/10.1111/j.1365-2934.2007.00780.x 17944603 194 Morris A Hewitt J Roberts CM Practical experience of using directly observed procedures, mini clinical evaluation examinations, and peer observation in pre-registration house officer (FY1) trainees Postgrad Med J 2006 82 966 285–288 10.1136/pgmj.2005.040477 Available from: http://dx.doi.org/10.1136/pgmj.2005.040477 16597818 195 Sh S Pooladi A BahramRezaie M Farhadifar F Khatibi R Evaluation of the Effects of Direct Observation of Procedural Skills (DOPS) on clinical externship students' learning level in obstetrics ward of kurdistan university of medical sciences J Med Ed 2009 13 1 29 33 196 Stosch C Wichelhaus AS Matthes J Die Portfolio-Methode: Modernes Assessment auf dem Prüfstand GMS Z Med Ausbild 2006 23 3 Doc43 Available from: http://www.egms.de/static/de/journals/zma/2006-23/zma000262.shtml 197 Buckley S Coleman J Davison I Khan KS Zamora J Malick S Moreley D Pollard D Ashcroft T Popovic C Sayers J The educational effects of portfolios on undergraduate student learning: a Best Evidence Medical Education (BEME) systematic review. BEME Guide No. 11 Med Teach 2009 31 4 282–298 10.1080/01421590902889897 Available from: http://dx.doi.org/10.1080/01421590902889897 19404891 198 Tochel C Haig A Hesketh A Cadzow A Beggs K Colthart I Peacock H The effectiveness of portfolios for post-graduate assessment and education: BEME Guide No 12 Med Teach 2009 31 4 299–318 10.1080/01421590902883056 Available from: http://dx.doi.org/10.1080/01421590902883056 19404890 199 Pocock I A new route for dental graduates Dent Update 2007 34 1 59 17348561 200 Kramer GA Albino JEN Andrieu SC Hendricson WD Henson L Horn BD Neumann LM Young SK Dental student assessment toolbox J Dent Educ 2009 73 1 12–35 19126764 201 Gadbury-Amyot CC McCracken MS Woldt JL Brennan RL Validity and reliability of portfolio assessment of student competence in two dental school populations: a four-year study J Dent Educ 2014 78 5 657–667 24789826 202 Michels NR Driessen EW Muijtjens AM Van Gaal LF Bossaert LL de Winter BY Portfolio assessment during medical internships: How to obtain a reliable and feasible assessment procedure? Educ Health 2009 22 3 313 203 O'sullivan PS Reckase MD McClain T Savidge MA Clardy JA Demonstration of portfolios to assess competency of residents Adv Health Sci Educ 2004 9 4 309–323 10.1007/s10459-004-0885-0 Available from: http://dx.doi.org/10.1007/s10459-004-0885-0 204 Melville C Rees M Brookfield D Anderson J Portfolios for assessment of paediatric specialist registrars Med Educ 2004 38 10 1117–1125 10.1111/j.1365-2929.2004.01961.x Available from: http://dx.doi.org/10.1111/j.1365-2929.2004.01961.x 15461658 205 McMullan M Students' perceptions on the use of portfolios in pre-registration nursing education: a questionnaire survey Int J Nurs Stud 2006 43 3 333–343 10.1016/j.ijnurstu.2005.05.005 Available from: http://dx.doi.org/10.1016/j.ijnurstu.2005.05.005 16038911 206 Burch VC Seggie JL Use of a structured interview to assess portfolio-based learning Med Educ 2008 42 9 894–900 10.1111/j.1365-2923.2008.03128.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2008.03128.x 18715487 207 Kadagad P Kotrashetti SM Portfolio: a comprehensive method of assessment for postgraduates in oral and maxillofacial surgery J Maxillofac Oral Surg 2013 12 1 80–84 10.1007/s12663-012-0381-7 Available from: http://dx.doi.org/10.1007/s12663-012-0381-7 24431818 208 Brett JF Atwater LE 360 degree feedback: accuracy, reactions, and perceptions of usefulness J Appl Psychol 2001 86 5 930–942 10.1037/0021-9010.86.5.930 Available from: http://dx.doi.org/10.1037/0021-9010.86.5.930 11596809 209 Lepsinger R Lucia AD The Art and Science of 360 Degree Feedback 2009 New York John Wiley & Sons 210 Donnon T Al Ansari A Al Alawi S Violato C The reliability, validity, and feasibility of multisource feedback physician assessment: a systematic review Acad Med 2014 89 3 511–516 10.1097/ACM.0000000000000147 Available from: http://dx.doi.org/10.1097/ACM.0000000000000147 24448051 211 Zhao Y Zhang X Chang Q Sun B Psychometric characteristics of the 360° feedback scales in professionalism and interpersonal and communication skills assessment of surgery residents in China J Surg Educ 2013 70 5 628–635 10.1016/j.jsurg.2013.04.004 Available from: http://dx.doi.org/10.1016/j.jsurg.2013.04.004 24016374 212 Joshi R Ling FW Jaeger J Assessment of a 360-degree instrument to evaluate residents' competency in interpersonal and communication skills Acad Med 2004 79 5 458–463 10.1097/00001888-200405000-00017 Available from: http://dx.doi.org/10.1097/00001888-200405000-00017 15107286 213 Archer JC Norcini J Davies HA Use of SPRAT for peer review of paediatricians in training BMJ 2005 330 7502 1251–1253 10.1136/bmj.38447.610451.8F Available from: http://dx.doi.org/10.1136/bmj.38447.610451.8F 15883137 214 Murphy DJ Bruce DA Mercer SW Eva KW The reliability of workplace-based assessment in postgraduate medical education and training: a national evaluation in general practice in the United Kingdom Adv Health Sci Educ 2009 14 2 219–232 10.1007/s10459-008-9104-8 Available from: http://dx.doi.org/10.1007/s10459-008-9104-8 215 Wenrich MD Carline JD Giles LM Ramsey PG Ratings of the performances of practicing internists by hospital-based registered nurses Acad Med 1993 68 9 680–687 10.1097/00001888-199309000-00014 Available from: http://dx.doi.org/10.1097/00001888-199309000-00014 8397633 216 Violato C Lockyer JM Fidler H Assessment of psychiatrists in practice through multisource feedback Can J Psychiatry 2008 53 8 525–533 18801214 217 Chandler N Henderson G Park B Byerley J Brown WD Steiner MJ Use of a 360-degree evaluation in the outpatient setting: the usefulness of nurse, faculty, patient/family, and resident self-evaluation J Grad Med Educ 2010 2 3 430–434 10.4300/JGME-D-10-00013.1 Available from: http://dx.doi.org/10.4300/JGME-D-10-00013.1 21976094 218 Hesketh EA Anderson F Bagnall GM Driver CP Johnston DA Marshall D Needham G Orr G Walker K Using a 360 degrees diagnostic screening tool to provide an evidence trail of junior doctor performance throughout their first postgraduate year Med Teach 2005 27 3 219–233 10.1080/01421590500098776 Available from: http://dx.doi.org/10.1080/01421590500098776 16011945 219 Ferguson J Wakeling J Bowie P Factors influencing the effectiveness of multisource feedback in improving the professional practice of medical doctors: a systematic review BMC Med Educ 2014 14 1 76 10.1186/1472-6920-14-76 Available from: http://dx.doi.org/10.1186/1472-6920-14-76 24725268 220 Brinkman WB Geraghty SR Lanphear BP Khoury JC Gonzalez del Rey JA Dewitt TG Britto MT Effect of multisource feedback on resident communication skills and professionalism: a randomized controlled trial Arch Pediatr Adolesc Med 2007 161 1 44–49 10.1001/archpedi.161.1.44 Available from: http://dx.doi.org/10.1001/archpedi.161.1.44 17199066 221 Weigelt JA Brasel KJ Bragg D Simpson D The 360-degree evaluation: increased work with little return? Current Surgery 2004 61 6 616–626 10.1016/j.cursur.2004.06.024 Available from: http://dx.doi.org/10.1016/j.cursur.2004.06.024 15590037 222 Garry A Stirling K Achieving 360° student feedback using SPaCE Clin Teach 2012 9 4 222–227 10.1111/j.1743-498X.2012.00550.x Available from: http://dx.doi.org/10.1111/j.1743-498X.2012.00550.x 22783853 223 Fabry G Medizindidaktik: ein Handbuch für die Praxis 2008 Karlsruhe Huber

PMC005xxxxxx/PMC5003143.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00106310.3205/zma001063Doc64urn:nbn:de:0183-zma0010637ArticleBasic practical skills teaching and learning in undergraduate medical education – a review on methodological evidence Lehren und Lernen praktischer Fertigkeiten im Medizinstudium – ein Review zur methodischen Evidenz Vogel Daniela 1Harendza Sigrid *11 Universitätsklinikum Hamburg-Eppendorf, III. Medizinische Klinik, Hamburg, Deutschland*To whom correspondence should be addressed: Sigrid Harendza, Universitätsklinikum Hamburg-Eppendorf, III. Medizinische Klinik, Martinistraße 52, D-20246 Hamburg, Deutschland, Phone: +49 (0)40/7410-5390, Fax: +49 (0)40/7410-40218, E-mail: harendza@uke.de15 8 2016 2016 33 4 Clinical skillsDoc6430 10 2015 09 5 2016 04 1 2016 Copyright © 2016 Vogel et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001063.shtmlObjective: Practical skills are an essential part of physicians’ daily routine. Nevertheless, medical graduates’ performance of basic skills is often below the expected level. This review aims to identify and summarize teaching approaches of basic practical skills in undergraduate medical education which provide evidence with respect to effective students’ learning of these skills. Methods: Basic practical skills were defined as basic physical examination skills, routine skills which get better with practice, and skills which are also performed by nurses. We searched PubMed with different terms describing these basic practical skills. In total, 3467 identified publications were screened and 205 articles were eventually reviewed for eligibility. Results: 43 studies that included at least one basic practical skill, a comparison of two groups of undergraduate medical students and effects on students’ performance were analyzed. Seven basic practical skills and 15 different teaching methods could be identified. The most consistent results with respect to effective teaching and acquisition of basic practical skills were found for structured skills training, feedback, and self-directed learning. Simulation was effective with specific teaching methods and in several studies no differences in teaching effects were detected between expert or peer instructors. Multimedia instruction, when used in the right setting, also showed beneficial effects for basic practical skills learning. Conclusion: A combination of voluntary or obligatory self-study with multimedia applications like video clips in combination with a structured program including the possibility for individual exercise with personal feedback by peers or teachers might provide a good learning opportunity for basic practical skills. Zusammenfassung Zielsetzung: Praktische Fertigkeiten sind ein wesentlicher Bestandteil des ärztlichen Arbeitsalltags. Dennoch liegt die Leistung von Absolventen eines Medizinstudiums bei der Durchführung praktischer Basisfertigkeiten häufig unter den erwarteten Anforderungen. Diese Übersichtsarbeit verfolgt daher das Ziel, Lehrmethoden für medizinische Basisfertigkeiten im Medizinstudium zu identifizieren und zusammenzufassen, die einen evidenzbasierten Nachweis für das effektive studentische Lernen dieser Fertigkeiten erbringen. Methoden: Praktische Basisfertigkeiten wurden als Basisfertigkeiten der körperlichen Untersuchung, Routinefertigkeiten, die im Laufe der Praxis besser werden, und als Fertigkeiten, die auch vom Pflegepersonal übernommen werden, definiert. PubMed wurde mit verschiedenen Begriffen durchsucht, die diese praktischen Basisfertigkeiten beschreiben. Insgesamt wurden 3467 identifizierte Publikationen gesichtet und 205 wurden schließlich auf ihre Eignung geprüft. Ergebnisse: 43 Studien, die mindestens eine praktische Basisfertigkeit, einen Vergleich zweier Gruppen von Medizinstudierenden und Wirkungen auf die studentische Leistung beinhalteten, wurden analysiert. Sieben praktische Basisfertigkeiten und 15 verschiedene Lehrmethoden konnten identifiziert werden. Die konsistentesten Ergebnisse in Bezug auf effektive Lehre und den Erwerb von praktischen Basisfertigkeiten wurden für strukturiertes Fertigkeitentraining, Feedback und selbstgesteuertes Lernen gefunden. Simulation war mit spezifischen Lehrmethoden wirksam und in mehreren Studien fanden sich keine Unterschiede in Bezug auf Lehreffekte zwischen Experten oder Peers als Lehrende. Multimedia-gestützte Instruktion zeigte bei Anwendung in geeignetem Rahmen ebenfalls positive Effekte für das Erlernen praktischer Basisfertigkeiten. Fazit: Eine Kombination von freiwilligem oder obligatorischem Selbststudium mit Multimedia-gestützten Anwendungen wie Video-Clips in Kombination mit einem strukturierten Programm, das die Möglichkeit für individuelle Übungen mit persönlichem Feedback von Peers oder Lehrenden beinhaltet, könnte eine gute Möglichkeit für das Erlernen praktischer Basisfertigkeiten bieten. basic practical skillsclinical skillsphysical examinationskills trainingundergraduate medical education ==== Body Introduction During undergraduate medical education knowledge, skills, and attitudes have to be acquired by medical students to provide competent patient care after graduation. The term “skills” often comprises communication skills, physical examination skills, practical skills, psychomotor skills, clinical skills, technical skills and others without further specification. A current approach in health profession education is the development of competence-based undergraduate curricula [56]. In Germany, a National Competence Based Catalogue of Learning Objective for Undergraduate Medical Education (NKLM) came into effect in June 2015 [10]. Many of the competences described in the NKLM include the acquisition of basic practical skills [http://www.nklm.de [accessed 19.9.2015]]. Regarding basic practical skills (i.e. accomplishing a task like knot tying or cardiac auscultation) [http://curriculum.racgp.org.au/media/12371/proceduralskills.pdf [accessed 19.9.2015]], complaints have been raised by medical stakeholders that medical graduates execute such skills below the expected level of performance [43]. Furthermore, third year undergraduate medical students reported their competence in core clinical skills like rectal examination or insertion of a nasogastric tube on average with 4.7 on a 6-point (1=excellent) Likert scale [11]. On the other hand, different basic clinical skills training programs seem to offer medical students different levels of preparedness with respect to physical diagnostic skills [42], suggesting that some teaching methods for practical skills might result in better performance. An obstacle for teaching practical skills well has been identified in some teachers’ lack of confidence in their own physical examination skills [38]. To develop or remodel an undergraduate medical curriculum with the goal of being competency-based, optimal and effective teaching strategies how to acquire basic practical skills need to be implemented. This review aims to identify and summarize teaching approaches of basic practical skills in undergraduate medical education, which provide evidence with respect to student learning of these skills. Methods Basic practical skills No unanimous term is used in medical education literature to describe basic practical skills. Neither is there a unique definition, which skills can be summarized under basic practical skills. The terms procedural skills, (basic) surgical skills, physical examination skills, (basic) clinical skills, hands-on skills, basic skills, technical skills, elementary techniques, motor skills, (basic) surgical techniques, psychomotor skills, psychomotor task, clinical technical skills, manual tasks, elementary procedures and physical diagnosis, and basic technical procedures are used inconsistently to describe similar or overlapping practical skills including either aspects of physical examination or procedures involving medical instruments, resembling the technical dimension of professional competence by Epstein and Hundert [8]. In 2011, the GMA Committee on Practical Skills published a consensus statement on 289 practical skills in undergraduate medical education [49]. Mastery of the different skills should either be achieved by medical students before starting clerkships, final year training or internship and the different levels of teaching and learning for the individual skill were defined as having watched a teacher while performing the respective skill, having conducted the skill oneself under supervision or being able to use the skill appropriately to the situation by oneself [49]. Based on the highest levels of this classification and excluding communication, emergency, and soft skills, skills included in this review were defined as being basic practical skills when they matched one of the following criteria: physical examination skills which every student should be able to perform independently of the intended postgraduate training (e.g. cardiac auscultation), simple routine medical skills which get better with practice (e.g. venipuncture), practical skills, which are also performed by nurses (e.g. bladder catheterization). Selection criteria We wished to identify studies that described teaching methods for any of the basic practical skills described in the definition above and provided evidence that the respective teaching methods showed an effect on students’ performance of the respective skill. Strategy of literature search Since a clear definition for basic practical skills is lacking, we searched PubMed using the search term “medical education” in combination with either “basic skills”, “basic technical procedures”, “clinical skills”, “clinical technical skills”, “hands-on skills”, “master learning”, “motor skills”, “physical examination skills”, “practical skills”, “procedural skills”, “psychomotor skills”, “surgical skills”, “surgical techniques”, or “technical skills” for articles in either English or German published between January 2000 and September 2015. The volumes 2000 to 2010 of the GMS Zeitschrift für Medizinische Ausbildung that are not listed in PubMed were searched individually. This original compiled search resulted in a total of 3467 publications. For further consideration, only full research articles with undergraduate medical students being the studies’ subjects were included; short papers, letters or comments were excluded. In this step all titles and abstracts were screened and only manuscripts including at least one of the desired basic practical skills were included for further screening, resulting in 205 articles. Duplicates were also excluded in this step. These 205 manuscripts were subdivided in the following categories: controlled studies, theoretical advices and reviews, evaluation studies and surveys, pre-post studies, observational studies, and qualitative studies. Results Manuscripts matching the different categories are displayed in Figure 1 (Fig. 1). We identified 43 publications as controlled studies including medical students and at least one basic skill matching the selection criteria, comparing at least two groups and using an assessment to measure skill performance. Table 1 (Tab. 1) shows the different basic practical skills covered by the 43 publications and specifies their respective teaching or learning methods. The number of papers attributed to a specific teaching or learning method is shown in Table 2 (Tab. 2). Structured Skills Training In general, student participation in structured skills training is associated with improved assessment outcomes - with respect to physical examination compared to students just following a clerkship [15] and with respect to injection and suturing skills compared to students not participating in a specific training program [26]. Different types of structured skills training programs have been developed leading to different outcomes with respect to students’ performance regarding physical examination skills or suturing. A structured bedside training where attendings received guidelines to demonstrate and observe students doing physical examinations, led to better student performance in half of the OSCE stations covering heart and lung examination [44], while another study with specific weekly bedside instructions in physical examination skills compared with the usual bedside teaching showed better OSCE results for musculoskeletal, pulmonary and gastrointestinal but not for cardiovascular exam [50]. Students being taught physical examination in a clinical coaching program with weekly structured teaching by paid general practitioners showed better OSCE results than students receiving weekly opportunistic teaching by unpaid hospital-based specialists without specific feedback [39]. Training in a skills lab with a set of specific exercises on dummies or paired peers versus standard bedside teaching was associated with better OSCE scores for abdominal examination but not for cardiac auscultation [20]. Another study reporting on an obligatory training program in a skills laboratory where students can practice skills on each other, on models, manikins, and standardized patients, shows better OSCE results for students participating in this program versus students from a traditional curriculum for lung and heart examination but not for examination of the abdomen and for injection and suturing techniques [33]. With respect to the latter, a specific surgical skills training workshop series demonstrated significant improvement for suturing skills [28], while another study identified the optimal instructor : student ratio to be one instructor for four students [7]. Different instructors The question, who might be the optimal instructor for clinical skills teaching, is addressed in nine controlled studies. For physical examination skills there is no difference in OSCE results between students taught by peers versus physicians [18], generalists versus specialists [63] or standardized physical examination teaching associates versus faculty [1]. Better OSCE results were reported for students taught by full-time faculty versus part-time faculty [64]. Regarding suturing skills, peer teaching and faculty teaching lead to equal practical test results [6] and there was also no difference between being taught by a non-surgical skills coach versus being taught by a surgeon [23]. For injections skills, peer-teaching lead to similar student skills like faculty teaching [59]. For bladder catheterization, one study showed equal results for students being taught by peers or faculty [55], while another study showed better performance of students being taught by experts [58]. Multimedia-assisted instruction With respect to multimedia-assisted instruction different aspects of the application of multimedia have been studied. Students who had access to standard video clips for different aspects of physical examination received better OSCE results than a cohort of students studying without being given this learning opportunity [32]. Students who learned with the “click-version” of an interactive program of abdominal exam performed better than students who worked with the “drag-version” of the same program [21]. When cardiac auscultation was learned with a CD-ROM in addition to the usual clinic rotation, auscultation skills significantly improved and this improvement lasted even until one year after the intervention [52]. Concerning suturing skills, self-study with interactive video instruction lead to similar performance than self-study with video and expert instruction [31]. Feedback including the possibility to watch one’s own performance on a video was associated with better suturing skills than verbal feedback alone [9]. Working alone with an interactive video on suturing skills was associated with better performance than working in student tandems with the same video [45]. With respect to different kinds of videos teaching suturing performance was best when students were shown videos with the correct task and videos explaining errors [46]. For bladder catheterization, computer-assisted learning was as effective as expert feedback in a simulation setting [58]. Simulation Students who received training of heart sounds with a high-fidelity simulator (Harvey) did not perform significantly better than students who trained with a low-fidelity simulator (CD) [5]. Students who were given the opportunity to train cardiac examination skills on standardized patients and a cardiac simulator (Harvey) performed significantly better in cardiac skills than a control group, who only worked with standardized patients [22]. Training of abdominal examination with standardized patients lead to better student performance in this examination skill than a lecture alone [12]. Training with a manikin (Laerdal SimMan 3G) resulted in better chest examination skills than performing chest examination on a peer [53]. With respect to gastric-tube insertion, being involved in role-play skills lab sessions did not result in better technical performance of this skill [30]. Simulator skills lab training for cannulation, venipuncture, and injection resulted in better performance of these skills compared to not having received simulator training [59]. Practicing injection skills on a manikin compared to another group who received additional training using a fellow student as surrogate patient did not lead to any differences with respect to the technical performance of an injection [4]. Feedback, self-guided learning and voluntary training Feedback has been identified as an important method to improve the learning of skills. One of the structured weekly programs for physical examination skills described above included ongoing formative assessment and feedback by the instructors for the students, who eventually showed better OSCE performance [39]. For suturing skills, different aspects of feedback – besides watching a video with one’s individual performance [9] – for this specific skill acquisition have been studied. Verbal feedback from an expert that is adapted to the personal situation of the learner was more effective than self-accessed computer generated feedback for suturing performance [34]. Furthermore, real-time feedback with an apparatus measuring the force applied by the learner’s hand while tying a knot led to an appropriate decrease of the force needed for this sensitive task compared to a group without this specific feedback [35]. While one study showed that suture training with feedback lead to better suturing skills than self-directed suture training [6], other studies reported that self-guided suturing practice [3] or a self-directed schedule for suturing practice [47] were associated with better suturing skill acquisition and additional expert feedback lead to no further skill improvement [31]. Furthermore, voluntary participation in reflective writing and skills practice [54] and voluntary practice with positively deviant peers [62] led to better clinical skill performance of participating students. Other teaching/learning methods Observing peers performing a physical examination was associated with better student performance in an assessment of physical examination skills than just receiving feedback from a patient instructor [27]. Using ultrasonography in learning clinical examination showed some improvement for correct lung and liver palpation but not for thyroid palpation [13]. Being taught knot-tying with the kinesthetic method has led to significant better performance of this task by novices compared to medical students, who watched a traditional video [17]. Furthermore, working with process goals while learning suturing leads to greater skill retention than working with outcome goals [3]. In addition, mental imagery technique appeared to transfer learning better from practice suturing sessions to actual surgical assessment than textbook study [48]. Better cannulation skills could be demonstrated by students using cumulative sum charts to log their cannulation attempts during their finial year [51]. When students were taught gastric-tube insertion with Peyton’s Four-Step Approach as teaching method they did not differ from their peers, who received standard instruction in terms of correct stepwise performance but sored better in global rating assessing professionalism [25]. Discussion Many different variables have been identified from controlled studies to influence students’ learning of basic practical skills ranging from more global factors like structured skills training, multimedia-assisted instruction or different instructors to specific teaching methods like feedback, mental imagery or Peyton’s Four-Step Approach. Besides very heterogeneous teaching methods, the teaching itself was applied in different phases of the undergraduate medical curriculum from first year to final year. Depending on the context of the study the influence of the same variable can lead to different results making it all the more difficult to give comprehensive recommendations which is the best method to teach which basic practical skill. One recommendation that can be given unrestrictedly is, that providing a skills training of any sort seems to lead to better skills learning in undergraduate medical education than just participating in usually unstructured clerkships or bedside teaching [15], [20], [50]. Whether the optimal instructor : student ratio (1:4) that was identified for learning suturing [7] will also be optimal for, e.g. learning cannulation, can still not be answered. Interestingly, when post-graduate year (PGY)-1 residents and graduating PGY-3 residents where compared in an OSCE on basic practical skills, a significant increase in skills was only seen between PGY-1 week 0 and PGY-1 week 4 residents, but not between the latter and PGY-3 residents [60]. When it comes to the question who is supposed to be the instructor for a certain basic skill, the recommendation that the most skilled clinicians should be recruited to teach physical examination [36] cannot be followed uncontradictedly. Several studies showed that being taught by instructed peers or trained personnel, generalists or specialists, leads to similar skill performance in medical students [1], [6], [18], [23], [59], [63]. However, one study demonstrated that students showed better performance of bladder catheterization after being taught by an expert [58] and students who had been taught by full-time faculty performed better than students who were taught by part-time faculty [64]. With the majority of studies providing evidence that basic practical skills can be taught equally well by educated peers or non-physicians we believe that the recommendation to recruit personnel other than physicians to teach basic practical skills can be given. This recommendation refers merely to the technical part of the teaching which was in the focus of this review. However, a more competence-based approach to undergraduate medical education, which the NKLM aims for, will eventually need the integration of physical examination skills with communication and clinical reasoning, which might require medical experts as role models to teach the students [14]. With respect to practical skills, which are also performed by nurses, injections, bladder catheterization, and gastric tube insertion were learned equally well when taught by peers compared to experienced faculty staff in a skills lab [59]. Meanwhile, 10 medical faculties in Germany have a peer-assisted learning program for gastric tube insertion in their skills lab [2]. On the other hand, these skills could be learned during the three months of nursing practice every undergraduate medical student in Germany has to complete [http://www.gesetze-im-internet.de/bundesrecht/_appro_2002/gesamt.pdf [accessed 28.9.2015]]. However, it can be assumed that nursing practice for medical students is a mix of many unstructured educational events like clerkships which have been identified in a focus group study to provide no real opportunity to train skills [40]. A possible structured learning approach for venipuncture, bladder catheterization etc. could be a multiprofessional program for medical and nursing students which showed a significant increase in self-assessed confidence levels for the taught skill in a pre-post evaluation [57]. In times of tight budgets, the result that a CD-ROM worked as well to learn cardiac auscultation as the much more expensive high-fidelity simulator Harvey [5] seems to be as welcome as the finding that such a CD in addition to a clinic rotation led to better and sustainable auscultation skills [52]. Since all medical faculties in Germany provide skills lab training as part of their undergraduate medical education [2], this might be the place were learning with CDs could be integrated in auscultation exercises. How students should work with multimedia, e.g. in teams or by themselves, might depend on the type of skill they are supposed to learn. For suturing, better performance was observed when students worked individually with an interactive video [45]. For more complex tasks involving competences like clinical reasoning, having worked with a tandem partner on virtual patient cases led to better results in a knowledge test [19]. Hence, studies on individual work or teamwork with multimedia cannot be generalized. Furthermore, when designing multimedia tools for studying skills great care has to be taken to choose the best application to enhance learning. When students learned with a “drag version” of an interactive program of abdominal exam they performed worse than their peers who had used the simpler “click-version” of the same program [21]. The difference might have been due to possible cognitive overload generated by dragging items on the screen rather than simplifying the task without decontextualizing it to adapt the intrinsic load of learning the specific skill to the developmental state of the learner [61]. Feedback in general and self-directed learning in the sense of organizing one’s own schedule, being aware which skill to practice, or participating in clearly defined voluntary exercises have been identified in this review to be beneficial for acquiring different basic practical skills. These findings are in line with the results that – moving away from the ‘see-one-do-one-teach-one’ approach – structured programs support basic skill learning and might solve the complaints that unstructured electives cannot be relied upon to provide students with practical skills learning [41]. Another study among German undergraduate medical students showed, that during electives procedures are often practiced unsupervised and students might acquire incorrect techniques in the absence of feedback [11]. Feedback is recommended in the clinical environment to provide learners with information about their performance for potential improvement and as guidance for students to reassess their attainment of goals [37], which can further support self-directed learning. An additional form of feedback, i.e. observing peers [27], has been found to be associated with better student learning than just the instructor’s feedback alone. Last, but not least, very special haptic feedback methods for practical skills like suturing [35] complete the list to underscore the importance of feedback as a teaching method of basic practical skills. Essential for feedback is the direct observation of the student by the instructor which is especially challenging in the current clinical environment and supported by practical tips [16]. However, for certain basic practical skills, peer teachers [6] and non-physician educators [23] seem to serve equally well as substitute instructors. With respect to applying specific teaching methods for learning basic practical skills, Peyton’s Four-Step Approach has been demonstrated to be effective for teaching gastric tube insertion using a manikin [25]. In another study on practicing the placement of a central venous catheter, an advanced practical skill according to the classification used in our review, Peyton’s Step 3 – the student explains each sub-step while the teacher follows the student’s instruction – was identified to play the most crucial role in contributing to students’ learning success [24]. However, Peyton’s method of instruction was designed for a 1:1 teaching which does not reflect the typical learning situation. In skills labs, mostly small group teaching takes place with the ideal instructor : student ratio for teaching suturing skills having been identified to be 1:4 [7]. In a descriptive study on skills lab training for the same advanced practical skill mentioned above, central venous catheter insertion, a modified Peyton’s Four-Step Approach for small group teaching was described as being well accepted by students and easy for instructors to realize [29]. Whether this approach works equally well for basic practical skills and leads to similar or improved learning has not been described yet. One limitations of our study is that many different search terms had to be used because “basic practical skills” as they were defined for our study are included in several other terms in the medical education literature and the list of search terms might still not have been complete. The focus of our review was on trying to summarize the evidence for how to teach basic practical skills in undergraduate medical education with the best effects. Hence, we only chose controlled studies which is a strength of our study. A weakness is, however, that we neglected to appraise the quality of the individual studies included in this review. Furthermore, despite individual assessment of all included papers by both authors, we might have been subject to errors during the data extraction and analysis and there also might have been a risk of reporting biases. Since the main focus of our study was on evidence-based teaching and learning methods in undergraduate medical education, we did not take influences of cognitive psychology and training of instructors into account as suggested influences were only found in manuscripts having been assigned to other categories. In addition, effective teaching methods of teaching skills might also be extracted from studies in postgraduate medical education or from studies on teaching more complex practical skills which were not in the focus of this study. In conclusion, our findings suggest that voluntary or obligatory self-study with multimedia applications like video clips of certain skills in combination with a structured program including the possibility for individual exercise with personal feedback by peers or teachers seems to provide a good learning opportunity for basic practical skills. Whether the combination of these different aspects of teaching which individually improve basic practical skills’ learning will lead to similar effects needs to be evaluated by further educational research. Competing interests The authors declare that they have no competing interest. Table 1 Teaching/learning methods for basic practical skills Table 2 Number of papers identified per teaching/learning method Figure 1 Strategy of literature search and study selection ==== Refs 1 Barley GE Fisher J Dwinnell B White K Teaching Foundational Physical Examination Skills: Study Results Comparing Lay Teaching Associates and Physician Instructors Acad Med 2006 81 10 95 97 10.1097/00001888-200610001-00024 Available from: http://dx.doi.org/10.1097/00001888-200610001-00024 2 Blohm M Lauter J Branchereau S Krautter M Köhl-Hackert N Jünger J Herzog W Nikendei C Peer-assisted learning" (PAL) in the Skills-Lab – an inventory at the medical faculties of the Federal Republic of Germany GMS Z Med Ausbild 2015 32 1 Doc10 10.3205/zma000952 Available from: http://dx.doi.org/10.3205/zma000952 25699102 3 Brydges R Carnahan H Safir O Dubrowski A How effective is self-guided learning of clinical technical skills? It's all about process Med Educ 2009 43 6 507 515 10.1111/j.1365-2923.2009.03329.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2009.03329.x 19493173 4 Chunharas A Hetrakul P Boonyobol R Udomkitti T Tassanapitikul T Wattanasirichaigoon D Medical students themselves as surrogate patients increased satisfaction, confidence, and performance in practicing injection skill Med Teach 2013 35 4 308 313 10.3109/0142159X.2012.746453 Available from: http://dx.doi.org/10.3109/0142159X.2012.746453 23228086 5 De Giovanni D Roberts T Norman G Relative effectiveness of high- versus low-fidelity simulation in learning heart sounds Med Educ 2009 43 7 661 668 10.1111/j.1365-2923.2009.03398.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2009.03398.x 19573189 6 Denadai R Toledo AP Oshiiwa M Saad-Hossne R Acquisition of suture skills during medical graduation by instructor-directed training: a randomized controlled study comparing senior medical students and faculty surgeons Updates Surg 2013 65 2 131 140 10.1007/s13304-013-0199-y Available from: http://dx.doi.org/10.1007/s13304-013-0199-y 23404432 7 Dubrowski A MacRae H Randomised, controlled study investigating the optimal instructor: student ratios for teaching suturing skills Med Educ 2006 40 1 59 63 10.1111/j.1365-2929.2005.02347.x Available from: http://dx.doi.org/10.1111/j.1365-2929.2005.02347.x 16441324 8 Epstein RM Hundert EM Defining and assessing professional competence JAMA 2002 287 2 226 235 10.1001/jama.287.2.226 Available from: http://dx.doi.org/10.1001/jama.287.2.226 11779266 9 Farquharson AL Cresswell AC Beard JD Chan P Randomized trial of the effect of video feedback on the acquisition of surgical skills Br J Surg 2013 100 11 1448 1453 10.1002/bjs.9237 Available from: http://dx.doi.org/10.1002/bjs.9237 24037563 10 Fischer MR Bauer D Mohn K NKLM-Projektgruppe Finally finished! National Competence Based Catalogues of Learning Objectives for Undergraduate Medical Education (NKLM) and Dental Education (NKLZ) ready for trial GMS Z Med Ausbild 2015 32 3 Doc35 10.3205/zma000977 Available from: http://dx.doi.org/10.3205/zma000977 26677513 11 Fischer T Chenot JF Simmenroth-Nayda A Heinemann S Kochen MM Himmel W Learning core clinical skills – a survey at 3 time points during medical education Med Teach 2007 29 3 397 399 10.1080/01421590701316563 Available from: http://dx.doi.org/10.1080/01421590701316563 17786759 12 Fletcher KE Stern DT White C Gruppen LD Oh MS Cimmino VM The Physical Examination of Patients With Abdominal Pain: The Long-Term Effect Of Adding Standardized Patients and Small-Group Feedback to a Lecture Presentation Teach Learn Med 2004 16 2 171 174 10.1207/s15328015tlm1602_9 Available from: http://dx.doi.org/10.1207/s15328015tlm1602_9 15276894 13 Fodor D Badea R Poanta L Dumitrascu DL Buzoianu AD Mircea PA The use of ultrasonography in learning clinical examination – a pilot study involving third year medical students Med Ultrason 2012 14 3 177 181 10.11152/mu.2013.2066.143.df177 Available from: http://dx.doi.org/10.11152/mu.2013.2066.143.df177 22957320 14 Goldstein EA Maclaren CF Smith S Mengert TJ Maestas RR Foy HM Wenrich MD Ramsey PG Promoting fundamental clinical skills: a competency-based college approach at the University of Washington Acad Med 2005 80 5 423 433 10.1097/00001888-200505000-00003 Available from: http://dx.doi.org/10.1097/00001888-200505000-00003 15851451 15 Güldal D Ozçakar N Yeniçeri N Dontlu C Ulusel B Comparison of Clinical Skills of 3rd-Year Students Who Completed Structured Clinical Skills Program With 6th-Year Students Who Acquired Clinical Skills in Unsystematic Way Teach Learn Med 2005 17 1 21 26 10.1207/s15328015tlm1701_5 Available from: http://dx.doi.org/10.1207/s15328015tlm1701_5 15691810 16 Hauer KE Holmboe ES Kogan JR Twelve tips for implementing tools for direct observation of medical trainees' clinical skills during patient encounters Med Teach 2011 33 1 27 33 10.3109/0142159X.2010.507710 Available from: http://dx.doi.org/10.3109/0142159X.2010.507710 20874011 17 Huang E Chern H O'Sullivan P Cook B McDonald E Palmer B Liu T Kim E A better way to teach knot tying: a randomized controlled trial comparing the kinesthetic and traditional methods Am J Surg 2014 208 4 690 694 10.1016/j.amjsurg.2014.05.028 Available from: http://dx.doi.org/10.1016/j.amjsurg.2014.05.028 25241957 18 Hudson JN Tonkin AL Clinical skills education: outcomes of relationships between junior medical students, senior peers and simulated patients Med Educ 2008 42 9 901 908 10.1111/j.1365-2923.2008.03107.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2008.03107.x 18694405 19 Jäger F Riemer M Abendroth M Sehnder S Harendza S Virtual patients: the influence of case design and teamwork on students' perception and knowledge – a pilot study BMC Med Educ 2014 14 137 10.1186/1472-6920-14-137 Available from: http://dx.doi.org/10.1186/1472-6920-14-137 25000965 20 Jünger J Schäfer S Roth C Schellberg D Ben-David M Nikendei C Effects of basic clinical skills training on objective structured clinical examination performance Med Educ 2005 39 10 1015 1020 10.1111/j.1365-2929.2005.02266.x Available from: http://dx.doi.org/10.1111/j.1365-2929.2005.02266.x 16178828 21 Kalet A Song HS Sarpel US Schwartz R Brenner J Ark TK Plass J Just enough, but not too much interactivity leads to better clinical skills performance after a computer assisted learning module Med Teach 2012 34 10 833 839 10.3109/0142159X.2012.706727 Available from: http://dx.doi.org/10.3109/0142159X.2012.706727 22917265 22 Kern DH Mainous AG 3rd Carey M Beddingfield A Simulation-Based Teaching to Improve Cardiovascular Exam Skills Performance Among Third-Year Medical Students Teach Learn Med 2011 23 1 15 20 10.1080/10401334.2011.536753 Available from: http://dx.doi.org/10.1080/10401334.2011.536753 21240777 23 Kim MJ Boehler ML Ketchum JK Bueno R Williams RG Dunnington GL Skills coaches as part of the educational team: A randomized controlled trial of teaching of a basic surgical skill in the laboratory setting Am J Surg 2010 99 1 94 98 10.1016/j.amjsurg.2009.08.016 Available from: http://dx.doi.org/10.1016/j.amjsurg.2009.08.016 20103072 24 Krautter M Dittrich R Safi A Krautter J Maatouk I Moeltner A Herzog W Nikendei C Peyton's four-step approach: differential effects of single instructional steps on procedural and memory performance – a clarification study Adv Med Educ Pract 2015 6 399 406 10.2147/AMEP.S81923 Available from: http://dx.doi.org/10.2147/AMEP.S81923 26060417 25 Krautter M Weyrich P Schultz JH Buss SJ Maatouk I Jünger J Nikendei C Effects of Peyton's Four-Step Approach on Objective Performance Measures in Technical Skills Training: A Controlled Trial Teach Learn Med 2011 23 3 244 250 10.1080/10401334.2011.586917 Available from: http://dx.doi.org/10.1080/10401334.2011.586917 21745059 26 Liddell MJ Davidson SK Taub H Whitecross LE Evaluation of procedural skills training in an undergraduate curriculum Med Educ 2002 36 11 1035 1041 10.1046/j.1365-2923.2002.01306.x Available from: http://dx.doi.org/10.1046/j.1365-2923.2002.01306.x 12406263 27 Martineau B Mamede S St-Onge C Rikers RM Schmidt HG To observe or not to observe peers when learning physical examination skills; that is the question BMC Med Educ 2013 13 55 10.1186/1472-6920-13-55 Available from: http://dx.doi.org/10.1186/1472-6920-13-55 23594455 28 Morris M Caskey R Mitchell M Sawaya D Surgical skills training restructured for the 21st century J Surg Res 2012 177 1 33 36 10.1016/j.jss.2012.03.060 Available from: http://dx.doi.org/10.1016/j.jss.2012.03.060 22534253 29 Nikendei C Huber J Stiepak J Huhn D Lauter J Herzog W Jünger J Krautter M Modification of Peyton's four-step approach for small group teaching – a descriptive study BMC Med Educ 2014 14 68 10.1186/1472-6920-14-68 Available from: http://dx.doi.org/10.1186/1472-6920-14-68 24690457 30 Nikendei C Kraus B Schrauth M Weyrich P Zipfel S Herzog W Jünger J Integration of role-playing into technical skills training: a randomized controlled trial Med Teach 2007 29 9 956 960 10.1080/01421590701601543 Available from: http://dx.doi.org/10.1080/01421590701601543 18158671 31 Nousiainen M Brydges R Backstein D Dubrowski A Comparison of expert instruction and computer-based video training in teaching fundamental surgical skills to medical students Surg 2008 143 4 539 544 10.1016/j.surg.2007.10.022 Available from: http://dx.doi.org/10.1016/j.surg.2007.10.022 32 Orientale E Kosowicz L Alerte A Pfeiffer C Harrington C Palley J Brown S Sapieha-Yanchak T Using Web-based Video to Enhance Physical Examination Skills in Medical Students Fam Med 2008 40 7 471 476 18928073 33 Peeraer G Scherpbier AJJA Remmen R De winter BY Hendrickx K van Petegem P Weyler J Bossaert L Clinical Skills Training in a Skills Lab Compared with Skills Training in Internships: Comparison of Skills Development Curricula Educ Health (Abingdon) 2007 20 3 1 9 34 Porte MC Xeroulis G Reznick RK Dubrowski A Verbal feedback from an expert is more effective than self-accessed feedback about motion efficiency in learning new surgical skills Am J Surg 2007 193 1 105 110 10.1016/j.amjsurg.2006.03.016 Available from: http://dx.doi.org/10.1016/j.amjsurg.2006.03.016 17188099 35 Rafiq A Tamariz F Boanca C Lavrentyev V Merrell RC Objective Assessment of Training Surgical Skills Using Simulated Tissue Interface with Real-Time Feedback J Surg Educ 2008 65 4 270 274 10.1016/j.jsurg.2008.05.012 Available from: http://dx.doi.org/10.1016/j.jsurg.2008.05.012 18707659 36 Ramani S Twelve tips for excellent physical examination teaching Med Teach 2008 30 9-10 851 856 10.1080/01421590802206747 Available from: http://dx.doi.org/10.1080/01421590802206747 18821164 37 Ramani S Leinster S AMEE Guide no. 34: Teaching in the clinical environment Med Teach 2008 30 4 347 364 10.1080/01421590802061613 Available from: http://dx.doi.org/10.1080/01421590802061613 18569655 38 Ramani S Orlander JD Strunin L Barber TW Whither bedside teaching? A focus-group study of clinical teachers Acad Med 2003 78 4 384 390 10.1097/00001888-200304000-00014 Available from: http://dx.doi.org/10.1097/00001888-200304000-00014 12691971 39 Régo P Peterson R Callaway L Ward M O'Brien C Donald K. Using a structured clinical coaching program to improve clinical skills training and assessment, as well as teachers' and students' satisfaction Med Teach 2009 31 12 e586 e595 10.3109/01421590903193588 Available from: http://dx.doi.org/10.3109/01421590903193588 19995160 40 Remmen R Denekens J Scherpbier AJ van der Vleuten CP Hermann I van Puymbroeck H Bossaert L Evaluation of skills training during clerkships using student focus groups Med Teach 1998 20 5 428 432 10.1080/01421599880517 Available from: http://dx.doi.org/10.1080/01421599880517 41 Remmen R Derese A Scherpbier A Denekens J Hermann I van der Vleuten C Van Royen P Bossaert I Can medical schools rely on clerkships to train students in basic clinical skills? Med Educ 1999 33 8 600 605 10.1046/j.1365-2923.1999.00467.x Available from: http://dx.doi.org/10.1046/j.1365-2923.1999.00467.x 10447847 42 Remmen R Scherpbier A van der Vleuten C Denekens J Derese A Hermann I Hoogenboom R Kramer A Van Rossum H Van Royen P Bossaert L Effectiveness of basic clinical skills training programmes: a cross-sectional comparison of four medical schools Med Educ 2001 35 2 121 128 11169083 43 Ringsted C Schroeder TV Henriksen J Ramsing B Lyngdorf P Jønsson V Scherpbier A Medical students' experience in practical skills is far from stakeholders' expectations Med Teach 2001 23 4 412 416 10.1080/01421590120043017 Available from: http://dx.doi.org/10.1080/01421590120043017 12098390 44 Roberts L Lu WH Go RA Daroowalla F Effect of Bedside Physical Diagnosis Training on Third-Year Medical Students' Physical Exam Skills Teach Learn Med 2014 26 1 81 85 10.1080/10401334.2013.857329 Available from: http://dx.doi.org/10.1080/10401334.2013.857329 24405350 45 Rogers DA Regehr G Gelula M Yeh KA Howdieshell TR Webb W Peer Teaching and Computer-Assisted Learning: An Effective Combination for Surgical Skill Training? J Surg Res 2000 92 1 53 55 10.1006/jsre.2000.5844 Available from: http://dx.doi.org/10.1006/jsre.2000.5844 10864482 46 Rogers DA Regehr G MacDonald J A role for error training in surgical technical skill instruction and evaluation Am J Surg 2002 183 3 242 245 10.1016/S0002-9610(02)00798-5 Available from: http://dx.doi.org/10.1016/S0002-9610(02)00798-5 11943119 47 Safir O Williams CK Dubrowski A Backstein D Carnahan H Self-directed practice schedule enhances learning of suturing skills Can J Surg 2013 56 6 E142 E147 10.1503/cjs.019512 Available from: http://dx.doi.org/10.1503/cjs.019512 24284153 48 Sanders CW Sadoski M van Walsum K Bramson R Wiprud R Fossum TW Learning basic surgical skills with mental imagery: using the simulation centre in the mind Med Educ 2008 42 6 607 612 10.1111/j.1365-2923.2007.02964.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2007.02964.x 18435713 49 Schnabel KP Boldt PD Breuer G Fichtner A Karsten G Kujumdshiev S Schmidts M Stosch C A consensus statement on practical skills in medical school – a position paper by the GMA Committee on Practical Skills GMS Z Med Ausbild 2011 28 4 Doc58 10.3205/zma000770 Available from: http://dx.doi.org/10.3205/zma000770 22205916 50 Smith MA Burton WB Mackay M Development, impact, and measurement of enhanced physical diagnosis skills Adv in Health Sci Educ 2009 14 4 547 556 10.1007/s10459-008-9137-z Available from: http://dx.doi.org/10.1007/s10459-008-9137-z 51 Smith SE Tallentire VR Spiller J Wood SM Cameron HS The educational value of using cumulative sum charts Anaesthesia 2012 67 7 734 740 10.1111/j.1365-2044.2012.07100.x Available from: http://dx.doi.org/10.1111/j.1365-2044.2012.07100.x 22420772 52 Stern DT Mangrulkar RS Gruppen LD Lang AL Grum CM Judge RD Using a Multimedia Tool to Improve Cardiac Auscultation Knowledge and Skills J Gen Intern Med 2001 16 11 763 769 10.1111/j.1525-1497.2001.10347.x Available from: http://dx.doi.org/10.1111/j.1525-1497.2001.10347.x 11722691 53 Swamy M Bloomfield TC Thomas RH Singh H Searle RF Role of SimMan in teaching clinical skills to preclinical medical students BMC Med Educ 2013 13 20 10.1186/1472-6920-13-20 Available from: http://dx.doi.org/10.1186/1472-6920-13-20 23394435 54 Tagawa M Imanaka H Reflection and selfdirected and group learning improve OSCE scores Clin Teach 2010 7 4 266–270 10.1111/j.1743-498X.2010.00377.x Available from: http://dx.doi.org/10.1111/j.1743-498X.2010.00377.x 21134204 55 Tolsgaard MG Gustafsson A Rasmussen MB HØiby P Müller CG Ringsted C Student teachers can be as good as associate professors in teaching clinical skills Med Teach 2007 29 6 553 537 10.1080/01421590701682550 Available from: http://dx.doi.org/10.1080/01421590701682550 17978968 56 Tse AM Iwaishi LK King CA Harrigan RC A collaborative approach to developing a validated competence-based curriculum for health profession students Educ Health (Abingdon) 2006 19 3 331 344 17178515 57 Tucker K Wakefield A Boggis C Lawson M Roberts T Gooch J Leraning together: clinical skills teaching for medical and nursing students Med Educ 2003 37 7 630 637 10.1046/j.1365-2923.2003.01558.x Available from: http://dx.doi.org/10.1046/j.1365-2923.2003.01558.x 12834421 58 Walsh CM Rose DN Dubrowski A Ling SC Grierson LE Backstein D Carnahan H Learning in the Simulated Setting: A Comparison of Expert-, Peer-, and Computer-Assisted Learning Acad Med 2011 86 10 Suppl 12 16 10.1097/ACM.0b013e31822a72c7 Available from: http://dx.doi.org/10.1097/ACM.0b013e31822a72c7 59 Weyrich P Celebi N Schrauth M Möltner A Lammerding-Köppel M Nikendei C Peer-assisted versus faculty staff-led skills laboratory training: a randomised controlled trial Med Educ 2009 43 2 113 120 10.1111/j.1365-2923.2008.03252.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2008.03252.x 19161480 60 Willett LL Estrada CA Castiglioni A Massie FS Heudebert GR Jennings MS Centor RM Does residency training improve performance of physical examination skills? Am J Med Sci 2007 333 2 74 77 10.1097/00000441-200702000-00002 Available from: http://dx.doi.org/10.1097/00000441-200702000-00002 17301584 61 Young JQ Van Marrienboer J Durning S Ten Cate O Cognitive Load Theory: implications for medical education. AMEE Guide No. 86 Med Teach 2014 36 5 371 384 10.3109/0142159X.2014.889290 Available from: http://dx.doi.org/10.3109/0142159X.2014.889290 24593808 62 Zaidi Z Jaffery T Shahid A Moin S Gilani A Burdick W Change in action: using positive deviance to improve student clinical performance Adv in Health Sci Educ 2012 17 1 95 105 10.1007/s10459-011-9301-8 Available from: http://dx.doi.org/10.1007/s10459-011-9301-8 63 Zakowski LJ Seibert C VanEyck S Skochelak S Dottl S Albanese M Can Specialists and Generalists Teach Clinical Skills to Second-year Medical Students with Equal Effectiveness? Acad Med 2002 77 10 1030 1033 10.1097/00001888-200210000-00019 Available from: http://dx.doi.org/10.1097/00001888-200210000-00019 12377682 64 Zeng J Zuo C Wang Y A Controlled Trial to Compare the Teaching Quality of Clinical-Skills Training Faculty: The Clinician-Educator Career Path in China Teach Learn Med 2014 26 2 146 152 10.1080/10401334.2014.892393 Available from: http://dx.doi.org/10.1080/10401334.2014.892393 24702550

PMC005xxxxxx/PMC5003144.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00106010.3205/zma001060Doc61urn:nbn:de:0183-zma0010608ArticleStudy on the Interrater Reliability of an OSPE (Objective Structured Practical Examination) – Subject to the Evaluation Mode in the Phantom Course of Operative Dentistry Studie zur Interrater-Reliabilität einer OSPE (Objective Structured Practical Examination) in Abhängigkeit vom Bewertungsmodus im Phantomkurs der Zahnerhaltungskunde Schmitt Laura 1Möltner Andreas 2Rüttermann Stefan 3Gerhardt-Szép Susanne *31 Goethe-University Frankfurt am Main, Carolinum Dental University Institute GmbH, Department of Orthodontics, Frankfurt/Main, Germany2 University Heidelberg, Medical Faculty, Competence Centre for Examinations in Medicine/Baden-Württemberg, Heidelberg, Germany3 Goethe-University Frankfurt am Main, Carolinum Dental University Institute GmbH, Department of Operative Dentistry, Frankfurt/Main, Germany*To whom correspondence should be addressed: Susanne Gerhardt-Szép, Goethe-University Frankfurt am Main, Carolinum Dental University Institute GmbH, Department of Operative Dentistry, D-60596 Frankfurt/Main, Germany, Phone: +49 (0)69/6301-7505, Fax: +49 (0)69/6301-3841, E-mail: s.szep@em.uni-freiburg.de15 8 2016 2016 33 4 Clinical skillsDoc6123 10 2015 03 6 2016 01 4 2016 Copyright © 2016 Schmitt et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001060.shtmlIntroduction: The aim of the study presented here was to evaluate the reliability of an OSPE end-of-semester exam in the phantom course for operative dentistry in Frankfurt am Main taking into consideration different modes of evaluation (examiner’s checklist versus instructor’s manual) and number of examiners (three versus four). Methods: In an historic, monocentric, comparative study, two different methods of evaluation were examined in a real end-of-semester setting held in OSPE form (Group I: exclusive use of an examiner’s checklist versus Group II: use of an examiner’s checklist including an instructor’s manual). For the analysis of interrater reliability, the generalisability theory was applied that contains a generalisation of the concept of internal consistency (Cronbach’s alpha). Results: The results show that the exclusive use of the examiner’s checklist led to higher interrater reliability values than the in-depth instructor’s manual used in addition to the list. Conclusion: In summary it can be said that the examiner’s checklists used in the present study, without the instructor’s manual, resulted in the highest interrater reliability in combination with three evaluators within the context of the completed OSPE. Zusammenfassung Einleitung: Ziel der vorliegenden Studie war es, die Reliabilität einer OSPE-Semesterabschlussprüfung im Phantomkurs der Zahnerhaltungskunde in Frankfurt am Main unter Berücksichtigung unterschiedlicher Bewertungsmodi (Prüfer-Checkliste versus Dozentenmanual) und PrüferInnenanzahl (drei versus vier) zu evaluieren. Methoden: Im Rahmen einer historischen monozentrischen Vergleichsstudie wurden zwei verschiedene Bewertungsmodi (Gruppe I: Verwendung ausschließlich einer Prüfer-Checkliste versus Gruppe II: Verwendung einer Prüfer-Checkliste inklusive eines Dozentenmanuals) im Rahmen einer realen Semesterabschlussprüfung, die in OSPE-Form abgehalten wurde, evaluiert. Zur Analyse der Interrater-Reliabilität wurde die Generalisierbarkeitstheorie verwendet, die eine Verallgemeinerung des Konzepts der internen Konsistenz (Cronbachs alpha) beinhaltet. Ergebnisse: Die Ergebnisse zeigen, dass die alleinige Verwendung der Prüfer-Checkliste zu höheren Interrater-Reliabilitätswerten führte als das zusätzlich zu der Liste verwendete ausführliche Dozentenmanual. Schlussfolgerung: Zusammenfassend kann festgehalten werden, dass die in der vorliegenden Studie verwendete Prüfer-Checkliste ohne Dozentenmanual im Rahmen der durchgeführten OSPE die höchste Interrater-Reliabilität ergab in Kombination mit der Anzahl von drei BewerterInnen. OSCEOSPEchecklistevaluatorinstructor's manualfeedbackdentistry ==== Body Introduction and Problem Definition Performance checks constitute a central element of teaching; their evaluation is characterised primarily by the quality criteria of objectivity, reliability and validity [1], [2]. A GMA (Society for Medical Education) guideline [1] existing for this purpose and the basic standards of the WFME (World Federation for Medical Examination) [3] indicate the following criteria: the examinations must be justiciable the examination procedure is based upon learning goals and the learning effect on students the examination procedures applied and the guidelines for passing the exams must be made known. In 2008, the Science Council recommended the creation of a functioning evaluation system on an international level for performance checks in universities. The task of the assessment tools applied was to analyse teaching performance clearly and dependably [http://www.wissenschaftsrat.de/download/archiv/8639-08.pdf, cited at 23.10.2015]. On the other hand, the current regulations on the licensing of dentists from 1955 contain no guidelines on the examinations held in the course of studies [http://www.gesetze-im-internet.de/z_pro/BJNR000370955.html, cited at 23.10.2015]. Because in the study of dentistry practical skills are reinforced, and thus also examined, we frequently deal with the implementation of competence-orientated methods of examination that can be characterised on the Miller pyramid by “shows how” or “acts” [4]. From this context, OSCE (Objective Structured Clinical Examination) and OSPE (Objective Structured Practical Examination) methods of examination are especially possible [4]. The OSCE method of examination was introduced in 1975 by Harden [5]. Initially conceived for examinations in medicine, today the OSCE is also used for examinations in dentistry. In a 1998 study, Mangour and Brown [6] presented the development and implementation of OSCEs in dentistry for the first time. The terms OSCE and OSPE are usually applied as equivalents and thus with no differentiation. Both Natkin and Guild [7], as well as the AMEE (Association for Medical Education in Europe) Guide No. 81 Part I [8] describe OSPE (as a variation of OSCE) as a method of examination used to test practical skills and knowledge in a non-clinical environment. The authors Wani and Dalvi [9] also noted that the OSPE is an exam form where both the strengths and weaknesses of students’ practical skill can be presented and reviewed. Students and examiners evaluate this exam form as positive and useful [1], [10], [11], [12], [13], [14]. In further studies, such as those of Smith et al. [15], Nayak et al. [16] and Abraham et al. [12], students described both OSCEs and OSPEs in comparison to written and oral examinations as fairer and less stressful exam forms, and preferred the OSPE to more “traditional” exam forms. A study by Schoonheim-Klein et al. [17] was also able to show that OSCEs, in a dental context in particular, promoted skills in the area of clinical competence and learning, as well as a more realistic self-assessment on the part of the students. In addition, the study by Nayak et al. [16] was able to show that through the OSPE, as well as the individual competencies of each student, the practical demonstration of facts and applied knowledge and learning behaviour could be positively influenced. Reliability values between 0.11 and 0.97 were given for the OSCEs [18]. The strongly varying results can be explained primarily by the fact that the parameters under which an OSCE is held (number of stations, number of examiners, length of the exam, type of evaluation mode) could be seen to vary considerably. Independently of the exam form, a differentiation is normally made in evaluation between the methods of “glance and grade” and evaluation based upon defined criteria. These methods were evaluated within the context of dental examination settings [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31]. The majority of the studies referred to above were not able to determine any significant differences between glance and grade and criteria-based methods. Furthermore, they did not take place in a real, but rather an artificial exam environment. There are hardly any studies on OSPEs which, as already mentioned, represent in the strict sense a variation of the OSCE on the assessment of parameters referred to above. It has not been investigated, for instance, to what extent the number of examiners and the type of evaluation methods influence the result of an OSPE. Against this background, the aims of this study were to evaluate the reliability of a real OSPE end-of-semester exam in the phantom course of operative dentistry in Frankfurt am Main, taking various evaluation modes and number of examiners into consideration. Material and Methods The phantom course of operative dentistry ran for a period of one semester (16 weeks). During this time, students had to complete practical work on a variety of simulation models (on extracted human and industrially manufactured artificial teeth). By means of previously defined treatment protocols, various treatment alternatives (for example fillings, laboratory restorations such as inlays, endodontic treatments, etc.) were practised step by step with the help of instructors. As soon as the predefined criteria were fulfilled, each step was ratified by the supervising instructor in a so-called certification booklet. The learning process was accompanied by formative feedback. At the end of the course, both an oral test of knowledge and a summative OSPE took place. The latter was carried out in the simulation unit of so-called “phantom patients”. Two plastic models (upper and lower jaw) were mounted in a “phantom head” consisting of 14 plastic upper jaw teeth and 14 plastic lower jaw teeth. The OSPE consisted of two examination parts, the “filling” (A) and the “inlay” (B), carried out on two different plastic teeth of each respective model. These divided into six “sub-units” (1. “primary preparation”; 2. “under filling and secondary preparation”; 3. “filling”; 4. “inlay”; 5. “filling overall” and 6. “overall grade”) which were each evaluated by the examiners (see Figure 1 (Fig. 1)). These subunits accorded to the criteria based on which the attendance certificates for the course were issued by the instructors. The examiner’s checklist, which contained the list of partial aspects (subunits) mentioned above, was tested out over four consecutive semesters (summer semester 2008 to winter semester 2009) in a regular examination scenario. During the test, the evaluation took place via inspection of the prescribed partial aspects, judged purely on the basis of the view of the examiners’ general quality criteria. School grades were awarded from 1 to 5 (1=very good to 5=insufficient). Each examiner evaluated each student in a real examination scenario (duration: 3 hrs.). This meant that the examiners assessed the students’ work directly at the workplace (on a phantom patient) in a predetermined order during the examination. The students signalled to the examiners that they were ready to submit a subunit for evaluation. During the OSPE, the examiners exchanged no information on the grades they had awarded. After the examiners had independently completed their individual examiner’s checklists, the evaluations were discussed in a joint meeting and it was determined which students should repeat the exam. This took place according to the Delphi principle [http://www.horx.com/zukunftsforschung/Docs/02-M-09-Delphi-Methode.pdf, cited at 23.10.2015]. Examination scenario of the study The present study relates to a period of two semesters (summer semester 2010 = Group I, summer semester 2012 = Group II). The composition of the study population is given in Table 1 (Tab. 1). The inclusion criteria were: students from the 6th semester participation in the phantom course for restorative dentistry examination skills present. The exclusion criteria were defined as follows: students from other semesters course dropouts and course repeaters examination skills not met. The difference in the respective group sizes (I versus II) resulted from the actual size of the semester which was subject to large variations and which was dependent upon the results of the preceding examination. A numerical adjustment of both groups was not feasible as all course participants, according to the study regulations, had to take the exam. The determination of the number of examiners was carried out prior to this study on application for ethical approval. The assignment of identical examiners for both groups was not practical for staffing reasons in the department. In group I, an examiner’s checklist was applied exclusively, as seen in Figure 1 (Fig. 1). In group II, the examiners used the identical examiner’s checklist, but in combination with a detailed instructor’s manual (see Figure 2 (Fig. 2)). This contained clearly defined criteria for the evaluation the individual school grades. In all, five examiners took part in the study (A-E), four women and one man. The examiners were all dentists in the Department for Operative Dentistry, had experience in teaching and in the evaluation of students’ work in the phantom course. Table 2 (Tab. 2) shows their distribution according to number and sex. Examiner A had passed the final examination in dentistry in 1990, examiners B, C, D, and E in 2007, 2008, 2010 and 2011 respectively. They all had experience in conducting the phantom course of operative dentistry. In addition to the others, only A had experience in conducting courses in patient treatment. The examiner’s checklist originated from subject areas that were presented as standard in the current course, and in textbooks for restorative dentistry. These were also similar to the units (filling, inlay) and subunits defined as relevant for examination in operative dentistry raised in Baumann’s study [32] on an interdisciplinary basis between four centres (the universities of Frankfurt, Freiburg, Leipzig and Munich). From the manual attached to group II, examiners were able to learn which evaluation criteria had to be fulfilled in order for a particular grade to be awarded. Train-the-Teacher In each semester, a 45 minute “train-the-teacher course” was held. In this course, examiners were prepared through practical exercises and theoretical instructions on situations in the OSPE and the use of the examiner’s checklist and the instructor’s manual. Thus in advance a relatively high measure of standardisation between the examiners could be achieved. Statistics and Application for Ethical Approval The results were evaluated according to the generalisability theory (G theory) with the statistic programmes SAS 9.2 (SAS Institute Inc., Cary, USA, PROC MIXED) and R (Version 2.15, Package lme4). The variance of the grades obtained is attributed to the influencing factors (in the terminology of the G theory “facets”) “students” and “examiners”, as well as to a measurement error component (see Figure 3 (Fig. 3)). From the variance proportions of the facet “examiner” and error variance relative to the facet “student”, the measurement reliability of the evaluations can be estimated. The generalisability coefficient represents an analogue to internal consistency (Cronbach’s alpha). In contrast to its usual application to various tasks, it is used here for several examiners. The G theory allows assessment of measurement reliability with the adoption of a different number of examiners to that in the actual investigation. In this way, both studies in which a varying number of examiners were involved can be made compatible (in analogy to the Spearman-Brown formula with which a standardisation of reliability for a certain number of tasks is possible). Similarly, the individual examiners (A-E) were evaluated amongst themselves with regard to the parameter “overall grade OSPE”. A sub-group analysis taking in all parameters of examiners A and B completed the statistical analysis. An application for ethical approval for the monocentric comparative study was given the approval number 135/35 by the Ethic Commission of the Department of Medicine of the Goethe University Results Table 3 (Tab. 3) shows the results of the determination of reliability from group I using the examiner’s checklist without the instructor’s manual. In this group, only in the case of three examiners were Cronbach’s alpha values under 0.6 determined for the two criteria “interior wall of the cavity” and “breadth/depth”. In all other subunits, the required value of 0.6 or larger than 0.6 for sufficient reliability could be attained. The subunit “adjacent tooth” achieved the value 1.0; this can be regarded as an ideal reliability value. Furthermore, table 3 (Tab. 3) shows the results of the determination of reliability from group II (using the examiner’s checklist and the instructor’s manual). In order to enable a comparison of the generalisability coefficients in both studies, these were each converted for numbers of both three and four examiners. Thus with the aid of the Spearman-Brown formula, for study I the reliability values for four examiners were determined from those for three examiners, and vice versa for group II. In group II the results for 4 examiners showed a high variance in the calculated Cronbach’s alpha values. For the first subunit “primary preparation” and the accompanying criteria (“proximal contact point” to “breadth/depth), Cronbach’s alpha values under 0.6 were calculated. The same was the case for the subunit “filling” and the accompanying criteria “contact points”, “occlusal design” and “smoothness”, for “inlay total” and accompanying criteria such as “cavity outer edge”, “cavity inner walls”, “breadth/depth”, “smoothness” and “adjacent tooth”. The remaining subunits and criteria were able to achieve the required value for sufficient reliability of 0.6. When comparing individual examiners regarding the parameter “overall grade OSPE”, for the summer semester 2010, correlation coefficients of 0.58 (A versus C), 0.64 (A versus B) and 0.68 (C versus B) were calculated. In the summer semester 2012, the corresponding values were lower (A versus B: 0.33; A versus E: 0.35; A versus D: 0.34; E versus D: 0.52; B versus D: 0.37 and E versus B: 0.35). The results of the subgroup analysis (A versus B, used in both study groups) can be seen in table 3 (Tab. 3). Discussion Limitations One limitation of the present study lies in the type of trial design selected (historical comparison group), as the study was carried out not within one particular semester with a particular student population, but rather in two successive semesters with different participants. Because of two different modes of assessment, a division of the summative examination within the semester was declared inadmissible by the faculty’s ethics commission. The authors see one further limitation in the fact that the examiners from both investigated groups were not equal either in number or team composition. Only two examiners (A and B) evaluated similarly in both study groups. Furthermore, despite the preceding train-the-teacher events, a difference in teaching experience must be assumed. This variation could, however, not be homogenised for staff reasons (expiry of contracts). The elaborate statistical analysis takes account of this limitation and standardises the unequal number of examiners. Modes of evaluation Based on current scientific information, no clear conclusion can be drawn on the benefit of an examiner’s checklist regarding the reliability of an examination. According to the latest research, there are only two studies which have dealt with the different modes of evaluation [19], [20], [26], [28], [29], [33]. In the present study, the best results could be determined regarding a high level of reliability by using the examiner’s checklist without the additional use of an instructor’s manual. A comparable result was achieved in a study by Bazan and Seale [34], where a similarly conceived examiner’s checklist for exam evaluation led to a similar reliability value for the exam. An explanation for this might be that the degree of differentiation in the evaluation guidelines was possibly too detailed to be applied by the examiner during the practical examination, and that the train-the-teacher event was apparently not able to set comparable evaluation standards for the examiners. This problem became particularly apparent in the partial step “inlay adjacent tooth” in which the extensive manual with the defined sub-criteria led to a massive deterioration in the Cronbach’s alpha values. This is also accords with the study by the authors Houpt and Cress [31], which found that the narrower the definition of the predetermined evaluation framework for a criterion was, the sooner discrepancies in measurement accuracy and examiner assessment occurred. A direct comparison of examiners A and B, who examined in both semesters, found that the use of the manual lowered the average correlation (0.68) recorded in summer semester 2010 to a value of 0.33. Despite this clarification, it is still necessary to establish why this partial step in particular caused such extreme deviations. Possibly the wording of the tooth structure definitions (enamel and dentine) resulted in confusion on the side of the examiners as the exam tasks were not carried out on natural teeth consisting of enamel and dentine, but rather on exam teeth made of plastic. Future studies should discuss the exact wording of the manual parameters in terms of content. Examination setting In contrast to the two studies already referred to, the examiners’ evaluation in the present study took place in a real exam situation. As a potential future alternative regarding study design, it would be feasible to give the examiners more time for evaluation. This, however, would require a fundamental revision of the end-of-semester exam at the University of Frankfurt am Main under study here. Considering that three hours were allowed for the whole examination, and that the individual steps were checked simultaneously ad hoc by the examiners with an average of = 22 students, more time spent on the evaluation could only be realised with difficulty. The question arises of why, during the real OSPE examination scenario, so much effort is expended and why the individual steps cannot be evaluated jointly by all the examiners after the exam. The reason for this is that many individual steps during the exam are no longer assessable owing to the succeeding phase, as they are then no longer visible. For example, the “primary preparation” step succeeding “under filling lining/secondary preparation” is no longer assessable as the former is partially concealed after putting in an under filling. This is the same for all partial steps so that at the end of the examination stage “filling”, only the final resulting step remains assessable. This procedure stands in stark contrast to all previously published OSPE examinations where in general the individual steps were both visible and assessable, even after the examination. Compared to the studies made by Goepferd and Kerber [26], Vann et al. [28] and Scheutzel [33] there is a clear difference, as in the examinations investigated there, the similarly complex revaluation form was able to be used under more favourable time conditions. This might explain the different results between the investigation carried out here and the studies previously referred to. Train-the-Teacher OSCE-based examinations show some disadvantages by way of analogy to the advantages already referred to above. According to Miller [4], [35], experience has shown that the OSCE is particularly training intensive and time consuming, and according to Nayak et al. [16], it requires intensive planning and team work. As a rule, the appointed examiners require intensive and systematic training in order to be able to fulfil the requirements of reliability and validity for an OSCE exam [35]. As a result, the OSCE is time consuming and cost intensive in comparison to other exam types such as multiple choice or oral exams [8], [35], [36]. In the context of the present study, a time-consuming preparation of the examiners in a train-the-teacher event was also carried out. As a result, resources of personnel and space, as well as financial resources in the clinical and organisational workflow within the department for restorative dentistry, would have to be found. The duration of a lecture unit (45 mins.) was realistic for this purpose and could be observed by all the examiners. However, the question arises as to how long preparation should effectively be in order to be able to homogenise different experiences in mixed teams in advance. In the summer semester of 2010, the three examiners amongst themselves showed an average correlation of between 0.58 and 0.68. In the summer semester of 2012, in the case of four examiners the identically long train-the-teacher events resulted in correlation values of 0.33 and 0.52. It can be assumed here that in the case of the application of the manual, the train-the-teacher event was not effectively utilised. Examiners On the basis of current data, examiners play an important role in the assessment of reliability. Until now, however, there have been no scientific studies known to us that have made any assessment of how high the minimum number of examiners for a OSPE should be. In this study, it was possible to attain sufficient reliability with three examiners in combination with checklists. According to the results of this investigation, the reliability value can be increased by a higher number of examiners. This increase in reliability values, however, is low in comparison to the number of examiners. In addition, a further increase in the number of examiners would result in greater complexity and expense with regard to organisation and financial costs. In this context, it has to be mentioned critically that no general recommendation can be made for other sites based upon the data available with regard to the number of examiners, as the possibility of having three to four examiners with long experience available for an OSPE examination is neither representative of normal circumstances nor feasible. The author groups Nikendei and Jünger [37] and Norcini et al. [38] came to a similar result. In their study, Natkin and Guild [39] were able to show a significant increase in reliability through a systematic preparation of the evaluators. Similar results were presented by Dhuru [25], in whose study examiners with many years of professional experience and using evaluation sheets achieved the most reliable examination results. In the present study, this can be confirmed only with the use of the checklist, as when the manual was used, the two examiners with the most years’ experience demonstrated only weak correlations. As shown in this investigation, the checklist appears to be capable of further increasing reliability, or of compensating for a lack of examining experience on the part of the evaluators. In Houpt and Kress’s [31] investigation, by contrast, reliability could not be increased for all evaluation criteria. Thus the authors believe that the train-the-teacher events on their own are not able to increase interrater reliability significantly. Training events of this type had the greatest effect with “non-expert” examiners, but relatively little influence with experienced evaluators [31]. Our study was able to confirm this. Exam tasks The number of examination tasks defined in this study, frequently equated with the term “stations” in the literature, should be looked at critically. In the present case only two separate tasks were involved (A. filling and B. inlay), but a total of 22 evaluations were obtained by the evaluators per student in and during the exam. Ultimately we are dealing with the definition of the term “station” in connection with the OSPE which based upon the evidence cannot be deduced from the literature. It must be noted critically that a value of 0.6 for Cronbach’s alpha only has a “sufficient” character. It must therefore also be asked just how valid an examination can then be, and whether it is suitable as a summative examination. According current scientific knowledge, it is our opinion that against this background, variant II cannot be recommend for high stakes examinations. Conclusion The following conclusions may be drawn from this study regarding the question of how an OSPE in dental teaching in a phantom course for operative dentistry can best be reliably designed: an examiner’s checklist without an instructor’s manual resulted in higher interrater reliability in the context of the OSPEs carried out the evaluation of students’ exam performance in the context of the OSPE should if possible be undertaken by at least three examiners. Acknowledgements The authors would like to thank the students of the 6th semester in the section for operative dentistry and the dental course assistants who also contributed to the evaluation of the OSPE. Competing interests The authors declare that they have no competing interests. Table 1 Composition of the study population. Table 2 Data for the examiners A to E, who evaluated group 1 in SS 2010 (A, B, C) and group 2 in SS 2012 (A, B, D, E) (f = female, m = male). Table 3 Results of group I and group II. The corresponding reliability values (Cronbach’s alpha) are given for three and four examiners. In the column “A vs. B”, the results of the subgroup analysis are presented. Identification with * means that differing from the real examination scenario, a conversion into another number of examiners (abbreviations: CL = cavity lining, vs. = versus). Figure 1 Examiner’s checklist group I and group II with both tasks A (filling) and B (inlay). The abbreviation UF stands for under filling. Figure 2 Instructor’s manual for group II with the evaluation criteria of both tasks A and B. The abbreviation p.a. signifies pulpal axial wall. The abbreviation UF stands for under filling, prox. = proximal. Figure 3 The facets of the variance analysis conducted in the study. ==== Refs 1 Fischer MR Gesellschaft für Medizinische Ausbildung Kompetenzzentrum Prüfungen Baden-Württemberg Leitlinie für Fakultätsinterne Leistungsnachweise während des Medizinstudiums: Ein Positionspapier des GMA-Ausschusses Prüfungen und des Kompetenzzentrums Prüfungen Baden-Württemberg GMS Z Med Ausbild 2008 25 1 Doc74 Available from: http://www.egms.de/static/de/journals/zma/2008-25/zma000558.shtml 2 Taylor CL Grey NJ Satterthwaite JD A comparison of grades awarded by peer assessment, faculty and a digital scanning device in a pre-clinical operative skills course Eur J Dent Educ 2013 17 1 16 21 10.1111/j.1600-0579.2012.00752.x Available from: http://dx.doi.org/10.1111/j.1600-0579.2012.00752.x 3 World Federation for Medical Education Basic Medical Education The 2012 Report 2012 Copenhagen WFME Office 4 Miller GE The assessment of clinical skills/competence/performance Acad Med 1990 65 S63 S67 10.1097/00001888-199009000-00045 Available from: http://dx.doi.org/10.1097/00001888-199009000-00045 2400509 5 Harden RM Stevenson M Downie WW Wilson GM Assessment of clinical competence using objective structured examination Br Med J 1975 1 447 451 10.1136/bmj.1.5955.447 Available from: http://dx.doi.org/10.1136/bmj.1.5955.447 1115966 6 Manogue M Brown G Developing and implementing an OSCE in dentistry Eur J Dent Educ 1998 2 2 51 57 10.1111/j.1600-0579.1998.tb00039.x Available from: http://dx.doi.org/10.1111/j.1600-0579.1998.tb00039.x 10865340 7 Natkin E Guild RE Evaluation of preclinical laboratory performance: a systematic study J Dent Educ 1967 31 2 152 161 5228846 8 Khan KZ Ramachandran S Gaunt K Pushkar P The Objective Structured Clinical Examination (OSCE): AMEE Guide No. 81. Part I: an historical and theoretical perspective Med Teach 2013 35 9 e1437 e1446 10.3109/0142159X.2013.818634 Available from: http://dx.doi.org/10.3109/0142159X.2013.818634 23968323 9 Wani P Dalvi V Objective Structured Practical Examination vs Traditional Clinical Examination in Human Physiology: Students perception Int J Med Sci Public Health 2013 2 3 522–547 10.5455/ijmsph.2013.080320133 Available from: http://dx.doi.org/10.5455/ijmsph.2013.080320133 10 Schoonheim-Klein M Muijtjens A Muijtens A Habets L Manogue M van der Vleuten C Hoogstraten J Van der Velden U On the reliability of a dental OSCE, using SEM: effect of different days Eur J Dent Educ 2008 12 3 131–137 10.1111/j.1600-0579.2008.00507.x Available from: http://dx.doi.org/10.1111/j.1600-0579.2008.00507.x 18666893 11 Hofer M Jansen M Soboll S Potential improvements in medical education as retrospectively evaluated by candidates for specialist examinations Dtsch Med Wochenschr 2006 131 8 373–378 10.1055/s-2006-932527 Available from: http://dx.doi.org/10.1055/s-2006-932527 16479467 12 Abraham RR Raghavendra R Surekha K Asha K A trial of the objective structured practical examination in physiology at Melaka Manipal Medical College. India Adv Physiol Educ 2009 33 1 21–23 10.1152/advan.90108.2008 Available from: http://dx.doi.org/10.1152/advan.90108.2008 19261756 13 Adome RO Kitutu F Creating an OSCE/OSPE in a resource-limited setting Med Educ 2008 42 5 525–526 10.1111/j.1365-2923.2008.03045.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2008.03045.x 14 Davenport ES Davis JE Cushing AM Holsgrove GJ An innovation in the assessment of future dentists Br Dent J 1998 184 4 192–195 9549918 15 Smith LJ Price DA Houston IB Objective structured clinical examination compared with other forms of student assessment Arch Dis Child 1984 59 1173 1176 10.1136/adc.59.12.1173 Available from: http://dx.doi.org/10.1136/adc.59.12.1173 6524948 16 Nayak V Bairy KL Adiga S Shenoy S Magazine BC Amberkar M Kumari M OSPE in Pharmacology: Comparison with the conventional Method and Students' Perspective Towards Br Biomed Bull 2014 2 1 218 222 17 Schoonheim-Klein ME Habets LL Aartman IH van der Vleuten CP Hoogstraten J van der Velden U Implementing an Objective Structured Clinical Examination (OSCE) in dental education: effects on students' learning strategies Eur J Dent Educ 2006 10 4 226 235 10.1111/j.1600-0579.2006.00421.x Available from: http://dx.doi.org/10.1111/j.1600-0579.2006.00421.x 17038015 18 Chenot JF Ehrhardt M Objective structured clinical examination (OSCE) in der medizinischen Ausbildung: Eine Alternative zur Klausur Z Allg Med 2003 79 2 437 442 19 Sharaf AA AbdelAziz AM El Meligy OA Intra- and inter-examiner variability in evaluating preclinical pediatric dentistry operative procedures J Dent Educ 2007 71 4 540 544 17468316 20 Kellersmann CT Zur Reliabilität der Beurteilung vorklinischer Phantomarbeiten bei Einsatz eines strukturierten Bewertungsbogens 2007 Münster Westfälischer Wilhelms-Universität Münster 21 Lilley JD ten Bruggen Cate HJ Holloway PJ Holt JK Start KB Reliability of practical tests in operative dentistry Br Dent J 1968 125 5 194 197 5244088 22 Fuller JL The effects of training and criterion models on interjudge reliability J Dent Educ 1972 36 4 19 22 4503280 23 Hinkelman KW Long NK Method for decreasing subjective evaluation in preclinical restorative dentistry J Dent Educ 1973 37 9 13 18 4517641 24 Gaines WG Bruggers H Rasmussen RH Reliability of ratings in preclinical fixed prosthodontics: effect of objective scaling J Dent Educ 1974 38 12 672 675 4530794 25 Dhuru VB Rypel TS Johnston WM Criterion-oriented grading system for preclinical operative dentistry laboratory course J Dent Educ 1978 42 9 528 531 280591 26 Goepferd SJ Kerber PE A comparison of two methods for evaluating primary class II cavity preparations J Dent Educ 1980 44 9 537 542 6931147 27 Feil PH An analysis of the reliability of a laboratory evaluation system J Dent Educ 1982 46 8 489 494 6954179 28 Vann WF Machen JB Hounshell PB Effects of criteria and checklists on reliability in preclinical evaluation J Dent Educ 1983 47 10 671 675 6225789 29 Bedi R Lo E King NM Chan T The effect of pictorial criteria upon the reliability of assessments of cavity preparations J Dent 1987 15 5 222 224 10.1016/0300-5712(87)90116-3 Available from: http://dx.doi.org/10.1016/0300-5712(87)90116-3 2445802 30 Jenkins SM Dummer PM Gilmour AS Edmunds DH Hicks R Ash P Evaluating undergraduate preclinical operative skill; use of a glance and grade marking system J Dent 1998 26 6 679 684 10.1016/S0300-5712(97)00033-X Available from: http://dx.doi.org/10.1016/S0300-5712(97)00033-X 9793290 31 Houpt MI Kress G Accuracy of measurement of clinical performance in dentistry J Dent Educ 1973 37 7 34 46 4514565 32 Baumann MP Evaluation von Bewertungskriterien für praktische Studentenarbeiten im Vergleich zur Bewertung per Augenschein 2015 München Medizinischen Fakultät der Ludwig-Maximilians-Universität München 33 Scheutzel P Einfluss des Bewertungssystems auf Objektivität und Reliabilität der Benotung zahnmedizinischer Studentenarbeiten am Phantompatienten GMS Z Med Ausbild 2007 24 1 Doc67 Available from: http://www.egms.de/static/de/journals/zma/2007-24/zma000361.shtml 34 Bazan MT Seale NS A technique for immediate evaluation of preclinical exercises J Dent Educ 1982 46 12 726 728 6960062 35 Barman A Critiques on the Objective Structured Clinical Examination Ann Acad Med Singapore 2005 34 8 478 482 16205824 36 Boursicot K Ware J Hazlett C Objective Structured Clinical Examination Objective Structured Practical Examination Med Educ 1979 31 41 54 37 Nikendei C Jünger J OSCE-praktische Tipps zur Implementierung einer klinisch-praktischen Prüfung GMS Z Med Ausbild 2006 23 3 Doc47 Available from: http://www.egms.de/static/de/journals/zma/2006-23/zma000266.shtml 38 Norcini JJ Maihoff NA Day SC Benson JA Trends in medical knowledge as assessed by the certifying examination in internal medicine JAMA 1989 262 17 2402–2404 10.1001/jama.1989.03430170064029 Available from: http://dx.doi.org/10.1001/jama.1989.03430170064029 2795825 39 Natkin E Guild RE Evaluation of preclinical laboratory performance: a systematic study J Dent Educ 1967 31 2 152 161 5228846

PMC005xxxxxx/PMC5003145.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00105610.3205/zma001056Doc57urn:nbn:de:0183-zma0010569ArticleVideo-based instructions for surgical hand disinfection as a replacement for conventional tuition? A randomised, blind comparative study Videobasierte Instruktionen zur chirurgischen Händedesinfektion als Ersatz für konventionellen Unterricht? Eine randomisierte, geblendete Vergleichsstudie Weber Uwe *1Constantinescu Mihai A. 2Woermann Ulrich 3Schmitz Felix 3Schnabel Kai 31 Bern, Schweiz2 Inselspital Bern, Universitätsklinik für Plastische- und Handchirurgie, Bern, Schweiz3 Universität Bern, Institut für Medizinische Lehre Abteilung für Unterricht und Medien, Bern, Schweiz*To whom correspondence should be addressed: Uwe Weber, Spitalackerstrasse 64, CH-3013 Bern, Schweiz, E-mail: uweweber@hotmail.com15 8 2016 2016 33 4 Clinical skillsDoc5701 3 2015 21 12 2015 08 12 2015 Copyright © 2016 Weber et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001056.shtmlIntroduction: Various different learning methods are available for planning tuition regarding the introduction to surgical hand disinfection. These learning methods should help to organise and deal with this topic. The use of a video film is an alternative to conventional tuition due to the real presentation possibilities of practical demonstration. Objective: This study examines by way of comparison which form of communication is more effective for learning and applying surgical hand disinfection for medical students in their first year of studies: video-based instruction or conventional tuition. Methodology: A total of 50 first-year medical students were randomly allocated either to the “Conventional Instruction” (CI) study group or to the “Video-based Instruction” (VI) study group. The conventional instruction was carried out by an experienced nurse preceptor/nurse educator for the operating theatre who taught the preparatory measures and the actual procedure in a two-minute lesson. The second group watched a two-minute video sequence with identical content. Afterwards, both groups demonstrated practically the knowledge they had acquired at an individual practical test station. The quality (a) of the preparation and (b) of the procedure as well as (c) the quality of the results was assessed by 6 blind experts using a check list. The acceptability of the respective teaching method was also asked about using a questionnaire. Results: The group performance did not differ either in the preparation (t=-78, p<0.44) or in the quality (t=-99, p<0.34). With respect to performance, it was possible to demonstrate a strong treatment effect. In the practical (t=-3.33, p<0.002, d=0.943) and in the total score (t=-2.65, p<0.011, d=0.751), the group with video-based instruction achieved a significantly better result. In response to the question as to which of the two learning methods they would prefer, the significant majority (60.4%) of students stated video instruction. Conclusion: In this study, the use of the video-based instruction emerged as the more effective teaching method for learning surgical hand disinfection for medical students and is preferable to conventional instruction. The video instruction is associated with a higher learning effectiveness, efficiency and acceptability. Zusammenfassung Einleitung: Für die Unterrichtsgestaltung zur Einführung der chirurgischen Händedesinfektion stehen unterschiedliche Lernmethoden zur Verfügung. Diese Lernmethoden sollen dabei helfen, das Unterrichtsthema zu strukturieren und zu bewältigen. Der Einsatz eines Videofilms ist, durch die realen Darstellungsmöglichkeiten der praktischen Demonstration, eine Alternative zum konventionellen Unterricht. Ziel: Mit der vorliegenden Studie wird vergleichend untersucht, welche Vermittlungsform effektiver für das Erlernen und Anwenden der chirurgischen Händedesinfektion von Medizinstudenten im 1. Studienjahr ist: videobasierte Instruktion oder konventioneller Unterricht. Methodik: Insgesamt wurden 50 Medizinstudierende im 1. Studienjahr per Zufall entweder der Lerngruppe „konventionelle Instruktion“ (KI) oder der Lerngruppe „videobasierte Instruktion“ (VI) zugeordnet. Die konventionelle Instruktion erfolgte durch einen erfahrenen Nurse Preceptors/Nurse Educators für den Operationssaal, der die vorzubereitenden Massnahmen sowie die eigentliche Prozedur im Rahmen einer zweiminütigen Lektion vermittelte. Die zweite Gruppe sah eine Videosequenz von 2 Minuten mit identischen Inhalten. Beide Gruppen demonstrierten ihr akkumuliertes Wissen im Anschluss praktisch an einer einzelnen praktischen Prüfungsstation. Die Güte (a) der Vorbereitung und (b) der Prozedur wurde ebenso wie (c) die Qualität des Ergebnisses von 6 geblendeten Experten anhand einer Checkliste beurteilt. Die Akzeptanz der jeweiligen Vermittlungsform wurde mittels eines Fragebogens erfragt. Ergebnisse: Die Gruppenleistungen unterschieden sich weder in der Vorbereitung (t=-78, p<0.44) noch in der Qualität (t=-99, p<0.34). In Hinblick auf die Performanz konnte ein starker Treatment-Effekt nachgewiesen werden. In der Durchführung (t=-3.33, p<0.002, d=0.943) sowie im Total Score (t=-2.65, p<0.011, d=0.751) erzielte die Gruppe mit videobasierter Instruktion ein signifikant besseres Ergebnis. Auf die Frage, welche der beiden Lernmethoden sie bevorzugen würden, wurde von den Studenten mit deutlicher Mehrheit (60.4%) die Videoinstruktion angegeben. Schlussfolgerung: Der Einsatz der videobasierten Instruktion stellte sich in dieser Studie als die effektivere Vermittlungsform zum Erlernen der chirurgischen Händedesinfektion für Medizinstudierende heraus und ist der konventionellen Instruktion vorzuziehen. Die Videoinstruktion ist mit höherer Lerneffektivität, Effizienz und Akzeptanz assoziiert. ==== Body 1. Introduction Various different learning methods are available for planning tuition regarding the introduction to surgical hand disinfection. These learning methods should help to organise and deal with this topic. A learning method should be selected that is suitable for teaching the topic. The transfer of nosocomial infectious agents takes place most frequently via the hands of staff. The prevalence studies for nosocomial infections revealed that an extrapolated 70,000 patients in Switzerland suffer from nosocomial infection, of whom 2,000 die [1]. The risk in the event of a perforation of surgical gloves can be reduced by correct surgical hand disinfection [2]. In her paper, Cloyd describes how students are inadequately prepared for work in the operating theatre and how they have the feeling that they are a burden and wish to have more effective preparation [3]. The question arises here as to how preparation for work in the operating theatre can be improved and which instruction options are available. In her paper “Basic Aseptic Techniques (BAT)”, Leeper describes the entrance level for the operating theatre. Maintenance of the sterile OP environment and surgical hand disinfection were named here as the two most important points for the entrance level that medical students must demonstrate. She describes how a reduction in post-operative wound infection was achieved as the outcome as a result of knowledge of BAT training [4]. The following learning methods are available for the introduction to surgical hand disinfection: Classical theoretical teaching with aims and objectives without any possibility of practical demonstration under guidance. This tuition can be supported by literature and image material. Self-study of the literature and image material without a practical demonstration. Theoretical input on the topic with a demonstration of surgical hand disinfection by a trained specialist (nurse preceptor/nurse educator), with subsequent opportunity for the students to practise in a simulated or real surgical unit. This method with a trained specialist (nurse preceptor/nurse educator) and subsequent opportunity to practise was also named the most often as the first choice for tuition for the skills (Leeper). The use of a video film, an alternative to the real presentation possibilities of practical demonstration. By using a video film, facts and connections can be illustrated better visually with the aid of digital text, graphics and animation and shown in a shorter time. People are visually inclined and visual intelligence plays a most important role [5]. The use of videos has many other advantages, not investigated in this study [6], such as the option of repeated viewing [7] and in specific preparation for examinations [8]. The use of video instructions has been shown in literature to have equally good or mostly positive effects in the teaching of skills in all fields. As early as 1993, Yoder showed that the video group demonstrated greater learning success when learning the cognitive principles of aseptic methods [9]. In a comparison between didactics, video instruction and computer-based training, the video instruction demonstrated a significant improvement in technical skills [10]. Xeroulis et al. demonstrated that video instruction with expert feedback can be an effective method for instructing basic technical skills [11]. In a comparison between computer-based video instruction and conventional tuition for learning basic surgical skills of suturing and knotting the medical students demonstrated no differences when learning the method [12]. In a further study by Shippey, three different learning methods, video instruction, practical with an instructor and independent learning, for learning suturing techniques were compared to each other. The results demonstrated significantly better effects in the case of video instruction [13]. The use of video instruction for simulation with a venous port catheter demonstrated that cognitive and technical knowledge increased [14]. 1.1. Question To what extent can video instruction for surgical hand disinfection be just as effective as conventional instruction? In this non-inferiority study we investigated which method a (VI=video instruction) or b (CI=conventional instruction) had the better effect for learning surgical hand disinfection among first-year medical students at the University of Bern. Based on the literature mentioned in the introduction, we expect that the students will demonstrate at least an equivalent or better performance after the video instruction (VI) than the students from the conventional instruction (CI). The directional hypothesis is thus: H(1):VI≥KI. The null hypothesis is: H(0):VI<KI 2. Methodology 2.1. Random Sampling and Design All 208 human medical students in the first semester at the University of Bern were sent a letter in the autumn of 2011 inviting them to take part in a voluntary study to examine two different teaching methods. The participants were not told the exact content of the teaching methods. 50 human medicine students signed up: 27 men and 23 women. The students were randomised [http://www.psychicscience.org] and divided into the observation groups Video Instruction (VI) and Conventional Instruction (CI). Design: Two different groups of students were compared (between-subjects). The verification of the learning results by the experts was blind. Description of treatment: an educational film regarding conduct in the operating theatre was developed to investigate the question [http://e-learning.studmed.unibe.ch/chirosurf/htmls/slide.html?chirosurf%7Coperation%7Ctheater%7Corganisation%7C1]. This educational film contains a commentated section that is nearly two minutes long (minute: 6:35-8:25) with visual overlays about surgical hand disinfection, which was used for investigating the question. An experienced nurse educator for the operating theatre was provided with the same text and content as in the film. The nurse educator was tasked with teaching the same content in a two-minute instruction with simultaneous demonstration of surgical hand disinfection. 2.2. Instruments A check list for assessing surgical hand disinfection for recording the success of the teaching surrounding surgical hand disinfection was modified according to a template by Wilkinson [15] and adapted to the latest handling methods [16] (c.f. Attachment 1: Check List for Surgical Hand Disinfection ). The checklist was submitted to three experts (one hygienist and two surgeons) who saw no need for further adjustments following these adaptations. It was possible to achieve the maximum scores shown in Table 1 (Tab. 1). Six experts worked through the items on the check list on tablets (iPad ®) using the OSCE-Eval [http://www.e-osce.ch/] application. All six advisors were blind with respect to the students’ instruction method (conventional or video-based) and were instructed beforehand regarding the check list with the observational criteria and the assessment. The students’ level of knowledge regarding surgical hand disinfection was evaluated using an additional questionnaire. After the investigation, the test persons were asked how satisfied they were with the different learning methods and which learning method (video, tutorial or both) they would prefer. 2.3. Practice 2.3.1. Intervention The interventions were started at the same time in two different rooms. The VI group (video instruction) was shown the two-minute video sequence as an introduction to surgical hand disinfection. The CI group (conventional instruction) was taught using the conventional lesson as has been customary to date with the same contents concerning the introduction to surgical hand disinfection by the nurse educator with a simultaneous demonstration. Carrying out the conventional instruction was likewise limited to two minutes. Afterwards the students from both groups were called up according to the allocation scheduled beforehand by the random generator to carry out the practical demonstration at an individual practical test state at the Bern interdisciplinary Skills and Simulated Patients Centre (BiSS) at the Institute for Medical Studies at the University of Bern (IML). Six test stations set up identically with the required materials (with hand disinfectant mixed with Visirub©, mask, protective glasses, UV lamp) were provided. Five minutes were calculated for each run (see Table 2 (Tab. 2)). One minute for orientation and preparation with the required material that the students had to select themselves. Three minutes to correctly carry out the surgical hand disinfection in good time. One minute to check the quality using the UV lamp. During the demonstration the preparation, practical and quality, as described, were assessed using the check list that had been compiled. Following the demonstration, the students attended the instruction of the respective other group. There they were presented with the respective other teaching method within the same timeframe. After the presentation, the students were asked about the different learning methods using our questionnaire (see Attachment 2: Questionnaire ) and about their satisfaction and acceptability. 3. Results The results of the 50 students were analysed. The statistical analysis of the data was carried out with the aid of Microsoft Office Excel 2010 and SPSS 20 for Windows. If one compares the participants in the video instruction (VI) with the conventional instruction (CI) in detail (Table 3), the following picture emerges in the observation areas: in the preparation, the participants from VI group did not perform any differently to the CI group. In the practical, the VI group performed significantly better than the CI group with an effect size of 0.943 (Cohen’s d). In quality, the VI group did not perform any differently to the CI group. In the total score, a significantly better result is demonstrated by VI group with an effect size of 0.751 (Cohen’s d) (see Table 3 (Tab. 3)). In the T-test (see Table 4 (Tab. 4)) a significant influence of the video instruction on the practical is demonstrated (p=0.002) and in the total score (p=0.011). 3.1. Interpretation of the Questionnaire The interpretation of the questionnaire on satisfaction with and the acceptability of the learning methods was voluntary and anonymised. We received 48 completed responses from the 50 participants. In the interpretation of the question regarding prior knowledge, it was ascertained that 7 out of 23 students from the CI group had already been taught surgical hand disinfection. In the VI group it was 3 out of 25 students who likewise made this statement (see Table 5 (Tab. 5)). In response to the question as to which of the interventions they would prefer, 29 (60.4%) answered in favour of the video and 19 (39.6%) in favour of the lesson. Therefore, five students from the conventional instruction group decided on the video having watched the video (see Table 6 (Tab. 6)). In response to the question as to which methods: lesson, video or both, they regarded to be suitable for the preparation for surgical hand disinfection “both” was clearly selected with 37 responses (see Table 7 (Tab. 7)). 4. Discussion The use of video instruction demonstrates, as described in the literature referred to, a positive effect in skills acquisition. Cognitive and technical skills can be improved using the real presentation possibilities of practical demonstration. 24 students were successfully taught in two minutes using the video instruction. If one regards the total expenditure in comparison and the possibility of repeated use, efficiency speaks clearly in favour of video instruction as a result of the greater effectiveness. When interpreting the questionnaire with regard to the question as to which learning method: video instruction or conventional instruction, they would prefer, it emerged that 17 students from the CI group decided in favour of the video and 8 in favour of the lesson. It was surprising that 2/3 of the CI group would prefer video instruction. All of the students selected the decision regarding the other learning method without knowing the results of their demonstration at the practical test station. In the VI group the result was balanced; 12 students preferred video instruction and 13 students preferred conventional instruction. Overall, it was expected that the respective learning method experienced before the practical test would be preferred by the majority. That the video instruction was preferred by the majority is in accord with the results in the literature cited in the introduction and can be ascribed to the known advantages of video instruction (general preference for digital media among the younger generation, the highlighting of key elements and standardisation of procedures) [6]. A clear statement in the interpretation was that a combination of conventional instruction and video instruction was regarded by the students as the ideal teaching unit. This was not unexpected, since an addition of both learning methods is naturally regarded to be a potential advantage by students. Some questions raised during the practical should be considered in further studies: would the differences in the practical between the two intervention groups possibly turn out differently if one had used several tutors each with a smaller group instead of one tutor? Would other learning methods, for example a demonstration with subsequent opportunities to practice in a simulated or real operating theatre, literature with the use of image material, lead to different results? It should also be considered critically that it was more the skills that were being tested in the practical test station and the students in both groups were only taught once for two minutes with the option of assisted practicals, by means of personal feedback and without any chance to ask questions. 5. Conclusions We can answer our question as to whether video instruction for surgical hand disinfection can be just as effective as conventional instruction in the affirmative and reject the null hypothesis. In the evaluation going beyond the pure question, a positive result is demonstrated in all areas in favour of video instruction. In the preparation and in the quality both groups achieved equally good results. The video instruction demonstrated a significantly better effect in the practical and in the total score. The evaluated sequence of surgical hand disinfection in the educational film demonstrates an effective and efficient transfer of knowledge. In response to the question as to which of the two methods they would prefer, the video instruction was selected clearly. A further relevant aspect is that a video film represents a meaningful enhancement to the choice of learning as a result of more extensive possibilities of presenting the teaching content and can increase cognitive and technical knowledge. Video instruction provides the opportunity to present relevant learning content realistically and for use in independent studies. Videos help the students to repeat learning content from the lesson independently at their own speed. Looking to the future, the study of the long-term effect and the different teaching methods would be of interest. 6. Acknowledgements I would like to thank the following people most sincerely for their support during this project: Mr Giovanni Ferrieri, Multimedia Designer Video Mr Jürgen Schmidt, Dipl.-Pflegw. (Technical University) Ms Margrit Catani and Ms Diana Schuler Dr. phil. nat. Banu Yürüker, MME Ms Sabine Richter, Technical Assistant Ms Gudrun Stopper, MME Mr Stephan Schallenberger, MAS in HCID Mr Kevin Gaunt, iOS Software Developer and Research Barbara Haldemann, Head of Gynaecological Operating Theatre at the Inselspital Bern Competing interests The authors declare that they have no competing interests. Supplementary Material Check list for surgical hand disinfection Questionnaire Table 1 Distribution of points Table 2 Process Schedule Table 3 Descriptive statistics of the observation dimensions separated by intervention group Table 4 Results of the T-test Table 5 Results of prior knowledge Table 6 Overall preference Table 7 Suitability of learning method for content ==== Refs 1 Balmelli C Eisenring MC Harbarth S Kuster S Marschall J Spicher VM Pittet D Ruef C Sax H Troillet N Widmer A Zanetti G Spitalinfektionen in der Schweiz: hochgerechnet rund 2000 Todesfälle pro Jahr 2013 Zürich Swissnoso Available from: http://www.jazumepg.ch/logicio/client/jazumepg/file/130823_spitalinfektionen_in_der_schweiz.pdf 2 Widmer A Troillet N Erfassung postoperativer Wundinfektionen: vom Prinzip zur Praxis1999 Swiss Noso 1999 6 1 1 8 Available from: http://www.swissnoso.ch/de/bulletin/articles/article/erfassung-postoperativer-wundinfektionen-vom-prinzip-zur-praxis 3 Cloyd J Holtzman D O'Sullivan P Sammann A Tendick F Ascher N Operating room assist: surgical mentorship and operating room experience for preclerkship medical students J Surg Educ 2008 65 4 275 282 10.1016/j.jsurg.2008.04.002 Available from: http://dx.doi.org/10.1016/j.jsurg.2008.04.002 18707660 4 Leeper K Stegall MS Stegall MH Basic aseptic technique for medical students: Identifying essential entry-level competencies Curr Surg 2002 59 1 69 73 10.1016/S0149-7944(01)00494-9 Available from: http://dx.doi.org/10.1016/S0149-7944(01)00494-9 16093107 5 Felder RM Soloman BA Learning Styles and Strategies 2005 Raleigh North Carolina State University Available from: http://www4.ncsu.edu/unity/lockers/users/f/felder/public/ILSdir/styles.htm 6 Dong C Goh PS Twelve tips for the effective use of videos in medical education Med Teach 2015 37 2 140 145 10.3109/0142159X.2014.943709 Available from: http://dx.doi.org/10.3109/0142159X.2014.943709 25110154 7 Aragon CE Zibrowski EM Does exposure to a procedural video enhance preclinical dental student performance in fixed prosthodontics? J Dent Educ 2008 72 1 67–71 18172237 8 Jang HW Kim KJ Use of online clinical videos for clinical skills training for medical students: benefits and challenges BMC Med Educ 2014 14 56 10.1186/1472-6920-14-56 Available from: http://dx.doi.org/10.1186/1472-6920-14-56 24650290 9 Yoder ME Transfer of cognitive learning to a clinical skill: linear versus interactive video West J Nurs Res 1993 15 1 115 117 10.1177/019394599301500109 Available from: http://dx.doi.org/10.1177/019394599301500109 8421915 10 Summers AN Rinehart GC Simpson D Redlich PN Acquisition of surgical skills: a randomized trial of didactic, videotape, and computer-based training Surg 1999 126 2 330 336 10.1016/S0039-6060(99)70173-X Available from: http://dx.doi.org/10.1016/S0039-6060(99)70173-X 11 Xeroulis GJ Park J Moulton CA Reznick RK Leblanc V Dubrowski A Teaching suturing and knot-tying skills to medical students: a randomized controlled study comparing computer-based video instruction and (concurrent and summary) expert feedback Surg 2007 141 4 442 449 10.1016/j.surg.2006.09.012 Available from: http://dx.doi.org/10.1016/j.surg.2006.09.012 12 Nousiainen M Brydges R Backstein D Dubrowski A Comparison of expert instruction and computer-based video training in teaching fundamental surgical skills to medical students Surg 2008 143 4 539 544 10.1016/j.surg.2007.10.022 Available from: http://dx.doi.org/10.1016/j.surg.2007.10.022 13 Shippey SH Chen TL Chou B Knoepp LR Bowen CW Handa VL Teaching subcuticular suturing to medical students: video versus expert instructor feedback J Surg Educ 2011 68 5 397 402 10.1016/j.jsurg.2011.04.006 Available from: http://dx.doi.org/10.1016/j.jsurg.2011.04.006 21821220 14 Cardoso AF Moreli L Braga FT Vasques CI Santos CB Carvalho EC Effect of a video on developing skills in undergraduate nursing students for the management of totally implantable central venous access ports Nurs Educ Today 2012 32 6 709 713 10.1016/j.nedt.2011.09.012 Available from: http://dx.doi.org/10.1016/j.nedt.2011.09.012 15 Wilkinson JM Van Leuven K Fundamentals of Nursing: Theory, Concepts, Applications, Vol 2 2007 Philadelphia FA Davis 16 Oldhafer K Jurs U Kramer A Martius J Weist K Mielke M Prävention postoperativer Infektionen im Operationsgebiet Bundesgesundheitsbl - Gesundheitsforsch - Gesundheitsschutz 2007 50 377 393 Available from: http://edoc.rki.de/documents/rki_ab/reKNpBgNk2ng/PDF/297Gug15s01A.pdf

PMC005xxxxxx/PMC5003146.txt

==== Front GMS J Med EducGMS J Med EducGMS J Med EducGMS Journal for Medical Education2366-5017German Medical Science GMS Publishing House zma00106210.3205/zma001062Doc63urn:nbn:de:0183-zma0010629ArticlePractical Clinical Training in Skills Labs: Theory and Practice Von Theorie und Praxis: Die Vermittlung klinisch-praktischer Fertigkeiten im Skills-Lab Bugaj T. J. 1Nikendei C. *11 University Hospital Heidelberg, Center for Psychosocial Medicine, Department of General Internal Medicine and Psychosomatics, Heidelberg, Germany*To whom correspondence should be addressed: C. Nikendei, University Hospital Heidelberg, Center for Psychosocial Medicine, Department of General Internal Medicine and Psychosomatics, Im Neuenheimer Feld, D-69120 Heidelberg, Germany, Phone: +49 (0)6221/56-38663, Fax: +49 (0)6221/56-5749, E-mail: christoph.nikendei@med.uni-heidelberg.de15 8 2016 2016 33 4 Clinical skillsDoc6328 9 2015 09 5 2016 15 2 2016 Copyright © 2016 Bugaj et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License.This article is available from http://www.egms.de/en/journals/zma/2016-33/zma001062.shtmlToday, skills laboratories or “skills labs”, i.e. specific practical skill training facilities, are a firmly established part of medical education offering the possibility of training clinical procedures in a safe and fault-forging environment prior to real life application at bedside or in the operating room. Skills lab training follows a structured teaching concept, takes place under supervision and in consideration of methodological-didactic concepts, ideally creating an atmosphere that allows the repeated, anxiety- and risk-free practice of targeted skills. In this selective literature review, the first section is devoted to (I) the development and dissemination of the skills lab concept. There follows (II) an outline of the underlying idea and (III) an analysis of key efficacy factors. Thereafter, (IV) the training method’s effectiveness and transference are illuminated, before (V) the use of student tutors, in the sense of peer-assisted-learning, in skills labs is discussed separately. Finally, (VI) the efficiency of the skills lab concept is analyzed, followed by an outlook on future developments and trends in the field of skills lab training. Zusammenfassung Skills-Labs (dt. „Fertigkeitenlabore“, also spezifische Übungsräume zum Erlernen praktischer Fertigkeiten) sind heute nicht mehr aus der medizinischen Ausbildungslandschaft wegzudenken, denn sie bieten die Möglichkeit, klinische Prozeduren noch vor der realen Anwendung am Krankenbett oder im Operationssaal in sicherer und geschützter Atmosphäre zu erproben. Der Unterricht im Skills-Lab erfolgt strukturiert, supervidiert und unter Berücksichtigung methodisch-didaktischer Rahmenaspekte, sodass im Idealfall das wiederholte, angst- und risikofreie Üben von Fertigkeiten gewährt wird. Die vorliegende selektive Literaturübersicht widmet sich in einem ersten Abschnitt der (I) Entstehung und Verbreitung des Skills-Lab-Konzepts. Es folgt (II) eine Darstellung der zugrundeliegenden Hintergrundidee sowie (III) eine Analyse der wichtigsten Wirksamkeitsfaktoren. Im Anschluss werden zunächst (IV) Effektivität und Transfer dieser Ausbildungsmethode beleuchtet, bevor (V) der Einsatz von studentischen Tutoren im Sinne des peer-assisted-learnings innerhalb des Skills-Labs gesondert darstellt wird. Im Anschluss erfolgt eine Analyse (VI) der Effizienz des Skills-Lab-Konzeptes, bevor in einem letzten Abschnitt ein Ausblick auf zukünftige Entwicklungen und Trends im Bereich der Skills-Labs gegeben wird. skills labskills lab trainingskillssimulationmedical education ==== Body Introduction In its broadest sense, the term “skills labs”, an abbreviation of skills laboratories, refers to specifically equipped practice rooms functioning as training facilities offering medical students, physicians in training and other medical staff alike a protected, fault-forgiving environment for the practice of clinical skills prior to their real life application. The present selective literature review aims to give the interested reader a brief summary of the scientific knowledge relevant to skills lab training in primary medical education and places the focus specifically on the former target group, namely, on medical students. In light of the fact that the majority of – often serious – mistakes come to effect in the first few working years after medical licensure and can often be put down to a lack of (practical) experience, this group appears to draw particular benefit from skills lab training for several reasons [1]. In addition, after years of mostly theoretically-oriented university education, medical students are only given a relatively short period of time to acquaint themselves with the practical clinical reality in the physician’s professional world. Ultimately, these promising young physicians have a particularly long period of responsibility ahead of them and will in turn be responsible for the training of future health worker generations [1]. Against this background, it seems crucial that clinical procedures are trained in a safe and fault-forgiving environment prior to real life application at bedside or in the operating room. Hence, skills labs play a key role in medical training quality assurance. Here, procedural skills can be trained, repeatedly practiced, and evaluated until the required minimum standard for patient treatment is ensured [2], [3]. Skills lab training provides medical students with the necessary basic skills for later clinical activity by the means of models, phantoms, and fellow students or with the help of standardized patients (SP) [4]. Skills lab training follows a structured teaching concept, takes place under supervision and in consideration of methodological-didactic concepts, ideally creating an atmosphere that allows the repeated, anxiety- and risk-free practice of targeted skills and ensures that all students are given the opportunity to perform these independently. In this selective literature review, the first section is devoted to the development and dissemination of the skills lab concept. There follows an outline of the underlying idea and (III) an analysis of key efficacy factors. Thereafter, the training method’s effectiveness and transfer are illuminated before the use of student tutors, in the sense of peer-assisted-learning, in skills labs is discussed separately. Finally, the efficiency of the skills lab concept is analyzed, followed by an outlook on future developments and trends in the field of skills lab training. Methods The present paper is based on a “narrative” or “selective” literature review, including author-specific expertise and teaching experience as well as a selective search of thematically relevant references in the database “PubMed”. The key reviews considered are summarized in Table 1 (Tab. 1). (I) Development and dissemination The first skills labs were established at the universities of Illinois and Maastricht in the 1970s with the aim of improving clinical practical skills [5], [6]. Whereon, the use of simulation-based teaching units has seen widespread development in medical education over the last 25 years [7]. Moreover, following the amendments to medical licensure laws in 2002, which allotted key importance to the training and evaluation of clinical skills at medical school, skills labs have been established nationwide and are part of medical education at all medical faculties in Germany [8], [9]. (II) The underlying idea Prior to the widespread introduction of skills labs, the acquisition of individual clinical practical skills, and hence successful medical education, was strongly dependent on “appropriate” patient encounters and well-trained lecturers [10]. Since access to these may be severely limited by availability and time problems [11], skills lab training provides a comparatively simple and almost always-available solution to the sketched dilemma. Furthermore, technical innovations in diagnosis and treatment, as well as large-scale patient safety initiatives have changed the demands on physicians’ skills over the past years, giving even further justification to skills lab training [12], [7]. For instance, technically sophisticated innovations, such as minimally invasive surgery, require the specialized training of psychomotor and spatial awareness skills. In a study from 2003, Ziv et al. point out that the use of simulation-based training in medical education, similar to its use in the fields of aviation or the military, represents, as it were, an ethical imperative and therefore see the acquisition of a basic routine prior to “real patient practice” as an indispensable step in medical training [13]. Hence, simulation-based medicine not only allows for the training of rare emergency measures but is, in itself, a crucial prerequisite to ensure good practice. Especially, the aspect of patient safety is of particular importance here: situations handled by medical students or young physicians with lacking adequate prior medical training have been shown to increase patient risk [14], [15]. Accordingly, in a study from 2006, Takayasu and colleagues were able to show that students particularly appreciated the aspect of experiential “practice without risk” of harming real patients in simulation-based training [16]. In addition to the provision of a fault-forgiving environment, professional feedback in particular is seen to make skills lab training a valuable experience for students. Accordingly, this important aspect and its influence on behavior are discussed separately below. (III) Key efficacy factors A Best Evidence Medical Education (BEME) collaboration review from 2005 [17] offers a list of key efficacy factors for simulation-based training in medical education, which also come to effect in skills lab training. For example, the importance of giving feedback directly during actual simulation-based training is emphasized. Furthermore, in line with the principle of “deliberate practice”, it is mentioned that repeated simulation-based practice is necessary in order to achieve best results. The reflections upon the required degree of realism, as in how close to real-life practice should and simulation training can come, also seem relevant in the establishment of skills lab training units. In addition to requiring coherent integration into the existing curriculum, simulation-based learning should take place in a “controlled” environment. Moreover, in accordance with the BEME guidelines, it is important that simulated skill performance outcome and assessment are measured in a standardized and clearly defined way. All mentioned points shall be discussed below in detail and the said BEME collaboration review key efficacy factors will be enhanced by some further important factors. Accordingly, a section on skills lab training preparation as well as a passage on standardization going beyond the simple definition of learning objectives by, for example, also considering the importance of unified teaching concepts and discussing appropriate assessment methods, can be found below. Finally, it should not go unmentioned that precisely the implementation of these key efficacy factors can often be held responsible for high costs, financial burden or problems related to skills labs operation or establishment. Setting The skills lab setting allows students to make mistakes and ideally to discover and correct these accordingly without having to fear adverse consequences for themselves or patients [17]. Accordingly, trainers also need not worry about real patients’ well-being during training and can thus concentrate on the skill development of individual trainees, for example, by highlighting “Teachable Moments” [15]. In addition to this “ideal environment” largely shaped by and attributable to the trainers attitude, there are also structural measures that make a skills lab a prosperous practice field and thus contribute to the establishment of an ideal learning setting in a wider sense: appropriately, todays skills labs often reflect the real-life medical working environment. This is achieved by, for example, furnishing the skills labs with identical or comparable equipment and facilities. Some skills labs also have adjoining observation rooms or installed cameras that allow the trainers to follow what is happening inside via a one-way mirror or monitor. Standardization In order to learn a practical skill, it is important that learners go through a standardized, trainer-led training program. In this initial phase, free practice time does not seem to be conducive [18]. A coherent didactic concept, such as standard instruction in accordance with Peyton (see below), can be seen as a valid standardized approach in this context. Furthermore, checklists have been shown to be conducive for the quality assurance of students' skills lab education [19] also including internal faculty standards that define examination procedures and principles [20], [21], [22]. Ultimately, the process of standardization provides the basis upon which competence-based assessment of student skills is made possible. Definition of learning objectives For the acquisition of procedural skills, learners generally have a larger chance of learning success if the learning objectives are clearly defined and the desired outcome has been clearly communicated in advance [17]. As a result, students are given the possibility to prepare themselves meticulously prior to skills lab training. However, the clear definition of learning objectives is also of essential interest to curriculum planning, as they can serve as blueprints for the appropriate definition of the necessary time frame under the careful consideration of the specific requirements and complexity of the particular activity and the learners’ previous experiences. The consensus statement “Practical Skills in Medical Studies” distinguishes three depth dimensions of learning objectives, which seem to be relevant for skills lab learning [23]: To have seen the implementation of the skill To have completed a skill itself several times under supervision To be able to perform a skill independently and routinely Training preparation In order to maximize the learning effect and to make optimal use of the often limited training periods, skills lab lessons must be specifically prepared in addition to their above-mentioned general integration into the curriculum. To this effect, students are frequently provided with preparatory texts in paper or electronic form in advance. However, other forms of student preparation have also been reported, such as the targeted analysis of virtual patients by the means of multimedia case studies [24]. Instruction During skills lab training, learners commonly practice the procedural skills’ psychomotor component under the trainers’ instruction, who have previously demonstrated the relevant skill. Subsequently, the skills are then performed by the learners themselves under supervision. Peyton’s Four-Step Approach has proven to be most helpful here. This systematic approach to learning practical skills has been increasingly used in the field of medical education and is also the routine approach in the European Society of Cardiology’s (ESC) practical training courses since 2000 [25], [26], [27], [28]. The Four-Step Approach consists of the following four clearly defined steps: The trainer demonstrates the skill in real time without giving instructions or explanatory words (“Demonstration”). The trainer repeats the procedure, this time describing all necessary sub-steps (“Deconstruction”). The trainer performs the skill for a third time, this time following the sub-steps only as described to him by the trainee (“Comprehension”). This step has been identified as the most important step of the Four-Step Approach in the past [29] as deeper processing mechanisms reflecting what was observed in the first two steps are necessary for the trainees’ to be able to give instructions. The trainee performs the skill on his/her own (“Performance”). Compared to the provision of simple instructions, Peyton’s approach has been shown to be more conducive, especially, in regard to observed professionalism and concomitant doctor-patient communication and also leads to the more rapid implementation of new activities [29], [30]. Though the procedure described above focuses a one-on-one learning situation, a modified version of the approach has also proven to be applicable for the instruction of small groups, typically found in skills lab training settings [31]. Deliberate Practice First described by K. Anders Ericsson in 1993, the idea of “Deliberate Practice” (DP) was initially derived from various studies with professional musicians, as the relationship between invested practice and achieved level of musical excellence presented itself particularly for empirically investigation [32]. Previously, it was often assumed that experts have innate skills not possessed by other people. However, according to Ericsson these skills are not innate or unchangeable but the result of lifelong and especially deliberate, as in systematic and goal-oriented, practice of an activity [32]. DP involves (a) repetitive practice of the intended skill, combined with (b) the thorough assessment of the skill so that the learner (c) can receive specific, informative feedback, which results in an increasingly (d) better performance of skill. So, according to Ericsson et al., the improved performance of an activity largely depends on how much time one spends actively practicing it – time alone does not suffice to achieve expert status. Even for skills of little complexity, repetitive practice seems very important [33] and is even indispensable for medium or highly-complex skills [34]. Hence, complex skills, for example in the field of anesthesia (plexus block, spinal or epidural anesthesia), have been demonstrated to require a significantly higher number of repetitions to achieve mastery [35]. DP therefore appears to play a major role in the learning of skills and thus is a key impact factor in skills lab training. However, it has been argued that fundamental basic skills could well be hereditary and that other factors, such as general cognitive abilities, are also conducive in achieving high skill performance [36], [37]. In order to ascertain the extent to which individual students learn using DP methods, Moulaert et al. conducted a survey with medical students. Subsequently, the authors compared the questionnaire study’s results with the students’ university performance. They were not only able to discover a relationship between DP and the academic performance but were also able to show that the high-performing students were more likely to practice DP learning than the lower performers [38]. In 2011, Duvivier et al. showed that students’ OSCE results improved with the increasing use of DP methods [39]. In turn, Griswold and colleagues were able to confirm that the DP method also improved practical skills increasing patient safety consecutively [14]. Moreover, McGhie even reported a strong dose-response relationship: the number of hours subjects spent practicing with high-fidelity simulators correlated with standardized learning outcomes [40]. Feedback Issenberg et al. list feedback, that is the knowledge of the personal performance in the carried out simulation, as the most important factor for effective simulation-based learning [17], which can either be generated by the simulator itself (for example “high-fidelity” simulators, as used in the fields of anesthesia and emergency medicine sporting a high degree of realism, are able to simulate a variety of vital signs) or be given by the trainer or other trainees in the vicinity in “real time”. In addition, post hoc feedback by viewing a recorded video footage of the performed skill also represents a possible feedback method [17]. A distinction is generally made between formative and summative feedback. In the context of simulation-based medical training, formative feedback is of higher priority, as it is usually not a question of “pass” or “fail” but of the trainees’ skills improvement [41]. Feedback helps trainees to reflect upon their own actions, to recognize possible errors and to subjectively evaluate and observe their learning progress [17]. Just recently, Bosse et al. were able to show that higher-frequency intermittent feedback is superior to low-frequency feedback for the acquisition of procedural skills during skills training [42]. However, other research has indicated that too intensive feedback can be inconducive, especially in the early stages of learning processes [43]. Fidelity In regard to the effectiveness of skills lab training, the importance of the degree of the simulations’ realism has been controversially discussed: while some studies have found evidence that the degree of realism achieved in a simulation positively impacts on learning effect [17], [44], others argue that the learners‘ motivation to engage in the simulation and to complement the fictitious situation by augmenting it with personal real-life experiences are further key factors in addition to the simulation’s fidelity [45]. Accordingly, other studies have shown no or only minimal training outcome advantages for high-fidelity simulation settings, hardly justifying the ensued considerable extra costs [46], [47], [48]. However, in order to integrate the implementation of procedural skills into a well-functioning role play, the provision of role assignments, role-specific instructions and case vignettes is crucial in skills lab training [19], [49]. A controlled study was able to show that the introduction of role play into skills lab training resulted in both a more realistic perception of the medical role, as well as in more intensive and improved communication accompanying the procedures [19]. Today, Miller’s Learning Pyramid is an often-cited guide to describe different levels of knowledge and skills acquisition [50] which distinguishes four levels of competence or training objectives: Knows Knows how Shows how Does Accordingly, the first two levels describe cognitive aspects, i.e. the acquisition of factual knowledge (level 1) and the application knowledge (level 2). The third and fourth levels refer to procedural skills. Skills lab training takes place on Miller’s Pyramid’s 3rd competence level, in detail on the level “shows how”/“show as if”. The acquisition of procedural skills in turn can take place in simulated environments with a varying degree of realism. In a publication from 2013, Russo and Nickel stress (49) that there is no such thing as an ideal degree of realism for simulations in medical training but that, depending on the training’s focus and simulators, suitable conditions for the rehearsal of specific action or movement sequences can be achieved with a relatively low degree of fidelity. However, the authors of this paper believe that even a good simulator, be it in combination with other factors aiming to further increase its degree of realism, such as accompanying role play, fails to achieve Miller’s fourth level of competence (“Does”) as no simulation can fully reflect and thus endeavor to replace real-life practical action. Curriculum integration In 2011, Parmar et al. reported that ideal skills lab training is most likely to be dependent on a combination of the provision of optimum curriculum in conjunction with a suitable phantom [51]. The above-mentioned BEME Collaboration overview from 2005 [17] also implies that simulation-based learning should not be seen as an optional offer but as an integral (and standardized) part of medical education, particularly as optional services, such as elective subjects, usually arouse much less interest among students [17]. Furthermore, in a study from 2012, Weller and colleagues come to the conclusion that simulated practice should necessarily be in temporal association with practical teaching at bedside allowing for the direct transference and consolidation of learned skills in practice [15]. The more complex the simulation, the greater the risk of frustrating or discouraging students, especially when practical skills lab training is not embedded in a coherent overall curriculum or adapted to the level of training [52]. Hence, simulation-based training should be deemed to be of equivalent importance as bedside teaching, lectures and/or problem-oriented learning, which is a challenge for curriculum development [41]. Already in 2009, Kneebone had indicated that the acquisition of practical skills should not be completely separated from the clinical context and that an excessive simplification of highly complex activities could lead to a superficial understanding in learners [53]. Therefore, it seems all the more important to find the right balance between simulation-based training and clinical reality in a commensurate curriculum conducive to learning. Linking the skills lab training with practical clinical examinations Depending on the focus, examinations represent a specific learning incentive to acquire theoretical and practical knowledge or manually-practical skills. The phenomenon known as “assessment drives learning” [54], [55] can be harnessed by creating congruence between the learning and examination format within the respective competence levels (knows, knows how; shows how; does; cf. Miller’s Pyramid [50]). Such “constructive alignment” [56] is achieved, for example, if the set-up of a peripheral IV catheter is practiced in skills lab training (level “shows how”) and the degree of competence acquired is then evaluated via a practical, not theoretical, examination on the same level of competency [57]: hence, the examination format adequately corresponds with the learning level. It has been shown that skills lab training learning content relevant for OSCE examinations (objective structured clinical examination) was continuously and increasingly practiced during the semester itself, whereas training opportunities without relevance to assessment were no longer made use of towards the end the semester [58]. (IV) Effectiveness and transfer in skills labs training Given the increase in simulation-based training opportunities in medical education, the question as to the effectiveness and transference of practiced skills in skills lab training arises. In 1999, Remmen et al. [59] showed that skills labs training increased the number of independently performed skills by medical students in everyday clinical practice, although the validity of the results is weakened by the small participant number. Also, the quality of the underlying skills was not considered. In 2005 in a prospective study, Jünger et al. [60] studied the impact of basic practical internal medicine skill training (incl. seven units of skills lab training lasting 90 min. each) on students‘ OSCE examination performance. The intervention group performed significantly better in the OSCE examination than the control group, which had only received classic bedside training. In 2012, Lund et al. were able to show that skills lab training not only led to an improved examination performance in practical clinical trials but also to an objective and significantly higher rated transfer capacity for clinical skills acquired in skills lab training (equivalent to the highest level of Kirkpatrick’s hierarchy for the evaluation of training effects [61]) via the example of the set-up of a plastic cannula [62]. In addition, several reviews and meta-analyzes on the effectiveness of skills lab training exist: in a review from 2007, Lynagh et al. [63], examined transference to the clinical reality in addition to the question of the effectiveness of skills lab training. Unfortunately, only 20 of the 44 studies included in the review considered the transfer of procedural skills. In turn, eight of the 20 studies used animal carcasses or live animals instead of real patients. Therefore, these studies are, strictly speaking, not able to answer the question of transfer to reality but rather address the question of transference of low fidelity simulations to high-fidelity simulations. However, of the twelve studies that actually examined the transference of procedural skills to clinical reality, eleven studies found that participants who had received a simulation-based training intervention significantly improved in performance compared to the control group. The research group led by Cook and colleagues contributed several important meta-analysis to assess the effectiveness simulation-based training in medicine: in 2011, Cook et al. [64] were able to show that simulation-based interventions were superior to no intervention settings, stressing that no further primary studies of this kind would be necessary as the large effect sizes in the areas of knowledge, attitude and skills and moderate effects on patient care outcome had been clearly proven for simulation-based training. In 2012, in a further meta-analysis, Cook and colleagues examined the superiority of simulation-based training in medicine over other instruction methods and could substantiate small to moderate effects over training forms which had dispensed simulation use [65]. In a meta-analysis in 2011, McGhie et al. [66] also investigated the effect of simulation-based training in medicine using Deliberate Practice (see above) compared to traditional teaching methods and were able to demonstrate its superiority, although they, similar to Cook et al., noted the relative lack of high-quality studies on the effectiveness of simulation-based teaching in medicine. In a randomized controlled trial in 2013, Hermann-Werner and colleagues were able to show that skills lab training learning effects also lasted longer than it seems to be the case with traditional teaching methods [67]. In this study, participants who received skills lab training performed the observed activities (insertion of a peripheral IV catheter and a gastric tube on a model) significantly better as compared to students who had previously only received training in a “see one – do one” intervention. However, this observation was not only limited to immediately after exercise; in follow-ups three or six months later, skills lab intervention group students required less time to carry out the skills and were also significantly more likely to be rated as clinically competent as compared to students that had been traditionally instructed. (V) Skills lab training and peer-assisted learning Skills lab training must not necessarily be led by medical faculty in order to yield long-term success. Moreover, implementing training by specially trained student tutors is quite widespread: according to a recent study published by Blohm et al. the principle of the PAL (peer-assisted learning) is used in about 90% of the skills labs of the surveyed German faculties. This could be down to the fact that several studies have been able to show that PAL is equivalent to medical faculty instructed training [68], [69], [70] as it allows for “eye level learning” [58], [71]. Depending on the focus, students’ cognitive, psychomotor and affective development has been proven to be supported by PAL [72]. Social and cognitive congruence between tutors and students have often been listed as key PAL efficiency factors [73], [74]. The construct of social congruence means that student tutors and students are able to communicate informally and in a particularly empathic manner simply because of their similar social roles [75]. In addition, student tutors and students often share a similar knowledge base and comparable learning experiences, commonly termed “cognitive congruence”. Accordingly, both speak the same “language” facilitating the delivery of explanations at an appropriate level [71]. (VI) Skills-Labs training efficiency Regardless of the above-discussed question of effectiveness, the question of this measure’s efficiency must inevitably arise. The obvious question of whether the – in some cases very high – effort and resource investment required to achieve the objective is justifiable is at the heart of this deliberation. Accordingly, the methods high cost is also the most frequently mentioned criticism of the concept of simulation-based training in medicine [76]. The exact cost of skills lab training is not easy to elicit, as costs reports are often irregular and / or incomplete according to Zendejas et al. observations [77]. The same meta-analysis from 2013 shows that from a total of 967 identified comparative studies only a fraction (6.1%) even mention the cost of the underlying intervention, with a median of two reported cost factors, although the meta-analysis' authors identify 21 relevant factors. These figures clearly underline significant deficiencies in the report of simulation costs. Only 1.6% of all studies identified by Zendejas et al. compared the cost of the simulation to those of different training forms. The authors of the meta-analysis list expenditure on simulators and accessories as the most commonly cited cost factors, followed by spending on consumables and simulation unit maintenance fees. However, the costs for room rental, furniture and employed tutors' wages are often overseen. Regardless of the exact cost, it is a already well-established fact that medical education is a particularly expensive field of education [78], which may be largely attributable to practical training. Hence, skills labs training should never be implemented as a mere end in itself but should always be subjected to the strict scrutiny of whether the intended learning objective can be reached by the means of other teaching methods, or if the use of simulators offers additional benefit, such as ensuring patient safety. Accordingly, deliberations on cost efficiency must consider that the renunciation of simulation-based training and the return to exclusive training "at bedside" may potentially ensue much higher costs due to the potential harm to patients [79]. Regardless of the obvious and aforementioned ethical concerns, several authors have also noted that the use of real patients for training is by no means inexpensive [80], [81], [82]: unlike in simulation-based training, materials, such as a catheter, rendered unsterile by careless actions can not be reused, therefore, increased material costs must be anticipated. Moreover, without prior simulator-based training, students require more time to carry out skills, which can cause problems even ensuing more costs in the usually tight schedules of clinics. Viewed differently, by offloading time-consuming teaching content from the clinical routine, simulations can also be seen to provide a possibility of saving cost not to be underestimated. The application of a skill in a real-life setting by a learner often ensues a higher level of iatrogenic injuries, which, in turn, require costly treatment. Outlook In 2002, the amendment of the medical licensing regulations already launched a number of developments in national medical education towards the more focused practical training of future physicians [https://www.gesetze-im-internet.de/_appro_2002/BJNR240500002.html]. Today, a further shift in curriculum content in favor of communicative and practical clinical skills training is highly probable in light of the introduction of the “National Competency-based Catalogue of Learning Objectives for Medical Education” [83] and the future “National Longitudinal Core Curriculum of Communications in Medicine”. In addition, the “Master Plan 2020” lists further measures promoting practical relevance in medical education [https://www.bundesgesundheitsministerium.de/ministerium/meldungen/2015/masterplan-medizinstudium-2020.html]. Consequently, medical education will continue to avert from the outdated model of “see-one, do-one, teach one” and, in all probability, skills lab training will increasingly advance to be a worldwide norm. In addition to further developments in simulation-based learning and training, an increase in methodologically convincing, randomized-controlled, double-blind studies investigating the effectiveness and transference of the procedural skills learned in skills lab in clinical practice as well as the associated costs, can be expected. Nevertheless, these conducive developments in the field of medical education, should never let us forget that simulation-based training always remains an approximation of real-life practice and can never replace real-life patient practice. Hence, even the most modern skills labs and the expected innovations in simulation-based medical education must always be seen as an essential complement – but can never be understood as a complete replacement of quality teaching at bedside and of supervision by experienced lecturers. Notwithstanding the foregoing, medical education today is no longer conceivable without the numerous opportunities and benefits of simulation-based training in skills labs only hinted at in this present work. Competing interests The authors declare that they have no competing interests. Table 1 Important skills-lab training reviews in medical education. ==== Refs 1 Datta R Upadhyay K Jaideep C Simulation and its role in medical education Med J Armed Forces India 2012 68 2 167 172 10.1016/S0377-1237(12)60040-9 Available from: http://dx.doi.org/10.1016/S0377-1237(12)60040-9 24623932 2 Du Boulay C Medway C The clinical skills resource: a review of current practice Med Educ 1999 33 3 185 191 10.1046/j.1365-2923.1999.00384.x Available from: http://dx.doi.org/10.1046/j.1365-2923.1999.00384.x 10211238 3 Bradley P Postlethwaite K Setting up a clinical skills learning facility Med Educ 2003 37 Suppl 1 6 13 10.1046/j.1365-2923.37.s1.11.x Available from: http://dx.doi.org/10.1046/j.1365-2923.37.s1.11.x 14641633 4 Barrows HS An overview of the uses of standardized patients for teaching and evaluating clinical skills. AAMC Acad Med 1993 68 6 443 451 10.1097/00001888-199306000-00002 Available from: http://dx.doi.org/10.1097/00001888-199306000-00002 8507309 5 Sajid A Lipson LF Telder V A simulation laboratory for medical education J Med Educ 1975 50 10 970 975 10.1097/00001888-197510000-00007 Available from: http://dx.doi.org/10.1097/00001888-197510000-00007 1099207 6 Van Dalen J Bartholomeus P Bender W Hiemstra R Scherpbier A Zwiestra R Training clinical competence in a skills laboratory Teaching and Assessing Clinical Competence 1990 Groningen Stichting TICTAC 7 Issenberg SB Scalese RJ Simulation in health care education Perspect Bio Med 2008 51 1 31 46 10.1353/pbm.2008.0004 Available from: http://dx.doi.org/10.1353/pbm.2008.0004 18192764 8 Kruppa E Junger J Nikendei C Innovative teaching and examination methods--taking stock at German medical faculties Dtsch Med Wochenschr 2009 134 8 371 372 19206056 9 Blohm M Lauter J Branchereau S Krautter M Kohl-Hackert N Jünger J Herzog W Nikendei C "Peer-assisted learning" (PAL) in the Skills-Lab--an inventory at the medical faculties of the Federal Republic of Germany GMS Z Med Ausbild 2015 32 1 Doc10 10.3205/zma000952 Available from: http://dx.doi.org/10.3205/zma000952 25699102 10 Scalese RJ Obeso VT Issenberg SB Simulation technology for skills training and competency assessment in medical education J Gen Int Med 2008 23 Suppl 1 46 49 10.1007/s11606-007-0283-4 Available from: http://dx.doi.org/10.1007/s11606-007-0283-4 11 Gordon JA Pawlowski J Education on-demand: the development of a simulator-based medical education service Acad Med 2002 77 7 751 752 10.1097/00001888-200207000-00042 Available from: http://dx.doi.org/10.1097/00001888-200207000-00042 12114174 12 Maran NJ Glavin RJ Low- to high-fidelity simulation - a continuum of medical education? Med Educ 2003 37 Suppl 1 22 28 10.1046/j.1365-2923.37.s1.9.x Available from: http://dx.doi.org/10.1046/j.1365-2923.37.s1.9.x 14641635 13 Ziv A Wolpe PR Small SD Glick S Simulation-based medical education: an ethical imperative Acad Med 2003 78 8 783 788 10.1097/00001888-200308000-00006 Available from: http://dx.doi.org/10.1097/00001888-200308000-00006 12915366 14 Griswold S Ponnuru S Nishisaki A Szyld D Davenport M Deutsch ES Nadkarni V The emerging role of simulation education to achieve patient safety: translating deliberate practice and debriefing to save lives Pediatr Clin North Am 2012 59 6 1329 1340 10.1016/j.pcl.2012.09.004 Available from: http://dx.doi.org/10.1016/j.pcl.2012.09.004 23116529 15 Weller JM Nestel D Marshall SD Brooks PM Conn JJ Simulation in clinical teaching and learning Med J Aust 2012 196 9 594 10.5694/mja10.11474 Available from: http://dx.doi.org/10.5694/mja10.11474 22621154 16 Takayesu JK Farrell SE Evans AJ Sullivan JE Pawlowski JB Gordon JA How do clinical clerkship students experience simulator-based teaching? A qualitative analysis Sim Healthcare 2006 1 4 215 219 10.1097/01.SIH.0000245787.40980.89 Available from: http://dx.doi.org/10.1097/01.SIH.0000245787.40980.89 17 Issenberg SB McGaghie WC Petrusa ER Lee Gordon D Scalese RJ Features and uses of high-fidelity medical simulations that lead to effective learning: a BEME systematic review Med Teach 2005 27 1 10 28 10.1080/01421590500046924 Available from: http://dx.doi.org/10.1080/01421590500046924 16147767 18 Fichtner A Pierre M Breuer G Lernen für die Praxis: Das Skills-Lab Simulation in der Medizin 2013 Heidelberg Springer Medizin 10.1007/978-3-642-29436-5_10 Available from: http://dx.doi.org/10.1007/978-3-642-29436-5_10http://dx.doi.org/10.1007/978-3-642-29436-5_10 19 Nikendei C Zeuch A Dieckmann P Roth C Schafer S Volkl M Schellberg D Herzog W Jünger J Role-playing for more realistic technical skills training Med Teach 2005 27 2 122 126 10.1080/01421590400019484 Available from: http://dx.doi.org/10.1080/01421590400019484 16019330 20 Pjontek R Scheibe F Tabatabai J Heidelberger Standarduntersuchung 2013 Heidelberg Medizinische Fakultät Heidelberg 21 Nikendei C Kadmon M Heidelberger Standardprozeduren 2015 1st ed. Heidelberg Medizinische Fakultät Heidelberg 22 Nikendei C Ganschow P Groener JB Huwendiek S Köchel A Köhl-Hackert N Pjontek R Rodrian J Scheibe F Stadler AK Steiner T Stiepak J Tabatabai J Utz A Kadmon M "Heidelberg standard examination" and "Heidelberg standard procedures" - Development of faculty-wide standards for physical examination techniques and clinical procedures in undergraduate medical education. GMS J Med Educ 2016 33 4 Doc54 10.3205/zma001053 Available from: http://dx.doi.org/10.3205/zma001053 23 Schnabel KP Boldt PD Breuer G Fichtner A Karsten G Kujumdshiev S Schmidts M Stosch C A consensus statement on practical skills in medical school - a position paper by the GMA Committee on Practical Skills GMS Z Med Ausbild 2011 28 4 Doc58 10.3205/zma000770 Available from: http://dx.doi.org/10.3205/zma000770 22205916 24 Lehmann R Bosse HM Huwendiek S Blended learning using virtual patients and skills laboratory training Med Educ 2010 44 5 521 522 10.1111/j.1365-2923.2010.03653.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2010.03653.x 20374472 25 Bullock I Skill acquisition in resuscitation Resuscitation 2000 45 2 139 143 10.1016/S0300-9572(00)00171-4 Available from: http://dx.doi.org/10.1016/S0300-9572(00)00171-4 10950322 26 Bullock I Davis M Lockey A Mackway-Jones K Pocket Guide to Teaching for Clinical Instructors 2015 Weinheim Wiley 10.1002/9781119088769 Available from: http://dx.doi.org/10.1002/9781119088769 27 American Heart Association R. C. Instructor's manual: advanced cardiovascular life support 2001 Dallas American Heart Association 28 Nolan J Baskett P Gabbott D Advanced life support course provider manual 2001 London Resuscitation Council (UK) 29 Krautter M Weyrich P Schultz JH Buss SJ Maatouk I Junger J Nikendei C Effects of Peyton's four-step approach on objective performance measures in technical skills training: a controlled trial Teach Learn Med 2011 23 3 244 250 10.1080/10401334.2011.586917 Available from: http://dx.doi.org/10.1080/10401334.2011.586917 21745059 30 Krautter M Dittrich R Safi A Krautter J Maatouk I Moeltner A Nikendei C Peyton's four-step approach: differential effects of single instructional steps on procedural and memory performance - a clarification study Adv Med Educ Pract 2015 6 399 406 10.2147/AMEP.S81923 Available from: http://dx.doi.org/10.2147/AMEP.S81923 26060417 31 Nikendei C Huber J Stiepak J Huhn D Lauter J Herzog W Jünger J Krautter M Modification of Peyton's four-step approach for small group teaching - a descriptive study BMC Med Educ 2014 14 68 10.1186/1472-6920-14-68 Available from: http://dx.doi.org/10.1186/1472-6920-14-68 24690457 32 Ericsson KA Krampe RT Tesch-Römer C The role of deliberate practice in the acquisition of expert performance Psychol Rev 1993 100 3 363 10.1037/0033-295X.100.3.363 Available from: http://dx.doi.org/10.1037/0033-295X.100.3.363 33 Lammers RL Learning and retention rates after training in posterior epistaxis management Acad Med 2008 15 11 1181 1189 10.1111/j.1553-2712.2008.00220.x Available from: http://dx.doi.org/10.1111/j.1553-2712.2008.00220.x 34 Ericsson KA An expert-performance perspective of research on medical expertise: the study of clinical performance Med Educ 2007 41 12 1124 1130 10.1111/j.1365-2923.2007.02946.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2007.02946.x 18045365 35 Konrad C Schupfer G Wietlisbach M Gerber H Learning manual skills in anesthesiology: Is there a recommended number of cases for anesthetic procedures? Anesth Analgesia 1998 86 3 635 639 10.1097/00132586-199810000-00057 Available from: http://dx.doi.org/10.1097/00132586-199810000-00057 36 Hambrick DZ Meinz EJ Limits on the predictive power of domain-specific experience and knowledge in skilled performance Curr Direct Psychol Sci 2011 20 5 275 279 10.1177/0963721411422061 Available from: http://dx.doi.org/10.1177/0963721411422061 37 Campitelli G Gobet F Deliberate Practice Necessary But Not Sufficient Curr Direct Psychol Sci 2011 20 5 280 285 10.1177/0963721411421922 Available from: http://dx.doi.org/10.1177/0963721411421922 38 Moulaert V Verwijnen MG Rikers R Scherpbier AJ The effects of deliberate practice in undergraduate medical education Med Educ 2004 38 10 1044 1052 10.1111/j.1365-2929.2004.01954.x Available from: http://dx.doi.org/10.1111/j.1365-2929.2004.01954.x 15461649 39 Duvivier RJ van Dalen J Muijtjens AM Moulaert VR van der Vleuten CP Scherpbier AJ The role of deliberate practice in the acquisition of clinical skills BMC Med Educ 2011 11 101 10.1186/1472-6920-11-101 Available from: http://dx.doi.org/10.1186/1472-6920-11-101 22141427 40 McGaghie WC Issenberg SB Petrusa ER Scalese RJ Effect of practice on standardised learning outcomes in simulation-based medical education Med Educ 2006 40 8 792 797 10.1111/j.1365-2929.2006.02528.x Available from: http://dx.doi.org/10.1111/j.1365-2929.2006.02528.x 16869926 41 McGaghie WC Issenberg SB Petrusa ER Scalese RJ A critical review of simulation-based medical education research: 2003-2009 Med Educ 2010 44 1 50 63 10.1111/j.1365-2923.2009.03547.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2009.03547.x 20078756 42 Bosse HM Mohr J Buss B Krautter M Weyrich P Herzog W Jünger J Nikendei C The benefit of repetitive skills training and frequency of expert feedback in the early acquisition of procedural skills BMC Med Educ 2015 15 22 10.1186/s12909-015-0286-5 Available from: http://dx.doi.org/10.1186/s12909-015-0286-5 25889459 43 Stefanidis D Korndorffer JR Jr Heniford BT Scott DJ Limited feedback and video tutorials optimize learning and resource utilization during laparoscopic simulator training Surgery 2007 142 2 202 206 10.1016/j.surg.2007.03.009 Available from: http://dx.doi.org/10.1016/j.surg.2007.03.009 17689686 44 Brydges R Carnahan H Rose D Rose L Dubrowski A Coordinating progressive levels of simulation fidelity to maximize educational benefit Acad Med 2010 85 5 806 812 10.1097/ACM.0b013e3181d7aabd Available from: http://dx.doi.org/10.1097/ACM.0b013e3181d7aabd 20520031 45 Dieckmann P "Ein bisschen wirkliche Echtheit simulieren": über Simulatorsettings in der Anästhesiologie 2005 Oldenburg Universität Oldenburg 46 De Giovanni D Roberts T Norman G Relative effectiveness of high- versus low-fidelity simulation in learning heart sounds Med Educ 2009 43 7 661 668 10.1111/j.1365-2923.2009.03398.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2009.03398.x 19573189 47 Grober ED Hamstra SJ Wanzel KR Reznick RK Matsumoto ED Sidhu RS Jarvi KA The educational impact of bench model fidelity on the acquisition of technical skill: the use of clinically relevant outcome measures Ann Surg 2004 240 2 374 381 10.1097/01.sla.0000133346.07434.30 Available from: http://dx.doi.org/10.1097/01.sla.0000133346.07434.30 15273564 48 Matsumoto ED Hamstra SJ Radomski SB Cusimano MD The effect of bench model fidelity on endourological skills: a randomized controlled study J Urology 2002 167 3 1243 1247 10.1016/S0022-5347(05)65274-3 Available from: http://dx.doi.org/10.1016/S0022-5347(05)65274-3 49 Dieckmann P Rall M Eich C Schnabel K Junger J Nikendei C Role playing as an essential element of simulation procedures in medicine Z Evid Fortbild Qual Gesundheitswes 2008 102 10 642 647 10.1016/j.zefq.2008.11.020 Available from: http://dx.doi.org/10.1016/j.zefq.2008.11.020 50 Miller GE The assessment of clinical skills/competence/performance Acad Med 1990 65 9 Suppl S63 S67 10.1097/00001888-199009000-00045 Available from: http://dx.doi.org/10.1097/00001888-199009000-00045 2400509 51 Parmar S Delaney CP The role of proximate feedback in skills training Surgeon 2011 9 Suppl 1 S26 S27 10.1016/j.surge.2010.11.006 Available from: http://dx.doi.org/10.1016/j.surge.2010.11.006 21549989 52 Russo SG Nickel EA Pierre M Breuer G Wie im wahren Leben: Simulation und Realitätsnähe Simulation in der Medizin 2013 Heidelberg Springer Medizin 10.1007/978-3-642-29436-5_12 Available from: http://dx.doi.org/10.1007/978-3-642-29436-5_12http://dx.doi.org/10.1007/978-3-642-29436-5_12 53 Kneebone R Perspective: Simulation and transformational change: the paradox of expertise Acad Med 2009 84 7 954 957 10.1097/ACM.0b013e3181a843d6 Available from: http://dx.doi.org/10.1097/ACM.0b013e3181a843d6 19550196 54 Muijtjens AM Hoogenboom RJ Verwijnen GM Van Der Vleuten CP Relative or Absolute Standards in Assessing Medical Knowledge Using Progress Tests Adv Health Sci Educ 1998 3 2 81 87 10.1023/A:1009728423412 Available from: http://dx.doi.org/10.1023/A:1009728423412 55 McLachlan JC The relationship between assessment and learning Med Educ 2006 40 8 716 717 10.1111/j.1365-2929.2006.02518.x Available from: http://dx.doi.org/10.1111/j.1365-2929.2006.02518.x 16869912 56 Biggs J What the student does: teaching for enhanced learning High Educ Res Develop 1999 18 1 57 75 10.1080/0729436990180105 Available from: http://dx.doi.org/10.1080/0729436990180105 57 Wass V Van der Vleuten C Shatzer J Jones R Assessment of clinical competence Lancet 2001 357 9260 945 949 10.1016/S0140-6736(00)04221-5 Available from: http://dx.doi.org/10.1016/S0140-6736(00)04221-5 11289364 58 Buss B Krautter M Moltner A Weyrich P Werner A Junger J Nikendei C Can the 'assessment drives learning' effect be detected in clinical skills training?--implications for curriculum design and resource planning GMS Z Med Ausbild 2012 29 5 Doc70 10.3205/zma000840 Available from: http://dx.doi.org/10.3205/zma000840 23255965 59 Remmen R Derese A Scherpbier A Denekens J Hermann I van der Vleuten C Van Royen P Bossaert L Can medical schools rely on clerkships to train students in basic clinical skills? Med Educ 1999 33 8 600 605 10.1046/j.1365-2923.1999.00467.x Available from: http://dx.doi.org/10.1046/j.1365-2923.1999.00467.x 10447847 60 Junger J Schafer S Roth C Schellberg D Friedman Ben-David M Nikendei C Effects of basic clinical skills training on objective structured clinical examination performance Med Educ 2005 39 10 1015 1020 10.1111/j.1365-2929.2005.02266.x Available from: http://dx.doi.org/10.1111/j.1365-2929.2005.02266.x 16178828 61 Kirkpatrick DL Craig RL Bittel LR Evaluation of training Training and Development Handbook 1967 New York McGraw Hill 87 112 62 Lund F Schultz JH Maatouk I Krautter M Moltner A Werner A Weyrich P Jünger J Nikendei C Effectiveness of IV cannulation skills laboratory training and its transfer into clinical practice: a randomized, controlled trial PloS One 2012 7 3 e32831 10.1371/journal.pone.0032831 Available from: http://dx.doi.org/10.1371/journal.pone.0032831 22427895 63 Lynagh M Burton R Sanson-Fisher R A systematic review of medical skills laboratory training: where to from here? Med Educ 2007 41 9 879 887 10.1111/j.1365-2923.2007.02821.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2007.02821.x 17696985 64 Cook DA Hatala R Brydges R Zendejas B Szostek JH Wang AT Erwin PJ Hamstra SJ Technology-enhanced simulation for health professions education: a systematic review and meta-analysis Jama 2011 306 9 978 988 10.1001/jama.2011.1234 Available from: http://dx.doi.org/10.1001/jama.2011.1234 21900138 65 Cook DA Brydges R Hamstra SJ Zendejas B Szostek JH Wang AT Erwin PJ Hatala R Comparative effectiveness of technology-enhanced simulation versus other instructional methods: a systematic review and meta-analysis Simul Healthc 2012 7 5 308 320 10.1097/SIH.0b013e3182614f95 Available from: http://dx.doi.org/10.1097/SIH.0b013e3182614f95 23032751 66 McGaghie WC Issenberg SB Cohen ER Barsuk JH Wayne DB Does simulation-based medical education with deliberate practice yield better results than traditional clinical education? A meta-analytic comparative review of the evidence Acad Med 2011 86 6 706 711 10.1097/ACM.0b013e318217e119 Available from: http://dx.doi.org/10.1097/ACM.0b013e318217e119 21512370 67 Herrmann-Werner A Nikendei C Keifenheim K Bosse HM Lund F Wagner R Celebi N Zipfel S Weyrich P "Best practice" skills lab training vs. a "see one, do one" approach in undergraduate medical education: an RCT on students' long-term ability to perform procedural clinical skills PloS One 2013 8 9 e76354 10.1371/journal.pone.0076354 Available from: http://dx.doi.org/10.1371/journal.pone.0076354 24086732 68 Weyrich P Celebi N Schrauth M Moltner A Lammerding-Koppel M Nikendei C Peer-assisted versus faculty staff-led skills laboratory training: a randomised controlled trial Med Educ 2009 43 2 113 120 10.1111/j.1365-2923.2008.03252.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2008.03252.x 19161480 69 Hudson JN Tonkin AL Clinical skills education: outcomes of relationships between junior medical students, senior peers and simulated patients Med Educ 2008 42 9 901 908 10.1111/j.1365-2923.2008.03107.x Available from: http://dx.doi.org/10.1111/j.1365-2923.2008.03107.x 18694405 70 Tolsgaard MG Gustafsson A Rasmussen MB Hoiby P Muller CG Ringsted C Student teachers can be as good as associate professors in teaching clinical skills Med Teach 2007 29 6 553 557 10.1080/01421590701682550 Available from: http://dx.doi.org/10.1080/01421590701682550 17978968 71 Yu TC Wilson NC Singh PP Lemanu DP Hawken SJ Hill AG Medical students-as-teachers: a systematic review of peer-assisted teaching during medical school Adv Med Educ Pract 2011 2 157 172 23745087 72 Secomb J A systematic review of peer teaching and learning in clinical education J Clin Nurs 2008 17 6 703 716 10.1111/j.1365-2702.2007.01954.x Available from: http://dx.doi.org/10.1111/j.1365-2702.2007.01954.x 18047577 73 Ten Cate O Durning S Dimensions and psychology of peer teaching in medical education Med Teach 2007 29 6 546 552 10.1080/01421590701583816 Available from: http://dx.doi.org/10.1080/01421590701583816 17978967 74 Lockspeiser TM O'Sullivan P Teherani A Muller J Understanding the experience of being taught by peers: the value of social and cognitive congruence Adv Health Sci Educ 2008 13 3 361 372 10.1007/s10459-006-9049-8 Available from: http://dx.doi.org/10.1007/s10459-006-9049-8 75 Schmidt HG Moust JH What makes a tutor effective? A structural-equations modeling approach to learning in problem-based curricula Acad Med 1995 70 8 708 714 10.1097/00001888-199508000-00015 Available from: http://dx.doi.org/10.1097/00001888-199508000-00015 7646747 76 Ker J Hogg G N M Walsh K Cost effective simulation Cost Effectiveness in Medical Education 2010 Abingdon Radcliffe 77 Zendejas B Wang AT Brydges R Hamstra SJ Cook DA Cost: the missing outcome in simulation-based medical education research: a systematic review Surgery 2013 153 2 160 176 10.1016/j.surg.2012.06.025 Available from: http://dx.doi.org/10.1016/j.surg.2012.06.025 22884087 78 Brugger P Threin M Wolters M Hochschulen auf einen Blick 2013 Reinbeck bei Hamburg Rowolt 79 Meyer O Pierre M Breuer G Simulators don't teach – Lernprozesse und Simulation Simulation in der Medizin 2013 Heidelberg Springer Medizin 10.1007/978-3-642-29436-5_5 Available from: http://dx.doi.org/10.1007/978-3-642-29436-5_5 80 Kohn LT Corrigan JM Donaldson MS Institute of Medicine (US) Committee on Quality of Health Care in America To Err is Human: Building a Safer Health System 2000 Washington (DC) National Academies Press 81 Haluck RS Krummel TM Simulation and virtual reality for surgical education Surg Techn Int 1999 8 59 63 82 Reznek M Harter P Krummel T Virtual reality and simulation: training the future emergency physician Acad Emerg Med 2002 9 1 78 87 10.1111/j.1553-2712.2002.tb01172.x Available from: http://dx.doi.org/10.1111/j.1553-2712.2002.tb01172.x 11772675 83 Fischer MR Bauer D Mohn K Finally finished! National Competence Based Catalogues of Learning Objectives for Undergraduate Medical Education (NKLM) and Dental Education (NKLZ) ready for trial GMS Z Med Ausbild 2015 32 3 Doc35 10.3205/zma000977 Available from: http://dx.doi.org/10.3205/zma000977 26677513

PMC005xxxxxx/PMC5003154.txt

==== Front J Lipid ResJ. Lipid ResjlrjlrjlrJournal of Lipid Research0022-22751539-7262The American Society for Biochemistry and Molecular Biology 27170732m06780110.1194/jlr.M067801Patient-Oriented and Epidemiological ResearchDifferential effects of EPA versus DHA on postprandial vascular function and the plasma oxylipin profile in men[S] McManus Seán 1*Tejera Noemi 1*Awwad Khader †http://orcid.org/0000-0001-5952-8756Vauzour David *§Rigby Neil §Fleming Ingrid †Cassidy Aedin *Minihane Anne Marie 2*Department of Nutrition and Preventive Medicine, Norwich Medical School,* University of East Anglia, Norwich NR4 7UQ, United KingdomInstitute for Vascular Signalling, Centre for Molecular Medicine,† Goethe University, 60590 Frankfurt, GermanyInstitute of Food Research,§ Norwich NR4 7UA, United Kingdom1 S. McManus and N. Tejera contributed equally to this article. 2 To whom correspondence should be addressed. e-mail: a.minihane@uea.ac.uk9 2016 9 2016 9 2016 57 9 1720 1727 14 3 2016 29 4 2016 Copyright © 2016 by the American Society for Biochemistry and Molecular Biology, Inc.2016Author’s Choice—Final version free via Creative Commons CC-BY license.Our objective was to investigate the impact of EPA versus DHA on arterial stiffness and reactivity and underlying mechanisms (with a focus on plasma oxylipins) in the postprandial state. In a three-arm crossover acute test meal trial, men (n = 26, 35–55 years) at increased CVD risk received a high-fat (42.4 g) test meal providing 4.16 g of EPA or DHA or control oil in random order. At 0 h and 4 h, blood samples were collected to quantify plasma fatty acids, long chain n-3 PUFA-derived oxylipins, nitrite and hydrogen sulfide, and serum lipids and glucose. Vascular function was assessed using blood pressure, reactive hyperemia index, pulse wave velocity, and augmentation index (AIx). The DHA-rich oil significantly reduced AIx by 13% (P = 0.047) with the decrease following EPA-rich oil intervention not reaching statistical significance. Both interventions increased EPA- and DHA-derived oxylipins in the acute postprandial state, with an (1.3-fold) increase in 19,20-dihydroxydocosapentaenoic acid evident after DHA intervention (P < 0.001). In conclusion, a single dose of DHA significantly improved postprandial arterial stiffness as assessed by AIx, which if sustained would be associated with a significant decrease in CVD risk. The observed increases in oxylipins provide a mechanistic insight into the AIx effect. augmentation indexblood pressurefish oilhydrogen sulfidelipidomicsnitric oxidenutritionomega-3 fatty acidspulse wave velocityeicosapentaenoic aciddocosahexaenoic acidBiotechnology and Biological Sciences Research Councilhttps://dx.doi.org/10.13039/501100000268BB/J004545/1Deutsche Forschungsgemeinschafthttps://dx.doi.org/10.13039/501100001659SFB 1039/A6 ==== Body An impact on arterial stiffness and compliance, associated with reduced blood pressure (1–4), is thought to be a significant contributor to the lower CVD risk associated with increased long chain n-3 PUFA (LC n-3 PUFA; EPA plus DHA) intake and status (4–6). In randomized controlled trials (RCTs) conducted to date, vascular function is typically measured in the fasting nonchallenged state. Given that adults following Westernized dietary patterns typically spend 16–18 h per day in the postprandial state, and that a postprandial phenotype characterized by an exaggerated lipemia and glycemia and inflammatory and oxidative stress is associated with vascular dysfunction (7), the postprandial vascular response is likely to be a more significant and discriminating marker of CVD risk. However, the impact of dietary factors and meal composition on acute postprandial vascular function is poorly understood. Increasing the total fat content of a test meal reduces both cardiac and peripheral postprandial vascular reactivity (7) and a small number of studies providing EPA plus DHA in the test meal have reported a positive effect of fish oil on vascular tone over 2–8 h (8–12). Although rarely considered, variability in the EPA and DHA ratio of the n-3 PUFA (fish or fish oil) source used in RCTs is likely to be a major determinant of the apparent inconsistency in findings reported. A limited number of chronic interventions that measured fasting vascular function indicate that DHA has a greater beneficial effect than EPA (4). However, with the exception of a single trial that examined relative impact on the cardiovascular hemodynamic responses to exercise (11), no study has directly compared the effects of EPA- versus DHA-rich oils on the postprandial response. Identification of the most bioactive LC n-3 PUFA or EPA/DHA ratio on “health” end points is important in order to refine current dietary recommendations and allow the stratification of intervention to suit an individual’s phenotype. Furthermore, with the emergence of nonfish sources of these fatty acids, such as separate EPA and DHA produced by marine microalga or in transgenic seed oils, and the availability of oil blends, all with a defined EPA and DHA content, information on their relative bioactivity is needed in order to inform these production systems (13, 14). The primary aim of this double-blind, placebo-controlled acute test meal study was to compare the impact of EPA versus DHA on postprandial (4 h) vascular function as assessed by blood pressure (BP), pulse wave velocity (PWV), reactive hyperemia index (RHI), and augmentation index (AIx). Underlying molecular mechanisms were investigated by quantifying potential circulating mediators of vascular function including nitrite [as a biomarker of nitric oxide (NO)], hydrogen sulfide (H2S), and LC n-3 PUFA hydroxy, epoxide, and thiol metabolites. Although shown to be potent modulators of vascular tone in cell, animal, and human ex vivo studies (15, 16), the role of these compounds in modulating EPA- and DHA-induced changes in vascular function in humans is unknown. MATERIALS AND METHODS This single-center dietary intervention study was conducted at the Norwich and Norfolk University Hospital Clinical Research Facility, at the University of East Anglia, between September 2012 and September 2013. All participants provided informed consent prior to commencing the study. It was conducted according to the guidelines described in the Declaration of Helsinki, and ethical clearance was provided by the English National Research Ethics Service, East of England Research Ethics Committee (reference 12/EE/0011). This trial was registered at www.clinicaltrials.gov (#NCT01692431). Subjects Twenty-six males between the ages of 35 and 55 years were recruited through local and targeted online advertisement. Volunteers were recruited if they possessed one or more of the following CVD risk factors: total cholesterol (TC) ≥6 mM, HDL cholesterol (HDL-C) ≤1.0 mM, systolic BP (SBP) >140 mmHg, diastolic BP (DBP) >90 mmHg, or waist circumference >102 cm, which individually were associated with an aged-adjusted relative risk (RR) of CVD ≥1.5, as previously described by Wilson et al. (17). Exclusion criteria included any clinically diagnosed disease, BP ≥160/100 mmHg, abnormal liver function or blood cell count, use of lipid-lowering or insulin-sensitizing medication, use of steroids or nonsteroidal anti-inflammatory drugs including aspirin, antibiotics use, or vaccinations within the previous 3 months, use of dietary fish oil or high dose antioxidant vitamin supplements, habitual consumers of more than one portion (140 g) of oily fish per week, heavy drinkers [>30 units (240 g) of alcohol per week], smokers or ex-smokers ceasing <3 months before screening. Study protocol On the three separate study days at least 4 weeks apart, meals containing EPA-rich oil (ERO), DHA-rich oil (DRO), or control oil (CO) were consumed. Treatment order for each participant was randomized by an independent researcher, who was not involved in the study design, data collection, or data analysis. For 3 days prior to their clinical visits, subjects were asked to follow a restricted diet that included avoidance of n-3 PUFA-rich foods and nitrate-, nitrite-, sulfite-, and sulfate-rich foods (supplementary Table 1) in order to standardize plasma n-3 PUFA, NO, and H2S status. Individuals were also asked to avoid alcohol or strenuous exercise in this 3-day period. The evening before their clinical visits, subjects were requested to consume a standard low-fat meal (<10 g of fat), which was provided. Adherence to the restrictions was assessed at each visit by questionnaire and a 24 h dietary recall. On the clinical visit, baseline (0 h) vascular and BP measures were taken along with a fasting (>10 h) blood sample. Following the consumption of the test meal, which participants were requested to consume within a 5 min period, the vascular and BP measures and blood sampling were repeated at 4 h. The 4 h time point was selected to coincide with the predicted peak plasma lipemia and EPA and DHA plasma concentrations following lipid consumption (7). Meal design Test meals were isocaloric [3,128 kJ; 9.5% of energy as protein, 51.0% of energy as fat (42.4 g total fat), and 39.5% of energy as carbohydrate] and matched for taste, smell, and appearance. They were composed of a high-fat chocolate-flavored milkshake containing the test fats and a peppermint extract to mask the fish oils, accompanied by toasted white bread (73 g) and a low fruit content jam (30 g). In the control meal, the milkshake was composed of 40 g of a 4:1 palm oil/soybean oil mixture, skimmed milk (150 g), chocolate-flavored powder (15 g), skimmed milk powder (15 g), and 2 g of peppermint oil extract. In the ERO- and DRO-test meals, 6.94 g and 8.33 g of the palm oil/soybean oil mixture was replaced by ERO or DRO (Epax®, FMC Health and Nutrition, Sandvika, Norway; see supplementary Table 2 for full fatty acid profile) to provide 4.16 g of EPA or DHA, respectively. Measures of BP and vascular function All vascular measures were taken in a temperature-controlled room following a 15 min rest period. BP was taken using an Omron 750 CP-II device (Omron, Milton Keynes, United Kingdom) in a supine position. Arterial stiffness and endothelial dependent reactive hyperemia were assessed using PWV and the EndoPAT RHI, respectively. AIx, a measure of the enhancement of central aortic pressure by a reflected pulse wave, which is dependent on the vascular tone of resistance arteries, was also undertaken. RHI was measured by arterial tonometry using an EndoPAT device (Itamar Medical Ltd., Caesarea, Israel) as previously described (18). Measures were taken for 5 min at rest, 5 min during brachial artery occlusion, and another 5 min subsequent to occlusion release. RHI was calculated by expressing the mean amplitude of peripheral arterial tone postocclusion relative to those at baseline. Nonocclusion effects were corrected for by accounting for changes in tone in the control arm probes. The RHI measure was also taken a second time at the postprandial (4 h) time point in response to 400 mg of sublingual nitroglycerine (GTN) in order to assess an impact of treatment on NO independent vasodilation. A Vicorder device (Skidmore Medical, Bristol, United Kingdom) was used for the measurement of the PWV and AIx as described previously (19, 20). In brief, measures of carotid-femoral PWV were taken via placement of a BP cuff at the level of the carotid and femoral artery. The distance between the two anatomical sites was measured in each individual in order to allow calculation of pulse wave propagation speed in meters per second. For AIx, the Vicorder cuff was positioned over the brachial artery and AIx established by analysis of the BP wave form. Specifically, AIx was calculated by expressing augmentation pressure (the difference between BP during occurrence of the anacrotic notch and maximum SBP) as a percentage of total pulse pressure. All BP and vascular measures were taken in triplicate apart from the RHI and RHI*GTN measures, which due to the basis of the methodology and the recovery time needed to return to premeasurement values, were measured only once. Serum/plasma lipids, glucose, nitrite, and H2S Blood samples for lipid and glucose analyses were left to sit at room temperature for 30 min and centrifuged at 2,100 g at 4°C for 10 min. Serum samples were stored at −80°C until analysis. Serum triacylglycerol (TAG), TC, HDL-C, NEFA, and glucose were measured on an ILab650 automated analyzer (Instrumentation Laboratory, Warrington, United Kingdom) with commercially available kits (Instrumentation Laboratory). Blood samples for nitrite and H2S quantification were drawn into ice-cold lithium heparin Vacutainers® and centrifuged immediately at 2,100 g at 4°C for 10 min. The obtained plasma samples were snap frozen and immediately stored at −80°C until analysis. Plasma nitrite analysis was conducted by reduction of nitrite to NO with chemiluminescent detection of NO undertaken via reaction with O3− as previously described (21). H2S was measured by reverse-phase HPLC coupled with fluorescence detection after derivatization with monobromobimane, following methodology detailed by Shen et al. (22). Plasma fatty acid and oxylipin analysis For fatty acid analysis, total lipids (TLs) were extracted and analyzed as described previously (23). In brief, TLs were extracted from 500 µl of plasma using chloroform-methanol (2:1, v/v). Fatty acid methyl esters (FAMEs) of the TL fraction were obtained by acid-catalyzed transmethylation, using nonadecanoic acid (19:0) (5% of the TL analyzed), as an internal standard. FAMEs were then purified by TLC and analyzed by gas-liquid chromatography. Oxylipin analysis of plasma samples was carried out at the Institute for Vascular Signaling (Frankfurt, Germany). Briefly, samples were spiked with the corresponding deuterated internal standards (Cayman Europe, Tallinn, Estonia), and lipid epoxide and diol derivatives were extracted twice into ethyl acetate (0.5 ml) and quantified using a Sciex API4000 mass spectrometer operating in the multiple reaction monitoring mode as previously described (24, 25). Statistical analysis Unless otherwise stated, data are presented as means ± SEM. Sample size calculations were undertaken with a power of 0.80 to detect clinically relevant changes at the 5% significance level for EndoPAT, PWV, and AIx. The primary outcome for the trial was the EndoPAT RHI index, with PWV and AIx as secondary outcomes. For EndoPAT, given an SD of 0.4 (26), a clinically relevant difference of 0.25 between treatments would be detectable in 21 subjects per treatment group. For PWV, given an SD of 1.2 m/s (27), a clinically relevant difference of 1.5 m/s would be detectable with a sample size of 11 participants per treatment group. For AIx, a clinically relevant 4% decrease in AIx would be detectable in a population of 13 individuals, assuming a population SD of 5% in AIx. We aimed to recruit 26 participants. This number was chosen to allow for a potential dropout rate of 15% and to provide a minimum of 22 participants completing the study. Statistical analysis was undertaken using the SPSS statistical software (version 22; Chicago, IL) with P < 0.05 taken as being significant. Prior to analysis, all data sets were checked for normality via visual inspection of normal Q-Q plots. Outliers were identified by analysis of normal Q-Q plots and variance in values greater than ± 2 SD. Two-factor repeated measures ANOVA was used to determine the independent and interactive impact of time and treatment on the outcome measures. Assumptions of sphericity were assessed via Mauchly’s test of sphericity. In cases where the assumptions of sphericity were not met, a Greenhouse-Geisser correction was applied. When significance was reached, subsequent post hoc analysis was undertaken to assess the significance of treatment group effect on the changes from baseline (4–0 h) using repeated measures ANOVA and Bonferroni adjustment. Pearson’s correlation analysis was used to establish the association between AIx responses and oxylipin levels. RESULTS Twenty-six individuals with a mean age of 45 ± 5 years and BMI of 27.4 ± 3.3 kg/m2 completed all three arms of the trial (Table 1). The DHA-rich or EPA-rich interventions had no significant effect on BP, PWV, RHI, or RHI in response to sublingual GTN administration (Table 2). For AIx, there was a statistically significant time (0 h vs. 4 h) effect (P < 0.010) and time × treatment interaction (P = 0.005) (Fig. 1), and post hoc analysis revealed significant (13.3%, P = 0.047) and borderline significant (11.3%, P = 0.06) change scores for the DRO and the ERO meal compared with control. No time × treatment interactions were evident for TAG, NEFA, or glucose (P = 0.36, 0.90, and 0.49, respectively). However, time effects were observed with 40% (P < 0.001) and 15% (P = 0.02) higher TAG and NEFA, respectively, and 9% lower glucose (P < 0.001) levels at 4 h for all treatment groups combined (Table 3). No impact of time or treatment was evident for nitrite or H2S concentrations. Fig. 1. AIx (%) at baseline and in response to treatment. Data are presented as mean ± SEM (n = 26). Two-factor repeated measures ANOVA. P, time < 0.010; P, time × treatment = 0.005. The asterisk indicates significantly different (P = 0.047) change scores for DRO when compared with CO. TABLE 1. Study population characteristics (n = 26) Characteristic Value Age (years) 45 ± 5 (36–54) Weight (kg) 87.0 ± 11.5 (73.5–129.4) BMI (kg/m2) 27.4 ± 3.3 (20.4–39.9) Waist circumference (cm) 95.5 ± 10.4 (76.8–128.0) TC (mM) 5.8 ± 0.9 (3.8–7.3) HDL-C (mM) 1.4 ± 0.3 (0.86–2.32) SBP (mmHg) 136 ± 10 (101–150) DBP (mmHg) 86 ± 7 (67–104) Data are presented as mean ± SEM with the range of values in brackets. TABLE 2. Vascular measurements at baseline (0 h) and in response to treatment (4 h) CO Meal ERO Meal DRO Meal P, Time × Treatmenta 0 h 4 h 0 h 4 h 0 h 4 h P, Timea DBP (mmHg) 77 ± 1 74 ± 2 77 ± 2 76 ± 2 75 ± 1 76 ± 1 0.25 0.15 SBP (mmHg) 125 ± 3 125 ± 2 127 ± 2 128 ± 2 125 ± 2 124 ± 1 0.93 0.38 RHI 2.4 ± 0.1 2.4 ± 0.1 2.4 ± 0.1 2.5 ± 0.1 2.4 ± 0.1 2.4 ± 0.1 0.75 0.92 RHI*GTN — 1.7 ± 0.1 — 1.5 ± 0.1 — 1.6 ± 0.1 — 0.59 PWV (m/s) 8.5 ± 0.2 8.3 ± 0.2 8.3 ± 0.1 8.4 ± 0.2 8.2 ± 0.2 8.2 ± 0.2 0.85 0.32 RHI*GTN, RHI after 400 mg of sublingual GTN. Data are presented as mean ± SEM (n = 26). a Two-factor repeated measures ANOVA. TABLE 3. Biochemical measures at baseline and in response to treatment CO Meal ERO Meal DRO Meal P, Time × Treatmenta 0 h 4 h 0 h 4 h 0 h 4 h P, Timea Triacylglycerol (mM) 1.7 ± 0.1 2.5 ± 0.2 1.6 ± 0.1 2.3 ± 0.2 1.6 ± 0.1 2.2 ± 0.1 <0.001 0.36 NEFA (μM) 359 ± 24 424 ± 36 335 ± 32 384 ± 35 351 ± 31 398 ± 33 0.020 0.90 Glucose (mM) 5.2 ± 0.1 4.7 ± 0.1 5.2 ± 0.1 4.7 ± 0.1 5.2 ± 0.1 4.9 ± 0.1 <0.001 0.49 Nitrite (mM) 88.7 ± 8.6 80.1 ± 7.2 84.3 ± 6.2 74.3 ± 6.5 78.4 ± 6.5 81.9 ± 0.1 0.37 0.46 H2S (mM) 348.3 ± 59.3 386.6 ± 62.0 369.2 ± 59.6 341.7 ± 39.4 377.5 ± 63.6 386.4 ± 61.1 0.76 0.35 Data are presented as mean ± SEM (n = 26). a Two-factor repeated measures ANOVA. The ERO and DRO meals increased plasma EPA and DHA by 158% (P = 0.03) and 88% (P = 0.05), respectively. Post hoc analysis showed no significant change in docosapentaenoic acid (DPA) in response to any of the meals after a Bonferroni correction was applied (Table 4). A total of 7 EPA and DHA hydroxy, epoxide, and diols metabolites were quantified (Table 4). The ERO meal significantly increased plasma concentrations of 15S-HEPE, 14,15-EpETE, 17,18-EpETE, 14,15-DiHETE, and 17,18-DiHETE with the DRO also increasing concentrations of the EPA-derived metabolites 14,15-DiHETE, 17,18-EpETE, 17,18-DiHETE, and the DHA-derived metabolite 19,20-DiHDPA. No changes were evident following consumption of the control meal. TABLE 4. Plasma LC n-3 PUFA concentrations along with select hydroxy, epoxide, and diol metabolites measured at baseline and in response to treatment (4 h) Control Meal ERO Meal DRO Meal P, Time × Treatmenta 0 h 4 h 0 h 4 h 0 h 4 h P, Timea Fatty acids (mg/ml)  EPA 3.0 ± 0.1 3.5 ± 0.1 3.3 ± 0.1 8.5 ± 0.3b 3.7 ± 0.1 5.1 ± 0.1 <0.001 0.49  DPA 0.9 ± 0.1 1.0 ± 0.1 1.1 ± 0.1 0.7 ± 0.1 0.8 ± 0.1 0.9 ± 0.1 0.86 0.02  DHA 6.6 ± 0.2 6.4 ± 0.1 6.7 ± 0.2 7.5 ± 0.2 6.4 ± 0.2 12.0 ± 0.3b 0.007 0.006 Oxylipins (ng/ml)  Hydroxy metabolites   15S-HEPE 0.2 ± 0.1 0.1 ± 0.0 0.8 ± 0.5 2.6 ± 0.7b 0.4 ± 0.2 0.9 ± 0.3 0.05 0.03   18S-HEPE 0.5 ± 0.2 0.2 ± 0.1 1.8 ± 1.0 6.5 ± 2.3 0.5 ± 0.3 3.8 ± 1.6 0.02 0.07  Epoxide metabolites   14,15-EpETE 0.6 ± 0.3 0.0 ± 0.0 0.4 ± 0.3 2.9 ± 0.8b 0.4 ± 0.2 0.5 ± 0.1 0.06 0.001   17,18-EpETE 0.9 ± 0.5 0.4 ± 0.0 0.6 ± 0.2 9.9 ± 1.8b 0.8 ± 0.3 2.6 ± 0.4b <0.001 <0.001  Diol metabolites   14,15-DiHETE 1.5 ± 0.1 1.5 ± 0.0 1.7 ± 0.1 3.2 ± 0.3b 1.6 ± 0.0 2.2 ± 0.1b <0.001 <0.001   17,18-DiHETE 3.0 ± 0.5 2.6 ± 0.2 2.9 ± 0.3 9.7 ± 1.0b 2.8 ± 0.2 6.0 ± 0.4b <0.001 <0.001   19,20-DiHDPA 0.9 ± 0.1 1.1 ± 0.1 0.8 ± 0.1 1.3 ± 0.1 0.9 ± 0.1 2.1 ± 0.2b <0.001 <0.001 DiHDPA, dihydroxydocosapentaenoic acid; DiHETE, dihydroxyeicosatetraenoic acid; EpETE, epoxyeicosatetraenoic acid; HEPE, hydroxyeicosapentaenoic. Data are presented as mean ± SEM (n = 26). a Two-factor repeated measures ANOVA. b Indicates a significant difference in change from baseline when compared with control. 14,15-DiHETE, 17,18-DiHETE, and 19,20-DiHDPA were significantly negatively correlated with AIx (r = −0.215, P = 0.007; r = −0.223, P = 0.005; and r = −0.199, P = 0.013, respectively) (supplementary Table 5) with the AIx absolute response to DHA correlated with baseline AIx values (r = −0.413, P = 0.036). DISCUSSION Arterial stiffness and vasodilation of the conduit arteries are important determinants of SBP, left ventricular hypertrophy, and overall CVD risk and are highly prognostic of future cardiovascular events (28, 29). Specifically AIx, a measure of pulse wave reflections influenced by vascular smooth muscle tone, affects central BP, with a reported hazard ratio of 1.68 [95% confidence interval (CI), 1.02–2.76] for all-cause mortality and 1.60 (95% CI, 1.07–2.39) for combined CVD end points for men in the highest versus lowest AIx tertile (30). In line with our recruitment of participants at above-average CVD risk, mean baseline AIx values of 24.6% were ∼30–50% higher than those observed in RCTs that included healthy males (31, 32) or in the lowest male CVD risk tertile in the Copenhagen City Heart Study (CCHS) (33) and were within the highest risk tertile for CCHS (33). Our main finding is an overall treatment effect on AIx, with the modest differences in responses between DHA and EPA intervention relative to control, although statistically different, unlikely to be of clinical significance. If sustained, the observed DHA-mediated 13.3% reduction would equate to a decrease in 10-year CVD risk from 3.3% to 2.8% in this population, using associations generated by the European Society of Cardiology (34). Furthermore EPA- and DHA-derived epoxides and diols were shown to be highly modifiable in the postprandial state and likely to contribute to the observed improved vascular function. While to the best of our knowledge no previous RCT has compared the impact of EPA versus DHA on AIx, a limited number of studies have reported a positive impact of chronic and acute combined EPA + DHA supplementation on this measure of arterial stiffness (9, 10, 35, 36). Using a comparable LC n-3 PUFA exposure (4.7 g EPA + DHA; EPA/DHA, 0.67), Chong et al. (9) reported that the inclusion of fish oil in a test meal reduced AIx in the 1.5–4 h postprandial period in healthy adults, with an improved stiffness index only evident in male participants. Purcell et al. (10) observed significant reductions in AIx in healthy men after fish oil (5 g EPA + DHA; EPA/DHA, 1.6) and algal oil (5 g DHA only) administration, during a 6 h postprandial assessment, which were most evident at 2 h and of comparable size effect to those observed in the current RCT. However, they did not include an EPA-only meal, which would allow direct comparison of the vasoactivity of EPA versus DHA on AIx and other measures of vascular function (10). In the current study, no effect of the EPA or DHA meals on PWV or RHI was observed. PWV is a proximal measure of gross arterial stiffness determined by the speed at which the pulse wave travels through the arterial tree, with an increased speed indicating increased stiffness and overall vascular dysfunction (28). In a meta-analysis of 17 studies, an increased PWV was associated with an increased risk of total cardiovascular events, cardiovascular mortality, and all-cause mortality [RR (95% CI) of 2.26 (1.89–2.70), 2.02 (1.68–2.42), and 1.90 (1.61–2.70), respectively] (37). The observed lack of impact of treatment on postprandial PWV in the present study corroborates previous observations showing that PWV is not acutely modified in response to altered fatty acid composition (38, 39). Peripheral arterial tonometry RHI as assessed by the proprietary EndoPAT device was used as an observer-independent, high-throughput technique to assess endothelium-dependent vasodilation (EDV) (40). Since this study was initiated, concerns have been expressed regarding the sensitivity of the EndoPAT technique, in particular its ability to assess (subtle changes in) endothelial function in health individuals or at the early stages of vascular dysfunction (41, 42). Future trials should use more sensitive techniques, such as flow-mediated dilatation, in order to assess the impact of EPA versus DHA on postprandial EDV. The prognostic value of a compromised flow-induced vasodilation (43), the observation of improved response to flow following chronic EPA + DHA supplementation (3) and of a differential and stronger impact of DHA relative to EPA using a less widely used and indirect measure of EDV, namely systemic vascular resistance as measured by the Finometer finger arterial BP monitor (11), provides strong justification for such research to be conducted. Much of the mechanistic focus centered around the modulation of vascular function by EPA and DHA has been on NO bioavailability and endothelial NO synthase expression and phosphorylation status (44). There has been a degree of inconsistency with regard to the capacity of LC n-3 PUFA to modify NO production postprandially, with some studies indicating an effect (8), whereas others have observed no changes (10). As NO is a labile compound, its status is often estimated via the concentrations of its oxidation products, most notably nitrite and nitrate, with nitrite thought to be the most reflective of changes in NO concentration (45). In agreement with previous investigations utilizing this method, there was no evidence of postprandial EPA- or DHA-mediated changes in plasma levels of nitrite or total NO metabolites (10). H2S has recently emerged as a novel endothelium-derived regulator of vascular function (16, 46). Although the impact of EPA and DHA on H2S status in the vasculature is almost completely unknown, a recent study reported an activation of cystathionine-γ-lyase, the main H2S biosynthesis enzyme, in response to DHA-rich tuna-oil supplementation in the lung tissue of Sprague-Dawley rats (47). However, no EPA- or DHA-mediated differences were evident in the current RCT, and it is unlikely that H2S contributes to reduced arterial stiffness associated with acute DHA consumption. Finally, our study was novel in its investigation of the impact of EPA versus DHA on the plasma concentration of selected vasoactive cytochrome P450 (CYP) enzyme-derived EPA and DHA metabolites (oxylipins), and for the first time concurrent assessment of the impact of LC n-3 PUFA on the vasoactive oxylipin profile and vascular responses in humans was conducted. Previous in vitro, ex vivo, and animal investigations have shown that the LC n-3 PUFA epoxides are particularly potent, and that the DHA epoxides account for ∼75% of DHA’s capacity to illicit vasodilation in in vitro models (48). Furthermore, the DHA epoxides have emerged as being more vasoactive than their EPA counterparts (15, 48, 49). In 2010, Shearer et al. (50) reported for the first time the presence of EPA- and DHA-derived epoxide and diols in human plasma and their modulation by EPA + DHA supplementation, which has been subsequently corroborated (51, 52). Consistent with the only previous acute assessment (52), which used a supraphysiological DHA-rich fish oil dose (26 g EPA + DHA; EPA/DHA, 0.12), it can be confirmed that the hydroxy, epoxide, and diol oxylipins are highly modifiable in the postprandial state with several fold higher concentration evident at 4 h postconsumption of test meals containing 4.16 g of EPA or DHA. The DHA-derived diol 19,20-epoxydocosapentaenoic acid (19,20-EpDPE) has emerged as a potentially potent vasomodulator with direct effects on vascular smooth muscle tone (48). A limitation of this study is that we were not able to detect 19,20-EpDPE, which fell below the limit of detection. Results from the human serum metabolome project suggest that the molar ratio of 19,20-EpDPE to 19,20-DiHDPA is in the range of 1 in 8 (53), which places plasma concentrations of 19,20-EpDPE in the low picomolar range. However, we did observe a highly significant 1.3-fold increase in the daughter 19,20-DiHDPA diol in response to DHA consumption, with 19,20-DiHDPA concentration negatively correlated with AIx responses. Given that 19,20-DiHDPA is directly and exclusively derived from 19,20-EpDPE via the activity of the soluble epoxide hydrolase, the higher postprandial 19,20-DiHDPA levels are likely reflective of increases in 19,20-EpDPE, which is proposed to have contributed to the vascular impact of DHA versus EPA. However, we cannot preclude the possibility that EPA-derived oxylipins also contributed to the vascular response following DRO intervention. The strong trend toward an increase in plasma EPA following DRO intervention, was likely due to some EPA being provided as part of the DRO intervention or the known tissue retroconversion of DHA to EPA (54). These sources provided sufficient amounts for the CYP epoxygenases that generally prefer EPA over DHA and arachidonic acid (55), with the DRO meal resulting in an increase in three EPA-derived oxylipin metabolites, with 14,15-DiHETE and 17,18-DiHETE correlated with AIx. In addition to the CYP-derived oxylipins, prostanoids and lipoxygenase-derived resolvins may contribute to the vascular response to n-3 PUFA intervention. Although potentially important in the vascular wall, this latter group of compounds has proved difficult to quantify in the human circulation. Therefore, following DHA intervention changes in the tissues status of both DHA and EPA and their associated oxylipins and other metabolites may act in a complementary and additive fashion (56) to mediate the impact on vascular function. The strengths of the current study were its crossover design, its inclusion of a range of measures of vascular function, with simultaneous analysis of vascular function and potential lipid- and nonlipid-derived mediators in the circulation at the clinical assessment time points. A limitation is the likely lack of sensitivity of the EndoPAT technique used to assess the impact of treatment on EDV. In summary, the results of this study show that DHA improves postprandial vascular function, with an effect size that would translate into a meaningful reduction in CVD risk, with a strong trend also evident following EPA intervention. The targeted lipidomic profiling indicates that postprandial changes in the LC n-3 PUFA epoxides and diol occur and are likely to play a mechanistic role as effector molecules of the vascular responses. Further research is needed to fully establish the individual impact of EPA versus DHA on clinical end point and disease biomarkers and to gain insight into the bioactive lipids modulating the effects. Ultimately, such information would allow the refinement of current EPA and DHA recommendations and the development of fish oil blends (with defined EPA and DHA content) to suit phenotype and would inform the design of much needed nonmarine sources of these fatty acids. Supplementary Material Supplemental Data The authors thank Epax®, FMC Health and Nutrition, Sandvika, Norway, for kindly providing the fish oil used in this clinical trial. Abbreviations: AIxaugmentation index BPblood pressure CIconfidence interval COcontrol oil CYPcytochrome P450 DBPdiastolic blood pressure DiHDPAdihydroxydocosapentaenoic acid DiHETEdihydroxyeicosatetraenoic acid DRODHA-rich oil EDVendothelium-dependent vasodilation EpDPEepoxydocosapentaenoic acid EpETEepoxyeicosatetraenoic acid EROEPA-rich oil GTNnitroglycerine HDL-CHDL cholesterol HEPEhydroxyeicosapentaenoic LC n-3 PUFAlong chain n-3 PUFA PWVpulse wave velocity RCTrandomized controlled trial RHIreactive hyperemia index SBPsystolic blood pressure TAGtriacylglycerol TCtotal cholesterol TLtotal lipid The work was supported by the Faculty of Medicine and Health Sciences, University of East Anglia PhD studentship program and Biotechnology and Biological Sciences Research Council Institute Strategic Program Grant BB/J004545/1. Institute for Vascular Signalling work was supported by a Deutsche Forschungsgemeinschaft (SFB 1039/A6). The authors have no conflicts of interest to disclose. [S] The online version of this article (available at http://www.jlr.org) contains a supplement. ==== Refs REFERENCES 1. Miller P. E. , Van Elswyk M. , and Alexander D. D. 2014 Long-chain omega-3 fatty acids eicosapentaenoic acid and docosahexaenoic acid and blood pressure: a meta-analysis of randomized controlled trials . Am. J. Hypertens. 27 : 885 –896 .24610882 2. Pase M. P. , Grima N. A. , and Sarris J. 2011 Do long-chain n-3 fatty acids reduce arterial stiffness? A meta-analysis of randomised controlled trials . Br. J. Nutr. 106 : 974 –980 .22005318 3. Wang Q. , Liang X. , Wang L. , Lu X. , Huang J. , Cao J. , Li H. , and Gu D. 2012 Effect of omega-3 fatty acids supplementation on endothelial function: a meta-analysis of randomized controlled trials . Atherosclerosis. 221 : 536 –543 .22317966 4. Mori T. A. 2014 Dietary n-3 PUFA and CVD: a review of the evidence . Proc. Nutr. Soc. 73 : 57 –64 .24119287 5. Mozaffarian D. , and Wu J. H. 2011 Omega-3 fatty acids and cardiovascular disease: effects on risk factors, molecular pathways, and clinical events . J. Am. Coll. Cardiol. 58 : 2047 –2067 .22051327 6. Saravanan P. , Davidson N. C. , Schmidt E. B. , and Calder P. C. 2010 Cardiovascular effects of marine omega-3 fatty acids . Lancet. 376 : 540 –550 .20638121 7. Jackson K. G. , Poppitt S. D. , and Minihane A. M. 2012 Postprandial lipemia and cardiovascular disease risk: interrelationships between dietary, physiological and genetic determinants . Atherosclerosis. 220 : 22 –33 .21955695 8. Armah C. K. , Jackson K. G. , Doman I. , James L. , Cheghani F. , and Minihane A. M. 2008 Fish oil fatty acids improve postprandial vascular reactivity in healthy men . Clin. Sci. 114 : 679 –86 .18052925 9. Chong M. F. , Lockyer S. , Saunders C. J. , and Lovegrove J. A. 2010 Long chain n-3 PUFA-rich meal reduced postprandial measures of arterial stiffness . Clin. Nutr. 29 : 678 –681 .20199827 10. Purcell R. , Latham S. H. , Botham K. M. , Hall W. L. , and Wheeler-Jones C. P. 2014 High-fat meals rich in EPA plus DHA compared with DHA only have differential effects on postprandial lipemia and plasma 8-isoprostane F2alpha concentrations relative to a control high-oleic acid meal: a randomized controlled trial . Am. J. Clin. Nutr. 100 : 1019 –1028 .25099540 11. Rontoyanni V. G. , Hall W. L. , Pombo-Rodrigues S. , Appleton A. , Chung R. , and Sanders T. A. 2012 A comparison of the changes in cardiac output and systemic vascular resistance during exercise following high-fat meals containing DHA or EPA . Br. J. Nutr. 108 : 492 –499 .22348439 12. West S. G. , Hecker K. D. , Mustad V. A. , Nicholson S. , Schoemer S. L. , Wagner P. , Hinderliter A. L. , Ulbrecht J. , Ruey P. , and Kris-Etherton P. M. 2005 Acute effects of monounsaturated fatty acids with and without omega-3 fatty acids on vascular reactivity in individuals with type 2 diabetes . Diabetologia. 48 : 113 –122 .15624100 13. Ruiz-Lopez N. , Usher S. , Sayanova O. V. , Napier J. A. , and Haslam R. P. 2015 Modifying the lipid content and composition of plant seeds: engineering the production of LC-PUFA . Appl. Microbiol. Biotechnol. 99 : 143 –154 .25417743 14. Salem N. Jr., and Eggersdorfer M. 2015 Is the world supply of omega-3 fatty acids adequate for optimal human nutrition? Curr. Opin. Clin. Nutr. Metab. Care. 18 : 147 –154 .25635599 15. Morin C. , Sirois M. , Echave V. , Rizcallah E. , and Rousseau E. 2009 Relaxing effects of 17(18)-EpETE on arterial and airway smooth muscles in human lung . Am. J. Physiol. Lung Cell. Mol. Physiol. 296 : L130 –L139 .18978038 16. Mustafa A. K. , Sikka G. , Gazi S. K. , Steppan J. , Jung S. M. , Bhunia A. K. , Barodka V. M. , Gazi F. K. , Barrow R. K. , Wang R. , 2011 Hydrogen sulfide as endothelium-derived hyperpolarizing factor sulfhydrates potassium channels . Circ. Res. 109 : 1259 –1268 .21980127 17. Wilson P. W. F. , D’Agostino R. B. , Levy D. , Belanger A. M. , Silbershatz H. , and Kannel W. B. 1998 Prediction of coronary heart disease using risk factor categories . Circulation. 97 : 1837 –1847 .9603539 18. Bonetti P. O. , Barsness G. W. , Keelan P. C. , Schnell T. I. , Pumper G. M. , Kuvin J. T. , Schnall R. P. , Holmes D. R. Jr., Higano S. T. , and Lerman A. 2003 Enhanced external counterpulsation improves endothelial function in patients with symptomatic coronary artery disease . J. Am. Coll. Cardiol. 41 : 1761 –1768 .12767662 19. Hickson S. S. , Butlin M. , Broad J. , Avolio A. P. , Wilkinson I. B. , and McEniery C. M. 2009 Validity and repeatability of the Vicorder apparatus: a comparison with the SphygmoCor device . Hypertens. Res. 32 : 1079 –1085 .19779487 20. Pucci G. , Cheriyan J. , Hubsch A. , Hickson S. S. , Gajendragadkar P. R. , Watson T. , O’Sullivan M. , Woodcock-Smith J. , Schillaci G. , Wilkinson I. B. , 2013 Evaluation of the Vicorder, a novel cuff-based device for the noninvasive estimation of central blood pressure . J. Hypertens. 31 : 77 –85 .23079681 21. Feelisch M. , Rassaf T. , Mnaimneh S. , Singh N. , Bryan N. S. , Jourd’Heuil D. , and Kelm M. 2002 Concomitant S-, N-, and heme-nitros(yl)ation in biological tissues and fluids: implications for the fate of NO in vivo . FASEB J. 16 : 1775 –1785 .12409320 22. Shen X. , Pattillo C. B. , Pardue S. , Bir S. C. , Wang R. , and Kevil C. G. 2011 Measurement of plasma hydrogen sulfide in vivo and in vitro . Free Radic. Biol. Med. 50 : 1021 –1031 .21276849 23. Vauzour D. , Tejera N. , O’Neill C. , Booz V. , Jude B. , Wolf I. M. , Rigby N. , Silvan J. M. , Curtis P. J. , Cassidy A. , 2015 Anthocyanins do not influence long-chain n-3 fatty acid status: studies in cells, rodents and humans . J. Nutr. Biochem. 26 : 211 –218 .25573539 24. Frömel T. , Jungblut B. , Hu J. , Trouvain C. , Barbosa-Sicard E. , Popp R. , Liebner S. , Dimmeler S. , Hammock B. D. , and Fleming I. 2012 Soluble epoxide hydrolase regulates hematopoietic progenitor cell function via generation of fatty acid diols . Proc. Natl. Acad. Sci. USA. 109 : 9995 –10000 .22665795 25. Michaelis U. R. , Fisslthaler B. , Barbosa-Sicard E. , Falck J. R. , Fleming I. , and Busse R. 2005 Cytochrome P450 epoxygenases 2C8 and 2C9 are implicated in hypoxia-induced endothelial cell migration and angiogenesis . J. Cell Sci. 118 : 5489 –5498 .16291720 26. Martin E. A. , Nelson R. E. , Felmlee-Devine M. D. , Brown T. E. , and Lerman A. 2011 Comparing EndoPAT and BIOPAC measurement of vascular responses to mental stress . Cell Biochem. Funct. 29 : 272 –278 .21671245 27. Asmar R. G. , London G. M. , O’Rourke M. E. , Safar M. E. ; REASON Project Coordinators and Investigators . 2001 Improvement in blood pressure, arterial stiffness and wave reflections with a very-low-dose perindopril/indapamide combination in hypertensive patient: a comparison with atenolol . Hypertension. 38 : 922 –926 .11641310 28. Mitchell G. F. 2014 Arterial stiffness and hypertension . Hypertension. 64 : 13 –18 .24752432 29. Schächinger V. , Britten M. B. , and Zeiher A. M. 2000 Prognostic impact of coronary vasodilator dysfunction on adverse long-term outcome of coronary heart disease . Circulation. 101 : 1899 –1906 .10779454 30. Janner J. H. , Godtfredsen N. S. , Ladelund S. , Vestbo J. , and Prescott E. 2013 High aortic augmentation index predicts mortality and cardiovascular events in men from a general population, but not in women . Eur. J. Prev. Cardiol. 20 : 1005 –1012 .22649125 31. Macready A. L. , George T. W. , Chong M. F. , Alimbetov D. S. , Jin Y. , Vidal A. , Spencer J. P. , Kennedy O. B. , Tuohy K. M. , Minihane A. M. , 2014 Flavonoid-rich fruit and vegetables improve microvascular reactivity and inflammatory status in men at risk of cardiovascular disease–FLAVURS: a randomized controlled trial . Am. J. Clin. Nutr. 99 : 479 –489 .24452238 32. Hallund J. , Madsen B. O. , Bugel S. H. , Jacobsen C. , Jakobsen J. , Krarup H. , Holm J. , Nielsen H. H. , and Lauritzen L. 2010 The effect of farmed trout on cardiovascular risk markers in healthy men . Br. J. Nutr. 104 : 1528 –1536 .20594395 33. Janner J. H. , Godtfredsen N. S. , Ladelund S. , Vestbo J. , and Prescott E. 2010 Aortic augmentation index: reference values in a large unselected population by means of the SphygmoCor device . Am. J. Hypertens. 23 : 180 –185 .19959999 34. Nürnberger J. , Keflioglu-Scheiber A. , Opazo Saez A. M. , Wenzel R. R. , Philipp T. , and Schäfers R. F. 2002 Augmentation index is associated with cardiovascular risk . J. Hypertens. 20 : 2407 –2414 .12473865 35. Shearer G. C. , Pottala J. V. , Hansen S. N. , Brandenburg V. , and Harris W. S. 2012 Effects of prescription niacin and omega-3 fatty acids on lipids and vascular function in metabolic syndrome: a randomized controlled trial . J. Lipid Res. 53 : 2429 –2435 .22892157 36. Siasos G. , Tousoulis D. , Oikonomou E. , Zaromitidou M. , Verveniotis A. , Plastiras A. , Kioufis S. , Maniatis K. , Miliou A. , Siasou Z. , 2013 Effects of omega-3 fatty acids on endothelial function, arterial wall properties, inflammatory and fibrinolytic status in smokers: a cross over study . Int. J. Cardiol. 166 : 340 –346 .22100606 37. Vlachopoulos C. , Aznaouridis K. , and Stefanadis C. 2010 Prediction of cardiovascular events and all-cause mortality with arterial stiffness: a systematic review and meta-analysis . J. Am. Coll. Cardiol. 55 : 1318 –1327 .20338492 38. Berry S. E. E. , Tucker S. , Banerji R. , Jiang B. , Chowienczyk P. J. , Charles S. M. , and Sanders T. A. B. 2008 Impaired postprandial endothelial function depends on the type of fat consumed by healthy men . J. Nutr. 138 : 1910 –1914 .18806100 39. Lithander F. E. , Herlihy L. K. , Walsh D. M. , Burke E. , Crowley V. , and Mahmud A. 2013 Postprandial effect of dietary fat quantity and quality on arterial stiffness and wave reflection: a randomised controlled trial . Nutr. J. 12 : 93 .23841960 40. Flammer A. J. , Anderson T. , Celermajer D. S. , Creager M. A. , Deanfield J. , Ganz P. , Hamburg N. M. , Luscher T. F. , Shechter M. , Taddei S. , 2012 The assessment of endothelial function: from research into clinical practice . Circulation. 126 : 753 –767 .22869857 41. Moerland M. , Kales A. J. , Schrier L. , van Dongen M. G. , Bradnock D. , and Burggraaf J. 2012 Evaluation of the EndoPAT as a tool to assess endothelial function . Int. J. Vasc. Med. 2012 : 904141 .22500237 42. Lind L. 2013 Relationships between three different tests to evaluate endothelium-dependent vasodilation and cardiovascular risk in a middle-aged sample . J. Hypertens. 31 : 1570 –1574 .23625114 43. Ras R. T. , Streppel M. T. , Draijer R. , and Zock P. L. 2013 Flow-mediated dilation and cardiovascular risk prediction: a systematic review with meta-analysis . Int. J. Cardiol. 168 : 344 –351 .23041097 44. Balakumar P. , and Taneja G. 2012 Fish oil and vascular endothelial protection: bench to bedside . Free Radic. Biol. Med. 53 : 271 –279 .22584102 45. Lauer T. , Preik M. , Rassaf T. , Strauer B. E. , Deussen A. , Feelisch M. , and Kelm M. 2001 Plasma nitrite rather than nitrate reflects regional endothelial nitric oxide synthase activity but lacks intrinsic vasodilator action . Proc. Natl. Acad. Sci. USA. 98 : 12814 –12819 .11606734 46. Wang R. 2011 Signaling pathways for the vascular effects of hydrogen sulfide . Curr. Opin. Nephrol. Hypertens. 20 : 107 –112 .21301337 47. Huang T. , Wahlqvist M. L. , and Li D. 2010 Docosahexaenoic acid decreases plasma homocysteine via regulating enzyme activity and mRNA expression involved in methionine metabolism . Nutrition. 26 : 112 –119 .19815389 48. Morin C. , Fortin S. , and Rousseau E. 2011 19,20-EpDPE, a bioactive CYP450 metabolite of DHA monoacyglyceride, decreases Ca(2)(+) sensitivity in human pulmonary arteries . Am. J. Physiol. Heart Circ. Physiol. 301 : H1311 –H1318 .21821782 49. Ye D. , Zhang D. , Oltman C. , Dellsperger K. , Lee H. C. , and VanRollins M. 2002 Cytochrome p-450 epoxygenase metabolites of docosahexaenoate potently dilate coronary arterioles by activating large-conductance calcium-activated potassium channels . J. Pharmacol. Exp. Ther. 303 : 768 –776 .12388664 50. Shearer G. C. , Harris W. S. , Pedersen T. L. , and Newman J. W. 2010 Detection of omega-3 oxylipins in human plasma and response to treatment with omega-3 acid ethyl esters . J. Lipid Res. 51 : 2074 –2081 .19671931 51. Schebb N. H. , Ostermann A. I. , Yang J. , Hammock B. D. , Hahn A. , and Schuchardt J. P. 2014 Comparison of the effects of long-chain omega-3 fatty acid supplementation on plasma levels of free and esterified oxylipins . Prostaglandins Other Lipid Mediat. 113–115 : 21 –29 . 52. Strassburg K. , Esser D. , Vreeken R. J. , Hankemeier T. , Muller M. , van Duynhoven J. , van Golde J. , van Dijk S. J. , Afman L. A. , and Jacobs D. M. 2014 Postprandial fatty acid specific changes in circulating oxylipins in lean and obese men after high-fat challenge tests . Mol. Nutr. Food Res. 58 : 591 –600 .24127338 53. Psychogios N. , Hau D. D. , Peng J. , Guo A. C. , Mandal R. , Bouatra S. , Sinelnikov I. , Krishnamurthy R. , Eisner R. , Gautam B. , 2011 The human serum metabolome . PLoS One. 6 : e16957 .21359215 54. Conquer J. A. , and Holub B. J. 1997 Dietary docosahexaenoic acid as a source of eicosapentaenoic acid in vegetarians and omnivores . Lipids. 32 : 341 –345 .9076673 55. Fischer R. , Konkel A. , Mehling H. , Blossey K. , Gapelyuk A. , Wessel N. , von Schacky C. , Dechend R. , Muller D. N. , Rothe M. , 2014 Dietary omega-3 fatty acids modulate the eicosanoid profile in man primarily via the CYP-epoxygenase pathway . J. Lipid Res. 55 : 1150 –1164 .24634501 56. Mozaffarian D. , and Wu J. H. Y. 2012 (n-3) Fatty acids and cardiovascular health: are effects of EPA and DHA shared or complementary? J. Nutr. 142 : 614S –625S .22279134

PMC005xxxxxx/PMC5003169.txt

==== Front Int J Food Sci Technol10.1111/(ISSN)1365-2621IJFSInternational Journal of Food Science & Technology0950-54231365-2621John Wiley and Sons Inc. Hoboken 10.1111/ijfs.13192IJFS13192ReviewReviewsA review of the impact of processing on nutrient bioaccessibility and digestion of almonds M. Grundy et al.Grundy Myriam Marie‐Louise 1 Lapsley Karen 2 Ellis Peter Rory 1 1 Diabetes and Nutritional SciencesKing's College LondonFranklin‐Wilkins Building, 150 Stamford StreetLondonSE1 9NHUK2 Almond Board of California1150 Ninth Street Ste.1500ModestoCA95354USA* Correspondent: E‐mail: peter.r.ellis@kcl.ac.uk31 7 2016 9 2016 51 9 10.1111/ijfs.2016.51.issue-91937 1946 24 3 2016 02 6 2016 © 2016 The Authors. International Journal of Food Science and Technology published by John Wiley & Sons Ltd on behalf of Institute of Food Science and TechnologyThis is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.Summary Almond kernels contain phytochemicals and nutrients that potentially have positive health benefits in relation to heart disease, diabetes and obesity. One important mechanism associated with these benefits is an imposed limit on bioaccessibility (release) of nutrients, such as lipids, from almond tissue during mastication and digestion. Recent studies have demonstrated the importance of food structure during the digestion of plant foods. In particular, in the almond kernel, depending on its structure and degree of processing, the amount of lipid released from the almond tissue matrix and the fatty acids produced from lipolysis has been found to vary substantially. This review aims at discussing the commercial methods of almond processing and the different almond forms produced for human consumption, mainly with respect to their impact on nutrient composition, digestion and metabolism. Almondsdietary fibrebioaccessibility/digestibilitylipidsprocessing effectsBBSRC DRINC ProjectBB/H004866/1 source-schema-version-number2.0component-idijfs13192cover-dateSeptember 2016details-of-publishers-convertorConverter:WILEY_ML3GV2_TO_NLMPMC version:4.9.4 mode:remove_FC converted:24.08.2016 ==== Body Introduction Almond seeds or kernels are highly versatile and can be eaten on their own or as part of a number of food products. Almonds are consumed world‐wide with the United States being the largest producer (Gradziel, 2011; Harris & Ferguson, 2013). There is a wide range of methods currently used to process almond seeds (e.g. heat processing and particle size reduction). These processes have led to the development of almond‐based products with enhanced organoleptic characteristics, but this is not without consequences for the nutritional properties of the almond tissue. From a nutritional perspective, almonds are a useful food and ingredient for other foods as they contain a range of macro‐ and micronutrients as well as phytochemicals. Epidemiological evidence and the results of numerous metabolic studies in humans have shown that the consumption of almonds and other nuts reduce a number of risk factors associated with noncommunicable disease, notably type 2 diabetes, obesity and cardiovascular disease (Richardson et al., 2009; Tan & Mattes, 2013; Nishi et al., 2014; Berryman et al., 2015). One crucial factor that seems to explain these putative health benefits is the physical behaviour of almonds in the gastrointestinal (GI) tract, especially how almonds are disassembled and the rate and extent to which they release macronutrients such as lipid. However, mechanisms that explain the physiological effects and the long term benefits of tree nuts like almonds are not well understood, particularly the properties of almond cell walls in each compartment of the GI tract (i.e. mouth, stomach and intestine). Obtaining information about the changes occurring to the structure of the almond tissue as the digestion process progresses and the mechanisms of lipid release is of crucial importance (Ellis et al., 2004; Mandalari et al., 2014; Grundy et al., 2015a,b). For instance, the size and microstructure of the particles following oral processing have a significant effect on nutrient bioaccessibility (release), digestion kinetics and other physiological processes in the GI tract (Grundy et al., 2015a). The purpose of this review is to present the most common processing techniques applied to almond kernels, and how they affect almond structure and their subsequent impact on the digestion of lipid and other nutrients. Almond anatomy and composition Macrostructure The sweet almond (Prunus dulcis (Mill.) D.A. Webb or Amygdalus communis L.) belongs to the Rosaceae family. Almond is a drupe1 of which the only edible part is the kernel or seed (Gradziel, 2011). The latter is composed of an embryo (two cotyledons), surrounded by a skin also called testa. The pericarp, which encloses the kernel, contains a green fleshy hull and a hard pitted shell (Fig. 1). Figure 1 Multiscale structure of almond fruit with kernel. Note that the size of the almond cell is about 35 μm and the oil body between 1 and 5 μm. Microstructure and composition The almond cotyledons (i.e. the white lipid‐bearing tissue) are made of rounded cells, principally parenchyma, with a relatively thin cell wall (~0.1–0.3 μm) (Fig. 2). Pigmented sclerenchyma (outer layer) and parenchyma cells as well as xylem tissue compose the testa (Mandalari et al., 2010a). The testa cells possess a secondary cell wall, which is confirmed by the presence of a significant amount of lignin (Femenia et al., 2001). A layer of aleurone cells, containing globoid crystals as well as protein and lipid bodies, forms the endosperm that separates the testa (spermoderm and perisperm) from the cotyledon (Winton & Winton, 1932; Young et al., 2004). Figure 2 Transmission electron micrograph image of almond kernel showing oil bodies (white inclusions), protein bodies (black inclusion) and the cell walls. Scale bar = 2 μm. Almonds are a valuable dietary source of lipid (comprising mainly monounsaturated fatty acids), protein, dietary fibre, vitamins (e.g. vitamin E), minerals, phenolic compounds and phytosterols (Bolling et al., 2011; Yada et al., 2011; Fernandez‐Cuesta et al., 2012) (Table 1). Table 1 Nutrient and total phenolic composition of almonds Ranges per 100 g of almond g mg Macronutrients Protein 16–23 Lipid 44–61 Saturated fats 3–4 Monounsaturated fats 31–35 Polyunsaturated fats 11–12 Carbohydrates Total sugars 4–6 Total dietary fibres 11–14 Water 4–5 Micronutrients Minerals Calcium 264–300 Magnesium 230–268 Phosphorus 440–510 Potassium 705–730 Zinc 3.0–4.1 Copper 0.9–1.3 Manganese 1.2–1.8 Vitamins Riboflavin 1.0–1.1 Vitamin E (α‐tocopherol) 25–27 Total phenolic compounds 260–350 Adapted from (Richardson et al., 2009; Bolling et al., 2011; Yada et al., 2011, USDA, 2015). John Wiley & Sons, LtdProtein The major storage protein found in almonds, sometimes called amandin or almond major protein, belongs to the legumin class of seed proteins, which itself is a part of the globulin family (Osborne & Campbell, 1896; Kshirsagar et al., 2011). Globulin proteins are classified according to their sedimentation coefficient, with the legumin type being 11S. Amandin accounts for about 70% of the total soluble proteins. It has a hexameric structure and each of the six subunits is composed of two polypeptides (α‐chain of about 45 kDa and β‐chain of about 20 kDa) linked by a disulphide bridge, with a molecular weight of approximately 450 kDa (Sathe et al., 2002). Along with 2S albumin, amandin, in particular the glutamine‐rich region of the protein, is responsible for the food allergy reactions observed in certain individuals following the consumption of almonds (Alasalvar & Shahidi, 2009; Willison et al., 2013). Lipids Almond lipids, composed predominantly of triacylglycerols (TAG), are assembled into oil bodies. These organelles are delimited by a monolayer of phospholipids in which oleosins, integral proteins, are embedded (Tzen et al., 1993; Beisson et al., 2001b). Compared with other tree nuts, the almond lipid has a low amount of saturated fatty acids, but nonetheless it contains a significant proportion of poly‐ and monounsaturated fatty acids, with oleic acid being the predominant fatty acid (Robbins et al., 2011). Thus, depending on the harvest and variety, the kernel is made of approximately 50% of lipids of which 70–80% is oleic acid, 15% linoleic acid and 5% palmitic acid (Yada et al., 2011). Carbohydrates and dietary fibre The contents of available carbohydrates (i.e. mostly sugars) and dietary fibre (i.e. cell walls) in almond kernels are about 5.5% and 11.8%, respectively (Ellis et al., 2004). Little is known, however, about the structural organisation of the almond cell walls in any part of the kernel. Also, as highlighted by our group (Grassby et al., 2014) and others (McDougall et al., 1996; Waldron et al., 2003), each cell type of edible plant tissues, including almond tissue, has a distinct cell wall composition. Furthermore, the precise molecular composition and spatial arrangement of the polysaccharides and noncarbohydrates in almond cell walls have not been completely delineated. A number of compositional studies have found that the cell walls of almond kernel cotyledon, following hydrolysis of the cell wall polysaccharides and analysis by gas‐liquid chromatography, are rich in arabinose, uronic acid, glucose, xylose and galactose, which implies that the cell wall is composed of arabinose‐rich polysaccharides, including pectic material (Femenia et al., 2001; Dourado et al., 2004; Ellis et al., 2004). The cell walls of almond testa contain arabinose, galacturonic acid, glucose, xylose and galactose, but their proportions are different to those in the cotyledon and mannose, rhamnose and fucose are also part of their composition (Ellis et al., 2004; Mandalari et al., 2010a). Micronutrients and phytochemicals The almond kernel is rich in vitamins and minerals, and is considered as a good source of vitamin E (tocopherols), riboflavin, calcium, magnesium, phosphorus, potassium, zinc, copper and manganese (Rodushkin et al., 2008; Richardson et al., 2009) (Table 1). Almonds also contain a wide variety of phenolic compounds, mainly proanthocyanidins, flavonoids and phenolic acids (Perez‐Jimenez et al., 2010; Bolling et al., 2011; Xie et al., 2012), which are located predominantly in the skin and are responsible for their antioxidant properties (Mandalari et al., 2010b). Phytosterols are also found in significant amounts (~270 mg 100 g−1) in almond kernels, β‐sitosterol being the predominant type (Fernandez‐Cuesta et al., 2012; Alasalvar & Bolling, 2015; Forcada et al., 2015). Evidence suggests that the phytosterols reduce blood concentrations of LDL cholesterol and so these compounds may also contribute to the reduced risk of cardiovascular disease associated with consuming almonds (Plat & Mensink, 2005; Berryman et al., 2015). Sweet almond contains trace amounts (~0.2 to 16 mg 100 g−1 of almond) of amygdalin, a poisonous cyanogenic glycoside, whereas bitter almond has a high level of this glycoside (3300 to 5400 mg 100 g−1) (Lee et al., 2013). Processing techniques and their impact on almond structure Almonds are consumed predominantly in the raw, sliced or roasted forms, although marzipan as well as almond butter, milk and oil are also commonly found (Wareing et al., 2000; Gradziel, 2011). They are principally eaten as a snack but they can contribute to the composition of various sweet (e.g. breakfast cereals, cakes and biscuits) and savoury (e.g. salads, curries and tajines) dishes and food products. According to the Food and Agriculture Organization (FAO), the annual world production of almonds has been estimated to be about 1 930 000 metric tons of shelled product in 2012 (Food and Agriculture Organization of the United Nations, 2012). The main producing countries are the USA (California), Spain, Syria and Italy; California produces ~80% of the world's almonds (Harris & Ferguson, 2013). The main processing techniques applied to almond and their effect on the structure and the composition of the nut are summarised in Table 2. Table 2 Main processing techniques and their effects on the chemical composition, structure and properties of almonds Processing Effect on almond structure and composition References Roasting Water loss Cell wall damage Changes in the cytoplasmic network Loss in oil body integrity (i.e. lipid coalescence) Distortion and aggregation of protein bodies Browning of the almond tissue due to Maillard reaction Lipid uptake (when oil used during roasting) Altan et al. (2011) Grundy et al. (2015b) Mandalari et al. (2014) Pascual‐Albero et al. (1998) Perren & Escher (2013) Varela et al. (2006) Blanching Alteration in cytoplasmic organisation Skin removal which leads to loss in some micronutrients (e.g. phenolic compounds) Water uptake Altan et al. (2011) Mandalari et al. (2010a) Pascual‐Albero et al. (1998) Particle size reduction Rupture of cell walls particularly on the surface of the almond particle Release of some of the nutrients Grundy et al. (2015a) Oil extraction Degradation of the almond tissue to extract the oil Loss in oil body integrity Gallier et al. (2014) Kamal‐Eldin & Moreau (2009) John Wiley & Sons, LtdRoasting Roasting is a thermal process that involves dehydration (Perren & Escher, 2013). Almonds can be roasted in different ways (e.g. hot air vs. oil roasting, variations in heating times and duration) to obtain the light, medium or dark roast depending on the colour and moisture content of the resulting almonds. The roasting process has to be performed under well‐defined conditions in order to preserve the almond nutritional properties and prevent off‐flavour formation due to oxidation of unsaturated fatty acids. The roasted almonds used in our recent studies (Grassby et al., 2014; Mandalari et al., 2014; Grundy et al., 2015a,b, 2016) were provided by the Almond Board of California and were roasted using a two‐step standard procedure of hot air roasting with typical temperatures ranging from ~130 to 154 °C (Almond Board of California, 2007b). The first step employed an intermediate temperature to stabilise the nut microstructure, and the second step was performed at a higher temperature in order to generate the distinctive roasted flavour and brown colour of the cotyledon. Thus, during roasting, some of the moisture is lost by evaporation, and the Maillard reaction takes place, which is a complex reaction between reducing sugars and amino acids and is responsible for the brown colour (Perren & Escher, 2013). This nonenzymic browning enhances the antioxidant capacity of the roasted almond. The hot air roasting process was shown to lead to very little weight variation in whole almond kernels; most of the loss being attributed to water evaporation (Perren & Escher, 2013). The decrease in water content in roasted almonds has been reported to be between 40.7 to 59.1% of the original moisture content of the raw almonds (Altan et al., 2011). However, the oil bodies and the endoplasmic network were largely destroyed, and the volume of extracellular pores enlarged. Roasting can therefore greatly affect the structure of almond cells, the cell walls as well as the intra‐cellular oil bodies (Pascual‐Albero et al., 1998; Varela et al., 2006; Mandalari et al., 2014; Grundy et al., 2015a,b). In these studies, roasted almond oil bodies appeared to coalesce to form larger oil droplets than the ones observed in raw almond cells. During oil roasting, similar observations were made, but lipid uptake (ranging from 7.2 to 10.3%) from the oil used during roasting was also found to take place (Altan et al., 2011). Moreover, roasting is reported to reduce the polyphenol content of the almond skin and subsequently its antioxidant capacity (Bolling et al., 2010). In terms of its physical behaviour during mastication, roasted almonds were found to be more brittle and crunchy and produced more loose particles postchewing than whole raw almonds (Varela et al., 2008; Vickers et al., 2014). The attributes of roasted almonds described by Vickers and colleagues are likely to be due to the loss of moisture occurring during the roasting process. Blanching Similar to roasting, the blanching procedure decreases potential contamination, such as bacterial and mould growth, and consists of a thermal process that removes almond skin, using either wet or dry methods (Wareing et al., 2000). One of the wet methods used consists in peeling off the almond skins after the kernels are bathed in water at 85–100 °C for 2–5 min (Almond Board of California, 2007a). Kernels are dried by hot air, and then cooled down to room temperature. As highlighted above, almond skin is rich in flavonoids and other phenolic compounds, which confer the skin's antioxidant properties. Therefore, removing the skin reduces some of the nutritional attributes of the almond kernel (Garrido et al., 2008). Compared with roasted almonds, blanched almonds have a greater water content (Vickers et al., 2014). Both the roasting and blanching processes have been demonstrated to have no effect on the allergenicity of almond proteins (Venkatachalam et al., 2002). Particle size reduction Whole natural, blanched and roasted almond can be further processed to obtain almond particles of different shape and size (Fig. 3). For instance, almonds can be sliced, diced, chopped, ground or slivered (Wareing et al., 2000). These almond products differ in the proportion of intact and ruptured cells (Grundy et al., 2015b). Particles of smaller size have more fractured cells and thereby greater nutrient release (bioaccessibility) than larger particles. Figure 3 Photographs of ground almond particles with different size ranges. Scale bars = 1 cm. Almond paste, or marzipan, is a mixture of sugar and ground almond (Gradziel, 2011). It can be eaten on its own or, more commonly, as part of confectionary and cake. Almond butter has a rich, creamy texture and can be used as an alternative to diary butter. The term ‘nut butter’ refers to a butter made from a nut, such as almond, containing at least 90% of a nut compound, which can be produced in the form of particles (chunk and/or flour), paste, oil or a combination thereof (Wilkes, 2012; Gorrepati et al., 2015). Almond butter is obtained from either raw or roasted almonds, with or without their skins. Homogenisation and oil extraction Almond milk can be used as a plant‐based alternative to cow's milk for individuals suffering from lactose intolerance and allergy to cow milk proteins. Almond milk is a colloidal dispersion obtained by the physical disintegration, such as grinding, of almond kernels with water. Commercially available almond milk is often submitted to high pressure and heat treatment, which has consequences on its physical properties (i.e. particles/droplet sizes, rheology and protein structure) and therefore its stability (Bernat et al., 2015), but also on the allergenic potential of almond proteins (Dhakal et al., 2014). Therefore, even though the microstructure of the oil bodies within the almond milk is intact following grinding (Gallier et al., 2014), it appears that the monolayer of phospholipids and proteins is disrupted during the subsequent heat treatment (Bernat et al., 2015). Almond oil is usually extracted by applying cold pressing to the almond kernels (Kamal‐Eldin & Moreau, 2009). Solvent or supercritical fluid extractions are other methods used to extract the oil. The oil yield is higher with these chemical extraction techniques, but the quality of the oil (i.e. purity and presence of micronutrients) is lower than the ones obtained by cold‐press. The cold‐pressed almond oil has a light and pale amber colour (Almond Board of California, 2013). During the extraction, the oil bodies completely lose their integrity. The vitamin E and phenolic compounds contained in the oil inhibit its oxidation. Effects of processing and storage on almond quality Almonds that are consumed or used in the raw form (not roasted or blanched) are required to be pasteurised in the USA to remove any contaminant, in particular bacteria, mould and fungi. A water activity below 0.65 (~6% moisture content) is required to prevent growth of microorganisms when the almond is stored (Harris & Ferguson, 2013). Lipid oxidation results from the breakdown of lipid either by enzymic activity or reaction with the atmospheric oxygen (Lin et al., 2012). Therefore, exposure to light and elevation in moisture content can lead to lipid oxidation. The process is minimal in almonds when the water activity is in the range of 0.25 and 0.35 (~3–4% moisture content) (Almond Board of California, 2014). Processing will have an impact on the moisture content of the almond and could promote lipid oxidation. This could be prevented by using low temperature and low humidity storage conditions (Lin et al., 2012). The techniques and conditions employed for the processing of almond kernels, briefly described above, can affect its macro‐ and microstructures, which in turn can impact on the behaviour of almond tissue in the GI tract postingestion. Behaviour of whole and processed almonds in the GI tract and implications for macronutrient bioaccessibility, postprandial metabolism and gut microflora Digestion of whole, raw almonds It has been previously shown that it is mainly the first outer layer of cells of almond particles that fracture by mechanical trituration or chewing, so that most of the parenchyma cells of almonds remain intact and therefore contain encapsulated lipid and protein (Ellis et al., 2004; Grundy et al., 2015a). However, in a study in ileostomy volunteers, the lipids present in the intact cells located underneath the fractured layers, appeared to ‘leach’ from the intact cells, but only after a prolonged incubation in the upper GI tract (Mandalari et al., 2008a). Indeed, ingested raw almonds collected from ileostomy volunteers after 12 h digestion showed cells with thicker (swollen) cell walls (~1.2 μm) than after 2 h digestion (~0.6 μm) and undigested cells (0.1–0.2 μm). This swelling of the cell wall may explain why intact cells lose lipid after longer retention times, suggesting that lipase, colipase and bile salt could diffuse into the intracellular compartment and then initiate lipolysis. However, lipase does not seem to diffuse through the intact cell walls even after prolonged incubation times (up to 20 h) as demonstrated by in vitro digestion experiments performed on laboratory‐separated almond cells (Grundy et al., 2016). Nevertheless, in small particles of masticated almond, there was some evidence of rupture and fissures in ‘damaged cells’ underlying the fractured surface and this may account for the lipid release that occurs after prolonged incubation in the GI tract (Mandalari et al., 2008a; Grundy et al., 2015a). It has been suggested that the cell wall swelling is mainly attributed to the degradation and solubilisation of pectic compounds present in the cell wall and middle lamella, a process that potentially could increase porosity of the cells (Baron‐Epel et al., 1988; Femenia et al., 2001; Waldron et al., 2003). Nonetheless, it remains unclear to what extent lipolysis occurs inside almond cells and whether the lipids are able to leave the cells as TAG molecules or hydrolysed products. Even though lipase appears to be able to penetrate inside some cells, most likely the damaged ones, much of the lipid (TAG and/or lipolytic products) remains encapsulated inside the almond cells (Grundy et al., 2016). Regardless of the mechanism involved, the rate and extent of digestion of the lipids present in these unfractured cells is significantly reduced, as that they are less accessible to emulsification and digestion by the lipases (Grundy et al., 2015b). What is very clear from these almond studies is that the cells of the almond cotyledon behave in a fairly predictable way as they fracture rather than separate after chewing (Ellis et al., 2004) or by mechanical processing such as cutting and milling (Grassby et al., 2014), most likely due to their strong cell‐cell adhesion (Waldron et al., 2003). Therefore, mechanical processing (mainly grinding) or mastication is necessary for the cells to rupture and allow intra‐cellular lipid and other nutrients (e.g. proteins) to be made available for digestion. The released lipids seem to coalesce and form droplets (size ~10–40 μm) at the surface of the ruptured cells, thus becoming available for lipolysis by the lipases (Ellis et al., 2004). A digestion model that simulates the gastric environment provided contradictory information on the behaviour of almond particles in the digestive tract (Kong & Singh, 2009). Almond cells appeared to separate following the acidic hydrolysis of the middle lamella, which lessened the cell‐cell adhesion. The authors also detected the presence of breach and breakage in cell walls causing the release of nutrients into the extracellular environment and/or the penetration of enzyme and digestive components into the cells. Collection of faeces after ingestion of almond kernels revealed the presence of significant amounts of almond tissues (cotyledon and testa) (Ellis et al., 2004). Some of the cells were found intact, whereas others contained bacteria that seemed to be utilising (i.e. fermenting) both intracellular nutrients (including lipid) and cell wall polysaccharides (notably pectic substances). Indeed, the erosion of cell walls, the presence of virtually empty cells (i.e. no intra‐cellular nutrients) and apparent bacterial replication provide some evidence for the potential role of almonds as a source of nutrients for the gut microflora. Mandalari and colleagues have confirmed the prebiotic role of almonds and that the lipid components of almonds are susceptible to fermentation (Mandalari et al., 2008b). Moreover, since the lipids provide most of the energy contained in the almond, undigested lipids excreted in the faeces could have an impact on energy metabolism. Evidence to support this hypothesis is provided by measurements of the metabolisable energy content of almonds in healthy human subjects (Novotny et al., 2012). These findings indicate that the energy values of raw almonds, calculated using the conventional Atwater factor, overestimate the amount of energy actually absorbed. Digestion of processed almonds In vitro (Mandalari et al., 2008a) and in vivo (Berry et al., 2008) studies have revealed marked variations in lipolysis rates and postprandial blood TAG concentrations between meals containing different forms of almond (whole natural, blanched, milled flour and free oil), which are mainly attributed to differences in lipid release (bioaccessibility). In the oil form, lipids were highly available and therefore fully digested (leading to a high concentration of TAG in the blood), whereas encapsulated nutrients (whole almonds) did not lead to a postprandial response as rapid and strong as the almond oil (Berry et al., 2008). These results strengthen the assumption that by increasing the number of fractured cells through either processing or mastication, the bioaccessibility of nutrients, especially lipids, is enhanced. More recent studies have confirmed that almonds consumed as the whole kernel (raw or roasted) were not fully digested, and the lipids were released slowly during the digestion process (Grassby et al., 2014; Mandalari et al., 2014; Grundy et al., 2015a,b, 2016). This behaviour is strongly linked to the resistance of almond tissue/cell walls to chemical and physical breakdown in the mouth, stomach and small intestine. When the oil bodies are released from the almond tissue, as this is the case in almond milk, they are highly digestible and the rate and extent of lipolysis is similar to emulsified almond oil (Beisson et al., 2001a; Gallier & Singh, 2012; Grundy et al., 2016). If not in the form of oil bodies, almond oil, like any other edible plant oil, is required to be emulsified and its susceptibility to digestion relies on the size and interfacial quality (e.g. molecules adsorbed and surface tension) of the oil droplets (Gallier et al., 2014; Grundy et al., 2015b). In an experiment performed in the pig, a useful animal model for studies of digestion and postprandial metabolism, no differences in plasma glucose or lipid levels were found between raw and roasted almonds (Bornhorst et al., 2013a). However, the same authors reported that gastric emptying of protein in pigs was more rapid for raw as compared with roasted almonds due to protein segregation. In more recent studies, it was shown that although the masticated bolus of roasted almonds contained a higher proportion of particles of small size compared with raw almond bolus (Grundy et al., 2015a), there were negligible differences in lipid release in the gastric compartment (Mandalari et al., 2014) and the time course of lipid digestion during the duodenal phase (Grundy et al., 2015b) between the two almond forms. Another study performed in pigs showed no difference in particle sizes and rheological behaviour between raw and roasted almonds during gastric digestion (Bornhorst et al., 2013b). It was also recently reported by Gallier and colleagues that there was no variation in ileal lipid digestibility in rats fed either crushed whole almonds, almond oil emulsion or almond oil bodies (Gallier et al., 2014). This surprising result may be ascribed to the fact that the gastric emptying rate of raw almonds was slower than almond cream and oil, leaving enough time for the almond tissue to be degraded. Finally, quantitative and qualitative analysis of the carbohydrates that comprise the cell walls of digested, finely ground almonds revealed that they were not degraded during digestion; however, some of the intracellular content was fermented by the microorganisms originating from the human large intestine (Mandalari et al., 2008b). By comparing the growth of faecal bacteria cultures between almond kernels with normal lipid content and defatted ones, these authors also confirmed the assumption made by Ellis and colleagues that gut bacteria utilise almond lipids as a source of energy for growth and maintenance (Ellis et al., 2004). Conclusions The beneficial health effects of almonds rely not only on their nutritional composition, as they are a good source of unsaturated fatty acids, vitamin E, polyphenols and phytosterols, but also on their structure and properties when ingested. Differences in the physical form of ingested almonds in particular, lead to variability in nutrient digestibility and consequently evoke different blood nutrient profiles and gut hormone responses. The potential cardioprotective effects of almonds and their high satiety value reported in the literature suggest that they would make a healthy snack, especially when consumed as whole kernels. Energy values of raw almonds calculated using Atwater factors have been shown to be an overestimate of their actual metabolisable energy. This finding together with the results from the studies presented in this review raise important nutritional questions about the validity of energy content values found on food labels, which are based on food composition data and Atwater correction factors. Conflict of interest Karen Lapsley is an employee of the Almond Board of California. Acknowledgments We thank Dr Ellen Lever for the illustration seen in Fig. 1. Two of the authors (Myriam Grundy and Peter Ellis) thank the BBSRC, UK for funding some of the research work described in this review (BBSRC DRINC Project BB/H004866/1). Note 1 Fruit that possesses simultaneously fleshy (mesocarp or hull) and stony (endocarp or shell) layers surrounding the kernel Armstrong, (2009). ==== Refs References Alasalvar , C. & Bolling , B.W. (2015 ). Review of nut phytochemicals, fat‐soluble bioactives, antioxidant components and health effects . British Journal of Nutrition , 113 , S68 –S78 .26148924 Alasalvar , C. & Shahidi , F. (2009 ). Tree Nuts: Composition, Phytochemicals, and Health Effects . Boca Raton : CRC Press . Almond Board of California . (2007a ). Guidelines for validation of blanching processes [Online] . Available at: www.almonds.com/consumers/about-almonds/global-almond-usage (Accessed 02/12/2015). Almond Board of California . (2007b ). Hot Air Roasting of Almonds [Online] . Available at: www.almonds.com/consumers/about-almonds/global-almond-usage (Accessed 02/12/2015). Almond Board of California . (2013 ). Varieties [Online] . Available at: www.almondboard.com/Growers/OrchardManagement/Varieties/Pages/Default.aspx (Accessed 05/09/2013). Almond Board of California . (2014 ). Almond Shelf Life Factors [Online] . Available at: www.almonds.com/consumers/about-almonds/global-almond-usage (Accessed 02/12/2015). Altan , A. , McCarthy , K.L. , Tikekar , R. , McCarthy , M.J. & Nitin , N. (2011 ). Image analysis of microstructural changes in almond cotyledon as a result of processing . Journal of Food Science , 76 , E212 –E221 .21535761 Armstrong , W.P. (2009 ). Fruits Called Nuts [Online] . Available at: http://waynesword.palomar.edu/ecoph8.htm (Accessed 14/06/2011). Baron‐Epel , O. , Gharyal , P.K. & Schindler , M. (1988 ). Pectins as mediators of wall porosity in soybean cells . Planta , 175 , 389 –395 .24221876 Beisson , F. , Ferté , N. , Bruley , S. , Voultoury , R. , Verger , R. & Arondel , V. (2001a ). Oil‐bodies as substrates for lipolytic enzymes . Biochimica et Biophysica Acta – Molecular and Cell Biology of Lipids , 1531 , 47 –58 . Beisson , F. , Ferté , N. , Voultoury , R. & Arondel , V. (2001b ). Large scale purification of an almond oleosin using an organic solvent procedure . Plant Physiology and Biochemistry , 39 , 623 –630 . Bernat , N. , Cháfer , M. , Rodríguez‐García , J. , Chiralt , A. & González‐Martínez , C. (2015 ). Effect of high pressure homogenisation and heat treatment on physical properties and stability of almond and hazelnut milks . LWT – Food Science and Technology , 62 , 488 –496 . Berry , S.E. , Tydeman , E.A. , Lewis , H.B. et al (2008 ). Manipulation of lipid bioaccessibility of almond seeds influences postprandial lipemia in healthy human subjects . American Journal of Clinical Nutrition , 88 , 922 –929 .18842777 Berryman , C.E. , West , S.G. , Fleming , J.A. , Bordi , P.L. & Kris‐Etherton , P.M. (2015 ). Effects of daily almond consumption on cardiometabolic risk and abdominal adiposity in healthy adults with elevated LDL‐cholesterol: a randomized controlled trial . Journal of the American Heart Association , 4 , e000993 .25559009 Bolling , B.W. , Blumberg , J.B. & Chen , O.C.Y. (2010 ). The influence of roasting, pasteurisation, and storage on the polyphenol content and antioxidant capacity of California almond skins . Food Chemistry , 123 , 1040 –1047 . Bolling , B.W. , Chen , C.Y. , McKay , D.L. & Blumberg , J.B. (2011 ). Tree nut phytochemicals: composition, antioxidant capacity, bioactivity, impact factors. A systematic review of almonds, Brazils, cashews, hazelnuts, macadamias, pecans, pine nuts, pistachios and walnuts . Nutrition Research Reviews , 24 , 244 –275 .22153059 Bornhorst , G.M. , Roman , M.J. , Dreschler , K.C. & Singh , R.P. (2013a ). Physical property changes in raw and roasted almonds during gastric digestion in vivo and in vitro . Food Biophysics , 9 , 39 –48 . Bornhorst , G.M. , Roman , M.J. , Rutherfurd , S.M. , Burri , B.J. , Moughan , P.J. & Singh , R.P. (2013b ). Gastric digestion of raw and roasted almonds in vivo . Journal of Food Science , 78 , H1807 –H1813 .24245891 Dhakal , S. , Liu , C. , Zhang , Y. , Roux , K.H. , Sathe , S.K. & Balasubramaniam , V.M. (2014 ). Effect of high pressure processing on the immunoreactivity of almond milk . Food Research International , 62 , 215 –222 . Dourado , F. , Barros , A. , Mota , M. , Coimbra , M.A. & Gama , F.M. (2004 ). Anatomy and cell wall polysaccharides of almond (Prunus dulcis D. A. Webb) seeds . Journal of Agricultural and Food Chemistry , 52 , 1364 –1370 .14995147 Ellis , P.R. , Kendall , C.W. , Ren , Y. et al (2004 ). Role of cell walls in the bioaccessibility of lipids in almond seeds . American Journal of Clinical Nutrition , 80 , 604 –613 .15321799 Femenia , A. , Garcia‐Marin , M. , Simal , S. , Rossello , C. & Blasco , M. (2001 ). Effects of supercritical carbon dioxide (SC‐CO2) oil extraction on the cell wall composition of almond fruits . Journal of Agricultural and Food Chemistry , 49 , 5828 –5834 .11743770 Fernandez‐Cuesta , A. , Kodad , O. & Velasco , L. (2012 ). Phytosterol variability in almond germplasm . Journal of the American Society for Horticultural Science , 137 , 343 –348 . Food and Agriculture Organization of the United Nations . (2012 ). FAOSTAT [Online] . Available at: www.faostat.fao.org (Accessed 06/05/2016). Forcada , I.F.C. , Velasco , L. , Company , R.S.I. & Marti , A.F.I. (2015 ). Association mapping for kernel phytosterol content in almond . Frontiers in Plant Science , 6 , 530 .26217374 Gallier , S. & Singh , H. (2012 ). Behavior of almond oil bodies during in vitro gastric and intestinal digestion . Food & Function , 3 , 547 –555 .22354453 Gallier , S. , Rutherfurd , S.M. , Moughan , P. & Singh , H. (2014 ). Effect of food matrix microstructure on stomach emptying rate and apparent ileal fatty acid digestibility of almond lipids . Food & Function , 5 , 2381 –2672 . Garrido , I. , Monagas , M. , Gomez‐Cordoves , C. & Bartolome , B. (2008 ). Polyphenols and antioxidant properties of almond skins: influence of industrial processing . Journal of Food Science , 73 , C106 –C115 .18298714 Gorrepati , K. , Balasubramanian , S. & Chandra , P. (2015 ). Plant based butters . Journal of Food Science and Technology , 52 , 3965 –3976 .26139864 Gradziel , T.M. (2011 ). Origin and dissemination of almond In: Horticultural Reviews (edited by Janick J. ). Pp. 23 –81 . Hoboken : John Wiley & Sons, Inc . Grassby , T. , Picout , D.R. , Mandalari , G. et al (2014 ). Modelling of nutrient bioaccessibility in almond seeds based on the fracture properties of their cell walls . Food & Function , 5 , 3096 –3106 .25310222 Grundy , M.M.‐L. , Grassby , T. , Mandalari , G. et al (2015a ). Effect of mastication on lipid bioaccessibility of almonds in a randomized human study and its implications for digestion kinetics, metabolizable energy, and postprandial lipemia . American Journal of Clinical Nutrition , 101 , 25 –33 .25527747 Grundy , M.M.‐L. , Wilde , P.J. , Butterworth , P.J. , Gray , R. & Ellis , P.R. (2015b ). Impact of cell wall encapsulation of almonds on in vitro duodenal lipolysis . Food Chemistry , 185 , 405 –412 .25952886 Grundy , M.M.‐L. , Carriere , F. , Mackie , A.R. , Gray , D.A. , Butterworth , P.J. & Ellis , P.R. (2016 ). The role of plant cell wall encapsulation and porosity in regulating lipolysis during the digestion of almond seeds . Food & Function , 7 , 69 –78 .26503478 Harris , L.J. & Ferguson , L. (2013 ). Improving the safety of almonds and pistachios In: Improving the Safety and Quality of Nuts (edited by Harris L.J. ). Pp. 350 –378 Cambridge : Woodhead Publishing . Kamal‐Eldin , A. & Moreau , R.A. (2009 ). Tree nut oils In: Gourmet and Health‐Promoting Speciality Oils (edited by Moreau R.A. & Kamal‐Eldin A. ). Pp. 127 –149 Urbana, USA : AOCS Press . Kong , F. & Singh , R.P. (2009 ). Digestion of raw and roasted almonds in simulated gastric environment . Food Biophysics , 4 , 365 –377 . Kshirsagar , H.H. , Fajer , P. , Sharma , G.M. , Roux , K.H. & Sathe , S.K. (2011 ). Biochemical and spectroscopic characterization of almond and cashew nut seed 11S legumins, amandin and anacardein . Journal of Agricultural and Food Chemistry , 59 , 386 –393 .21138244 Lee , J. , Zhang , G. , Wood , E. , Rogel Castillo , C. & Mitchell , A.E. (2013 ). Quantification of amygdalin in nonbitter, semibitter, and bitter almonds (Prunus dulcis) by UHPLC‐(ESI)QqQ MS/MS . Journal of Agricultural and Food Chemistry , 61 , 7754 –7759 .23862656 Lin , X. , Wu , J. , Zhu , R. et al (2012 ). California almond shelf life: lipid deterioration during storage . Journal of Food Science , 77 , C583 –C593 .22582861 Mandalari , G. , Faulks , R.M. , Rich , G.T. et al (2008a ). Release of protein, lipid and vitamin E from almonds seeds during digestion . Journal of Agricultural and Food Chemistry , 56 , 3406 –3416 . Mandalari , G. , Nueno‐Palop , C. , Bisignano , G. , Wickham , M.S. & Narbad , A. (2008b ). Potential prebiotic properties of almond (Amygdalus communis L.) seeds . Applied and Environmental Microbiology , 74 , 4264 –4270 .18502914 Mandalari , G. , Tomaino , A. , Arcoraci , T. et al (2010a ). Characterization of polyphenols, lipids and dietary fibre from almond skins (Amygdalus communis L.) . Journal of Food Composition and Analysis , 23 , 166 –174 . Mandalari , G. , Tomaino , A. , Rich , G.T. et al (2010b ). Polyphenol and nutrient release from skin of almonds during simulated human digestion . Food Chemistry , 122 , 1083 –1088 . Mandalari , G. , Grundy , M.M.‐L. , Grassby , T. et al (2014 ). The effects of processing and mastication on almond lipid bioaccessibility using novel methods of in vitro digestion modelling and micro‐structural analysis . British Journal of Nutrition , 112 , 1521 –1529 .25351860 McDougall , G.J. , Morrison , I.M. , Stewart , D. & Hillman , J.R. (1996 ). Plant cell walls as dietary fibre: range, structure, processing and function . Journal of the Science of Food and Agriculture , 70 , 133 –150 . Nishi , S. , Kendall , C.W. , Gascoyne , A.M. et al (2014 ). Effect of almond consumption on the serum fatty acid profile: a dose‐response study . British Journal of Nutrition , 112 , 1137 –1146 .25138064 Novotny , J.A. , Gebauer , S.K. & Baer , D.J. (2012 ). Discrepancy between the Atwater factor predicted and empirically measured energy values of almonds in human diets . American Journal of Clinical Nutrition , 96 , 296 –301 .22760558 Osborne , T.B. & Campbell , G.F. (1896 ). Conglutin and vitellin . Journal of the American Chemical Society , 18 , 609 –623 . Pascual‐Albero , M.J. , Perez‐Munuera , I. & Lluch , M.A. (1998 ). Cotyledon structure of raw, soaked and roasted almond (Prunus amygdalus L.) . Food Science and Technology International , 4 , 189 –197 . Perez‐Jimenez , J. , Neveu , V. , Vos , F. & Scalbert , A. (2010 ). Identification of the 100 richest dietary sources of polyphenols: an application of the Phenol‐Explorer database . European Journal of Clinical Nutrition , 64 , S112 –S120 .21045839 Perren , R. & Escher , F.E. (2013 ). Impact of roasting on nut quality In: Improving the Safety and Quality of Nuts . (edited by Harris L.J. ). Pp. 173 –197 Cambridge : Woodhead Publishing . Plat , J. & Mensink , R.P. (2005 ). Plant stanol and sterol esters in the control of blood cholesterol levels: mechanism and safety aspects . American Journal of Cardiology , 96 , 15d –22d . Richardson , D.P. , Astrup , A. , Cocaul , A. & Ellis , R.P. (2009 ). The nutritional and health benefits of almonds: a healthy food choice . Food Science and Technology Bulletin: Functional Foods , 6 , 41 –50 . Robbins , K.S. , Shin , E.C. , Shewfelt , R.L. , Eitenmiller , R.R. & Pegg , R.B. (2011 ). Update on the healthful lipid constituents of commercially important tree nuts . Journal of Agricultural and Food Chemistry , 59 , 12083 –12092 .21985331 Rodushkin , I. , Engstrom , E. , Sorlin , D. & Baxter , D. (2008 ). Levels of inorganic constituents in raw nuts and seeds on the Swedish market . Science of the Total Environment , 392 , 290 –304 .18155749 Sathe , S.K. , Wolf , W.J. , Roux , K.H. , Teuber , S.S. , Venkatachalam , M. & Sze‐Tao , K.W.C. (2002 ). Biochemical characterization of amandin, the major storage protein in almond (Prunus dulcis L.) . Journal of Agricultural and Food Chemistry , 50 , 4333 –4341 .12105967 Tan , S.Y. & Mattes , R.D. (2013 ). Appetitive, dietary and health effects of almonds consumed with meals or as snacks: a randomized, controlled trial . European Journal of Clinical Nutrition , 67 , 1205 –1214 .24084509 Tzen , J.T.C. , Cao , Y.Z. , Laurent , P. , Ratnayake , C. & Huang , A.H.C. (1993 ). Lipids, proteins, and structure of seed oil bodies from diverse species . Plant Physiology , 101 , 267 –276 .12231682 USDA . (2015 ). National Nutrient Database for Standard Reference Release, Release 28 [Online] . Available at: http://ndb.nal.usda.gov/ndb/foods/show/3635?fgcd=&manu=&lfacet=&format=&count=&max=35&offset=&sort=&qlookup=almond (Accessed 15/02/2016). Varela , P. , Chen , J. , Fiszman , S. & Povey , M.J.W. (2006 ). Crispness assessment of roasted almonds by an integrated approach to texture description: texture, acoustics, sensory and structure . Journal of Chemometrics , 20 , 311 –320 . Varela , P. , Salvador , A. & Fisman , S. (2008 ). On the assessment of fracture in brittle foods: the case of roasted almonds . Food Research International , 41 , 544 –551 . Venkatachalam , M. , Teuber , S.S. , Roux , K.H. & Sathe , S.K. (2002 ). Effects of roasting, blanching, autoclaving, and microwave heating on antigenicity of almond (Prunus dulcis L.) proteins . Journal of Agricultural and Food Chemistry , 50 , 3544 –3548 .12033826 Vickers , Z. , Peck , A. , Labuza , T. & Huang , G. (2014 ). Impact of almond form and moisture content on texture attributes and acceptability . Journal of Food Science , 79 , S1399 –S1406 .24962063 Waldron , K.W. , Parker , M.L. & Smith , A.C. (2003 ). Plant cell walls and food quality . Comprehensive Reviews in Food Science and Food Safety , 2 , 101 –119 . Wareing , P.W. , Nicolaides , L. & Twiddy , D.R. (2000 ). Nuts and nut products In: The Microbiological Safety and Quality of Food . (edited by Lund B.M. , Baird‐Parker T.C. & Gould G.W. ). Pp. 919 –940 Gaithersburg, USA : Aspen Publishers . Wilkes , R.S. (2012 ). Nut butter and related products enriched with omega‐3 . US patent application. Willison , L.N. , Zhang , Q. , Su , M. , Teuber , S.S. , Sathe , S.K. & Roux , K.H. (2013 ). Conformational epitope mapping of Pru du 6, a major allergen from almond nut . Molecular Immunology , 55 , 253 –263 .23498967 Winton , A.L. & Winton , K.B. (1932 ). The Structure and Composition of Foods. Volume I: Cereals, Starch, Oil Seeds, Nuts, Oils, Forage Plants . New York , John Wiley & Sons . Xie , L. , Roto , A.V. & Bolling , B.W. (2012 ). Characterization of ellagitannins, gallotannins, and bound proanthocyanidins from California almond (Prunus dulcis) varieties . Journal of Agricultural and Food Chemistry , 60 , 12151 –12156 .23167850 Yada , S. , Lapsley , K. & Huang , G.W. (2011 ). A review of composition studies of cultivated almonds: macronutrients and micronutrients . Journal of Food Composition and Analysis , 24 , 469 –480 . Young , C.T. , Schadel , W.E. , Pattee , H.E. & Sanders , T.H. (2004 ). The microstructure of almond (Prunus dulcis (Mill.) D.A. Webb cv. ‘Nonpareil’) cotyledon . LWT – Food Science and Technology , 37 , 317 –322 .

PMC005xxxxxx/PMC5003317.txt

==== Front 10166641444247J Ga Public Health AssocJ Ga Public Health AssocJournal of the Georgia Public Health Association2471-977310.21633/jgpha.6.103nihpa811242ArticleFormative research to develop a lifestyle application (app) for African American breast cancer survivors Smith Selina A. PhD, MDiv12Whitehead Mary S. MPH, CHES3Sheats Joyce Q. RN, MPH1Fontenot Brittney MPH1Alema-Mensah Ernest PhD, DMin4Ansa Benjamin MD, MSCR11 Institute of Public and Preventive Health, Augusta University, Augusta, GA2 Department of Family Medicine, Medical College of Georgia, Augusta University, Augusta, GA3 SISTAAH Talk Breast Cancer Support Group, Mami, FL4 Department of Community Health and Preventive Medicine, Morehouse School of Medicine, Atlanta, GACorresponding Author: Dr. Benjamin Ansa, CJ2300 1120 15th Street, Augusta, GA 30912, 706-721-6141, BANSA@augusta.edu19 8 2016 Summer 2016 29 8 2016 6 1 50 59 This is an open-access article distributed under the terms of the Creative Commons Attribution Non-Commercial No-Derivatives License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work ("first published in the Journal of the Georgia Public Health Association…") is properly cited with original URL and bibliographic citation information. The complete bibliographic information, a link to the original publication on http://www.gapha.jgpha.org/, as well as this copyright and license information must be included.Background There is a proliferation of lifestyle-oriented mobile technologies; however, few have targeted users. Through intervention mapping, investigators and community partners completed Steps 1–3 (needs assessment, formulation of change objectives, and selection of theory-based methods) of a process to develop a mobile cancer prevention application (app) for cancer prevention. The aim of this qualitative study was to complete Step 4 (intervention development) by eliciting input from African American (AA) breast cancer survivors (BCSs) to guide app development. Methods Four focus group discussions (n=60) and three individual semi-structured interviews (n=36) were conducted with AA BCSs (40–72 years of age) to assess barriers and strategies for lifestyle change. All focus groups and interviews were recorded and transcribed verbatim. Data were analyzed with NVivo qualitative data analysis software version 10, allowing categories, themes, and patterns to emerge. Results Three categories and related themes emerged from the analysis: 1) perceptions about modifiable risk factors; 2) strategies related to adherence to cancer prevention guidelines; and 3) app components to address barriers to adherence. Participant perceptions, strategies, and recommended components guided development of the app. Conclusions For development of a mobile cancer prevention app, these findings will assist investigators in targeting features that are usable, acceptable, and accessible for AA BCSs. Lifestyle modificationintervention mappingcancer prevention guidelinesbreast cancer survivorssmartphone app ==== Body INTRODUCTION Among African American (AA) women in the U.S., breast cancer is the most commonly diagnosed cancer, and the second most common cause of cancer death. In 2016, an estimated 30,700 new cases and 6,310 deaths from breast cancer are expected to occur among AA women (American Cancer Society (ACS), 2016). Incidence rates continue to increase in AA women at about 0.5% per year, and death rates are 42% higher compared to White women (ACS, 2016). The increase in incidence and mortality rates in AA women may in part reflect the rising prevalence of obesity in this group. AA women have the highest rates of being overweight or obese compared to other ethnic groups in the U.S. Recent national data show that 82.0% of Black women are overweight or obese compared to 63.2% of White women (National Center for Health Statistics ((NCHS), 2016). Obesity is a risk factor for breast cancer recurrence and poor survival (Strickler, 2013). Healthy lifestyles (e.g., keeping physically active and healthy diet) are widely recognized as contributors to cancer survivorship. Health organizations such as the American Cancer Society (ACS) and the World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) routinely publish dietary and lifestyle recommendations aimed at cancer prevention (Kushi et al., 2012; WCRF/AICR, 2007). Both the ACS guidelines and the WCRF/AICR recommendations are concerned with maintaining a healthy weight throughout life, consuming a plant-based diet, adopting a physically active lifestyle, and limiting consumption of red meat and alcohol. Adherence to the ACS dietary guidelines and the WCRF/AICR guidelines is associated with a decrease in the incidence of cancer and mortality (McCullough et al., 2011; Hastert et al., 2013). Current approaches aimed at modifying lifestyle risk behaviors appear to be inadequate for producing sustained adherence to cancer prevention recommendations (Lyons et al., 2016). As smartphones become more available, they provide a promising means of implementing and disseminating cancer prevention interventions to a wide range of individuals at a low cost. In the USA, 68.0% of adults owned a smartphone in 2015. Among ethnic groups, AAs, a population that is disproportionately affected by obesity and breast cancer disparities, have the highest smartphone ownership (68.0%) (Pew Research Center, 2015). There has been a proliferation of lifestyle-oriented mobile apps, however, few have targeted users. Conducted in three phases, the present study was designed to develop an educational intervention to promote adherence to cancer prevention recommendations among AA breast cancer survivors (BCSs) and to evaluate the feasibility and acceptability of delivering the intervention through a smartphone app. The study phases include: 1) a mixed-methods assessment of the lifestyle needs and experiences of the targeted population (2013–2015); 2) a qualitative study, which is the subject of this report, to elicit stakeholder input to guide intervention development (2015–2016); and 3) a feasibility study piloting the app to assess its usability (2018). METHODS The research protocol, A Community-Engaged Approach to Developing a Mobile Cancer Prevention App (mCPA), has been described elsewhere (Smith et al., 2016). Briefly, the study engaged members of a breast cancer support group, Survivors Involving Supporters to Take Action in Advancing Health (SISTAAH) Talk, in a partnership to create content for a smartphone app. The outcome of this study will be a theory-based, culturally tailored, mobile app for testing and future dissemination and implementation. Mapping Intervention Development Framework Through intervention mapping, investigators partnered with the SISTAAH Talk breast cancer support group to conduct the study. Consistent with community-based participatory research (CBPR) principles (Smith et al., 2015), intervention mapping was selected as a framework to guide development of the intervention (Corbie-Smith et al., 2010). Intervention mapping is an iterative process that begins with a needs assessment and continues by fostering collaborations with stakeholders during intervention development, implementation, and evaluation (McEachan et al., 2008). This systematic process combines theory, empirical scientific literature, and community data to develop culturally appropriate interventions (Bartholomew et al., 1998). The SISTAAH Talk app heuristic intervention map (Figure 1) outlines an iterative process that includes six steps. Step 1, Needs Assessment, included a literature review, secondary data analysis, lifestyle assessment, and focus group discussions. The literature review established that, although lifestyle changes can reduce the risk of recurrence by one half and the risk of breast-cancer associated mortality by one third (Howell et al., 2014), many BCSs do not engage in such strategies. Lifestyle factor data from the Behavioral Risk Factor Surveillance System (BRFSS) revealed three disparity risk categories for AA women: 1) obesity (35.7% vs. 23.7% for whites); 2) inadequate fruit and vegetable consumption (12.6% vs. 17.4% for whites); and 3) physical inactivity (63.8% vs. 50.3% for whites) (Centers for Disease Control and Prevention (CDC), 2015). In a secondary data analysis of the National Health Interview Survey Cancer Control Supplement, the health-related quality of life for female AAs aged 35 and older (n=62) was compared to AA female survivors of other cancers (SOCs) (n=74) and to AA women of the same age with no history of cancer (NHCs) (n=1,566) (Smith et al., 2016; Claridy et al., 2016). There were no statistically significant differences between BCSs and NHCs, but SOCs reported poorer mental health relative to NHCs [3.3 t-points 95% CI 0.6 – 5.9]. A comparison of differences between SOCs and NHCs showed that there were three physical health items in which SOCs were more likely to report poorer physical health relative to NHCs (ability to carry out physical activities [OR=3.4; 95% CI 1.7 – 6.7], level of fatigue [OR=2.0; 95% CI 1.1 – 3.7], and level of pain [OR=3.3; 95% CI 1.3 – 3.9]). AA BCSs (n=240; mean age: 56.9 years; standard deviation [SD]: 11.8; range: 25–92 years) completed a lifestyle assessment survey. More than half were overweight/obese (68.0%); did not limit portion sizes to control weight (89.0%); consumed <5 vegetables and fruits/day (75.0%); and had >5 servings of red (75.0%) and processed meats/week (94.0%). Participants in four focus group discussions (n = 42; mean age: 45.7 years; standard deviation [SD]: 7.9; range: 35–75 years old) identified barriers to and intervention approaches for enhancing dietary intake. Themes emerging from content analysis converged into the following categories: “talk” as central; peer-facilitated sessions; support group approach; no “pamphlet only” control group; “hands on” or interactive nutrition education; supporters (co-survivors); and community-based (not “community placed”) research (Smith et al., 2015). Step 2, Formulation of Change Objectives was accomplished through contributions from a Community Advisory Board (CAB), with representatives from community coalitions and breast cancer support groups in Miami, Chicago, Houston, Los Angeles, and Philadelphia (Table 1). Employing the iterative characteristic of CBPR, the CAB formulated objectives for lifestyle changes based on results from the previously described needs assessment and AICR prevention guidelines for cancer survivors (WCRF/AICR, 2007). These include: 1) Be as lean as possible without becoming underweight; 2) Be physically active for at least 30 minutes every day to help prevent cancer and prevent recurrence of cancer; 3) Avoid sugary drinks and limit consumption of energy-dense foods (particularly processed foods high in added sugar, low in fiber, or high in fat); 4) Eat more of a variety of vegetables, fruits, whole grains, and legumes such as beans, taking up at least 2/3 of the space on your plate; 5) Limit consumption of red meats (such as beef, pork, and lamb) and avoid processed meats, taking up only 1/3 or less of your plate; 6) If consumed at all, limit alcoholic drinks to two for men and one for women a day; and 7) Limit consumption of salty foods and foods processed with salt (sodium) by substituting herbs and spices high in phytochemicals (e.g., basil, turmeric, paprika, thyme, and dill). For Step 3, Theory-Based Methods, the health belief model (HBM) and theory of planned behavior (TPB) were selected to undergird the SISTAAH Talk app. The HBM suggests that health-related cognitions for determining behavior considers belief of BCSs that lifestyle behaviors affect breast cancer recurrence, how severe the recurrence would be, and the cost/benefits of lifestyle change (Glanz et al., 2008). The TPB posits that health behavior is affected by past breast cancer experience and social norms (i.e., lifestyle practices) more so than beliefs (i.e., a link between breast cancer recurrence and lifestyle) (Ajzen, 1991). Components of the HBM and TPB that support app features are outlined in Table 2. Step 4, Intervention Development, the subject of this report, included participatory engagement. SISTAAH Talk members reviewed themes from focus group discussions during three telephone support group meetings with the principal investigator. They then engaged with researchers in videotaped and audiotaped experiential cooking and exercise sessions, which will serve as content for the app. A feasibility trial and research-to-practice demonstration (Step 5, Adoption and Implementation) and Evaluation (Step 6), assessed through the Reach, Efficacy/Effectiveness, Adoption, Implementation, and Maintenance (RE-AIM) framework (Glasgow et al., 2004), will be completed. Investigators and participants anticipate development of an acceptable, feasible, and accessible mobile app, which will be available for intervention testing in 2018. Study Population Established in 1995 at the University of Miami Sylvester Comprehensive Cancer Center as the first breast cancer support group for women of color in South Florida, SISTAAH Talk has a goal of providing a forum for African-Americans to communicate about and make sense of their diagnosis and treatment in order to achieve improved physical and mental health outcomes. SISTAAH Talk includes women from across Miami-Dade and Broward counties, reaching an average of 20 women each month through education, outreach, and research. A purposefully selected sample of 12 SISTAAH Talk members, treated for >1 year for Stages I-IIIc breast cancer, 75 years of age or younger, and English speaking/writing, were identified by leaders of the support group as role models or “coaches” to participate in developing mCPA content. Each applicant was interviewed by the principal investigator and support group facilitator to determine their comfort level in participating in focus group discussions, semi-structured interviews, and development of app content (e.g., video taping cooking demonstrations and physical activity). Procedures The Institutional Review Board of Augusta University approved this research plan. Following informed consent, SISTAAH Talk coaches engaged in experiential educational sessions (e.g., cooking and exercise) and audio- and videotaped facilitated discussions. Cooking demonstrations followed cancer prevention guidelines relative to nutrition and dietary intake (e.g., portion control and weight control, vegetables and fruits, red and processed meats, and whole grains). Exercise sessions focused on guidelines specific to physical activity (e.g., 150 minutes per week). Qualitative methods were used to determine preferences related to cancer prevention. Following the experiential educational sessions, focus group discussions (FGDs) were completed to generate ideas and determine app content. Individual semi-structured interviews (SSIs), which included an open set of questions that allowed new ideas to evolve based on participant responses, were completed to determine barriers to lifestyle modification, potential strategies to address them, and desired app components. The principal investigator, a BCS experienced in qualitative data collection methods, facilitated the FGDs and SSIs. Sample size was determined based on the principle of saturation, which suggests that, with as few as four discussions, no additional information will be obtained (Carlsen and Glenton, 2011). This qualitative sampling technique was used to ensure that perspectives across age groups were obtained. Six 90-minute FGDs were conducted with the SISTAAH Talk coaches. A semi-structured and open-ended FGD guide was used to ensure adequate content coverage. The interview guide explored: a) perceptions about modifiable risk factors; b) feelings related to inclusion of cancer prevention guidelines in a smartphone app; and c) strategies for lifestyle modification. Following the three experiential educational sessions, thirty-six 30-minute SSIs were completed with the SISTAAH Talk coaches to assess perceptions about dietary intake and physical activity, to determine strategies for their modification, and to garner users’ preferred components for the app. To learn more about the content of lifestyle apps, SISTAAH Talk coaches reviewed existing smartphone apps. They were then asked during SSIs to evaluate salient features of the apps to determine which, if any, assisted in addressing individual barriers to lifestyle behavior change. Topic guides for group discussions and individual interviews appear in the Appendix. Prior to their use, the SISTAAH Talk leadership ensured that the FGD and SSI interview guides were written at an appropriate comprehension level. The guides were pilot-tested for appropriateness and language accuracy. Each discussion and interview was digitally recorded, transcribed verbatim, manually coded, and summarized. The FGD and SSI process ended at the point of saturation or when the collection of new qualitative information no longer shed light on the issues under investigation. Theoretical Approach and Data Analysis Data analysis was guided by a grounded-theory approach to allow categories, themes, and patterns to emerge. This tactic, which uses an inductive method to generate substantive ideas that suggest more focused research questions (Corbin and Strauss, 2008), served as the theoretical basis to formative data analysis. Data were analyzed using Qualitative Content Analysis (Schreier, 2012). Coding steps included developing preliminary themes, creating additional codes based on themes that arose, developing non-substantive codes, and producing detailed codes for analysis of specific topics. NVIVO 10 (qualitative data analysis computer software) was used to facilitate the coding process (i.e., assessing the degree of agreement/disagreement across themes and calculating inter-rater reliability scores) (2015). Recurring themes were identified, the research team came to a consensus on coded themes, and themes were summarized for analysis. RESULTS A total of 12 SISTAAH Talk coaches participated in the study (Table 3). Participants were more than 50 years old (60.0%), college educated (75.0%), widowed/divorced (42.0%), and earned an annual income $25,000 or higher (75.0%). The mean period since breast cancer diagnosis was 8.7years. Data were generated from FGDs and SSIs. Recurring themes that emerged were organized into three categories: 1) perceptions about modifiable risk factors; 2) strategies related to cancer prevention guideline adherence; and 3) app components to address barriers to lifestyle modification. Perceptions about modifiable risk factors In general, FGD participants had a clear concept and understanding of the cancer prevention guidelines specific to dietary intake and physical activity. Dietary intake Racial-ethnic disparities in modifiable breast cancer risk factors (obesity and low consumption of fruits and vegetables) are large and persistent, particularly between White and AA women (CDC, 2015). The comments of one BCS reflects this understanding: “I have to say that as cancer survivors, we have the choice of making our lives better. Like we have the choice of making decision for ourselves, that’s really a benefit for me. I feel like choosing to eat well; to eat less meat, more vegetables and fruits I think it’s best for us. I enjoy eating well because healthy food tastes really well.” The prevalence of overweight or obesity among AA women is 82% relative to 63% for white women (NCHS, 2016). Obesity and weight gain after breast cancer diagnosis are associated with poorer outcomes, including decreased quality of life, increased recurrence, breast cancer deaths, and all-cause mortality (McCollough et al., 2011). For overweight and obese women, a sustained weight loss of 10% of initial weight reduces risk of recurrence of a new primary breast cancer. Comments by one participant reflect this perception: “Well, it’s something that our people have a big problem with. Sometimes when I tell people I am trying to make a change, they say, ‘why you doing that?’ Let’s face it, I am not sure that I can do anything about it, but I think if I do, it could save my life.” Adherence to diet-related cancer prevention guidelines is relevant to primary prevention and recurrence of breast cancer (ACS, 2016). Limiting consumption of red and processed meats is recommended. Participants shared the following comments: “I was a little surprised to learn that even with all the talk about how red meat affects cancer cells, that people still eat so much of it.” “I was saying if I could just try to cut out the red meat because it does not break down in our bodies for two or three days, which makes me constipated, it would help.” “Processed meats, like sausage, are cheap and you can cook them fast. But, they contain a lot of salt, a lot of fats. When we are buying those things from the store, we have no idea of what’s in them.” Consuming five or more fruit and vegetable servings daily is recommended for cancer prevention (ACS, 2016). Participants’ perceptions related to this guideline are captured in the following comments: “One of the things that we can do to eat healthy is eat more vegetables and fruits.” “I woke up one morning after chemo and was really sick to my stomach. I did a lot of research and decided to gradually start to eat more plant-based foods. Two weeks turned into two months and now, I’ve been doing it for two years. After one month, I could just feel my body losing all this fat, like it was melting.” Challenges to meeting this recommendation were reflected in these statements: “I don’t have trouble eating more vegetables. The problem is not adding butter, gravy, processed cheese, and those things that make them taste good.” “People generally get offended when you say ‘no thank you’ to food. It’s hard to say no to vegetables cooked with added meat.” Replacing refined grains with whole grains, portion control, avoiding hidden sugar, and the limited sodium content of food are components of the cancer prevention guidelines. Two statements from a participant illustrate this understanding: “I have to remember to eat smaller meals, more often. That’s what works for me. I need to learn how to use herbs because they definitely make a difference in how food tastes.” “I would say cooking and eating well is pretty simple. You don’t have to have a lot of ingredients to eat well. It’s not about how much you eat, it’s how well it is prepared and how good it is for you.” Barriers to adherence to diet-related cancer prevention guidelines were related to access to healthy foods, costs, and cultural food preferences. Physical Activity AA BCSs are less likely to report adherence to physical activity guidelines than women of other races (ACS, 2016). Compared to other races/ethnic groups, AA women have the largest potential reductions in breast cancer risk from physical activity (41%) (Friedenreich, 2010), yet they are generally less physically active and more sedentary than women of other racial/ethnic groups. In the present study, the recommendation of 150 minutes of moderate physical activity per week was well received, as reflected in the following: “Walking not only helps you physically, you are able to release emotionally, you’re able to share and feel alive” “The first thing people talk about is fatigue and the lack of energy so by walking and partnering up, you can do it. The challenge is to get out and walk and to have a good time.” “Walking is aerobic; your heart rate goes up.” The primary barrier to physical activity for FGD participants was limitations due to physical outcomes related to treatment (e.g., lymphedema (swelling in an arm), arthralgia (pain in a joint) and neuropathy (numbness or weakness)). The following statements captured these barriers: “I tend to baby my surgical side, I’m very cautious with it. Even to the point when I go to the doctor they told me that you don’t do blood pressure from that side, you don’t do anything from that side, nothing over 5 pounds.” “When you have nothing in the area, the body tends to shift and close up.” “I gradually starting to stretch and the beginning up exercise, I was like oh my God I can’t wait to get home and try in private” “Having a tram flap restricted some of my movement. Sometimes I get very frustrated with that and I just don’t want to do. I realized today, just modify it. I did a lot of things the normal way and when I got to the point where I couldn’t do it, I modified it with and it was a revelation to me.” For BCSs, physical activity improves fitness and alleviates cancer-related fatigue and enhances physical functioning (Knobf et al., 2014). Another theme that arose from the FGDs as a secondary barrier to physical activity was fatigue. According to one BCS: “Since my chemo ended two years ago, I still have problems with fatigue. When I get home from work, I am really too tired to exercise.” For some respondents, closely related to fatigue was the extreme hot temperature experienced in South Florida as a barrier to physical activity, as illustrated in one BCS remarked: “Living in the hottest place in the country [Miami], I hate to sweat. Exercise wears me out.” Strategies related to cancer prevention guideline adherence Participants shared the following strategies for promoting guideline adherence: “The feedback from the leader was really encouraging.” “Have two pairs of walking shoes and leave one at work so that when a colleague says ‘let’s go walk’ you won’t say that you don’t have any shoes.” “I wasn’t able to go the complete distance but the distance I went was good for me. I bonded with two of my sistaahs and I enjoyed it. Exercising with another breast cancer survivor really helps.” “If you have lymphedema, you have to be careful about using heavy weights [for strength training] and you need to wear your sleeves because it’s very important.” App components to address barriers to lifestyle modification Table 4 summarizes the main findings from the SSIs. Strategies were categorized as major themes for inclusion in the mobile cancer prevention app. Components were selected based on apps reviewed by participants. Many participants commented on the need for social support and advice from experts. Further, they wanted the app to be easy to operate, include a tutorial, link to related resources, and operate at no cost to users. Culture and tradition were identified as major themes for lifestyle modification. Related to food, participants wanted the app to include soul food, southern, and Caribbean recipes; for physical activity, they felt that simple, easy-to-do exercises should be the focus. DISCUSSION Twelve AA BCSs (coaches) with a mean age of 50 years, participated in a qualitative study to elicit stakeholder advice to guide the development of a mobile cancer prevention app. Responses from participants were organized into 3 categories: 1) perceptions about modifiable risk factors; 2) strategies related to adherence to cancer prevention guidelines; and 3) app components to address barriers to adherence. Participants were aware of the racial/ethnic disparities that exist in modifiable lifestyle risk for breast cancer, and they understood the importance of choosing to eat healthy, avoiding red and processed meat, consuming more fruits and vegetables, and reducing meal portion sizes. They also understood the importance of walking and aerobic exercise. Barriers to meeting dietary recommendations were making food taste good, access to healthy foods, costs, and preference for cultural foods. For exercise, barriers include limitations from complications of cancer treatment such as lymphedema, joint pain, and neuropathy. The strategies listed for adhering to cancer prevention include partnering with survivors to exercise, having an extra pair of walking shoes for convenience, wearing sleeves, being careful about using heavy weights in cases of lymphedema, and recipes for healthy meals. The app components included educational materials about WCRF/AICR prevention guidelines, a diary and reminders, a BMI calculator, links to social media, internet educational videos, and flags for lapses. There has been a proliferation of health-related apps, especially ones targeting exercise, diet, and weight. The advantages that web/mobile phone apps hold over standard face-to-face programs include ready availability, less burden, acceptability by target populations, greater program adherence, low cost, self-monitoring, and reaching a large target population. A limitation is determining how to design an appealing app that will maintain user engagement over time (Pelligrini et al., 2015). Health promotion interventions should be culturally appropriate and tailored to meet the needs of the targeted population (Ansa et al., 2015). The present study is among the first to consider the socioeconomic and cultural needs as well as the health challenges of its intended users in guiding the development of a mobile lifestyle app. This approach is likely to enhance user engagement over time and increase adherence to recommendations for cancer prevention. Previous studies involving technology-based interventions have shown that individually customized messages are more effective than non-customized messages in improving self-efficacy and dietary behaviors (Kroeze et al., 2006). The small sample size and non-generalizability of results to other breast cancer groups are limitations of the study. CONCLUSIONS Among AA BCSs, complex cultural and socioeconomic issues may hinder optimal adherence to cancer prevention guidelines. General prevention measures may not be sufficient for eliminating disparities that exist as a result of such issues. Researchers should work collaboratively with survivors and support groups to ensure that interventions that are developed and tested are relevant and practical in meeting the needs of this group. The findings from this study will assist investigators in the development of a mobile cancer prevention app with features that are usable, acceptable, and accessible to AA BCSs. The authors wish to thank the women of SISTAAH Talk for their contributions to developing the app. The National Cancer Institute (R01CA166785) and the Medical College of Georgia Foundation funded this work. APPENDIX Focus Group Discussion Topic Guides Physical Activity What did you expect to get out of today’s walk (strength training) (yoga) demonstration? Were you looking forward to it? Why or why not? What did you like the most about the walk (strength training) (yoga) demonstration? What did you like least about the walk (strength training) (yoga) demonstration? Do you have any limitations or problems related to walking (strength training) (yoga)? If yes, please describe them. Tell us what you learned about walking (strength training) (yoga); will it motivate you to incorporate it into you daily activities? How did you feel about partnering with a breast cancer survivor for walking (strength training) (yoga)? Were there advantages/disadvantages to partnering? What did you gain from today’s walking (strength training) (yoga) demonstration? Would you advise breast cancer survivors to walk (strength training) (yoga)? Why or why not? Dietary Intake What did you expect to get out of the cooking (grains) (red meat) (fruits and vegetables) demonstration? What did you like most about the cooking (grains) (red meat) (fruits and vegetables) demonstration? What did you like least about the cooking (grains) (red meat) (fruits and vegetables) demonstration? Tell us how you feel the affects of how you eat and breast cancer? How relevant are the cancer prevention guidelines to the foods that you select? What would you say are the main challenges you have faced in changing your eating habits? Why is reading food labels important to you? What changes to your diet, if any, have you made since your breast cancer diagnosis? How do you feel about letting people around you know that you are trying to change the way that you eat? What are some of the ways that you have been able to get the kind of support that you need to make changes to your diet? Semi-Structured Interview Topic Guides Internet Usage Do you have daily access to the Internet? If so, do you use it to access information about diet/nutrition? Physical activity? Do you have a smartphone? Do you use apps on your smartphone? If so, do you use apps to obtain information about diet/nutrition? Physical activity? How comfortable are you in using a computer or smart phone to obtain information about diet/nutrition? Physical activity? Which features are important to you to include on the SISTAAH Talk app? How do you feel about developing the SISTAAH Talk app? What do you hope to gain from this experience? Smartphone Apps What type of smartphone do you have? Android or iPhone? Did you complete the assignment to test health and fitness apps? How many apps did you review? (total number) If yes, did you review diet or nutrition apps? If yes, did you review physical activity or diet apps? What did you like most about the apps that you tested? What did you like least about the apps that you tested? Of the apps you reviewed, which features would you like to see on the SISTAAH Talk app? If no, please tell me why you did not test health and fitness apps? Now, I would like for you to tell me how comfortable you were testing the smartphone apps. On a scaled of 1 to 10, where 1 is not very comfortable and 10 is extremely comfortable, how comfortable were you in using health and fitness apps? Would you say that your comfort level was a 1 _ _ _ _ _ _ _ _ or 10? Based on your response, what do you feel is needed to improve your comfort level? Is there anything else you would like for us to consider in developing the SISTAAH Talk app? Experiential Educational Sessions Did you find the educational sessions useful in managing your dietary intake? How did you feel about the SISTAAH Talk discussions following the cooking demonstrations? Did participating in developing content for the app inspire you to do more physical activity? Why or why not? If so, did you set a physical activity goal; if yes, what was the goal that you set? Did the content of the cooking demonstrations fit your needs? Please explain how your needs were met. Did the content of the exercise demonstrations fit your needs? Please explain how your needs were met. How confident do you feel about the information that you learned during the sessions? Were the cancer prevention guidelines well explained? If not, which components do you need more information about? After participating in activities to develop the app, are you inspired to eat healthier? To exercise more? Why or why not? Which components inspired you and how were you inspired? Figure 1 SISTAAH Talk App Intervention Map Table 1 Community advisory board representation Community Coalitions & Support Groups by City Miami Florida Resources for Enhancing & Sustaining Health SISTAAH Talk Chicago National Black Leadership Initiative on Cancer Sisters Working it Out CT Joiner Foundation Houston National Black Leadership Initiative on Cancer Angels Surviving Cancer, Inc. Reconstruction of a Survivor Los Angeles Black Women for Wellness Sisters Breast Cancer Survivors Network Celebrate Life Philadelphia National Black Leadership Initiative on Cancer Women of Faith and Hope Linda Creed Breast Cancer.Org Table 2 SISTAAH Talk app theoretical components Component Health Belief Modela Theory of Planned Behaviorb HBM1 HBM2 HBM3 TPB4 TPB5 Education ✓ ✓ ✓ Instructions ✓ Goal setting ✓ Social support ✓ Provide feedback ✓ Prompt review ✓ Self-monitoring ✓ ✓ Teach use of cues ✓ Action planning ✓ a Health Belief Model: HBM1 (perceived costs); HBM2 (health benefits); HBM3 (cues for action). b Theory of Planned Behavior: TPB4 (subjective norms/social support); TPB5 (behavior control). Table 3 Sociodemographic characteristics of SISTAAH Talk coaches Characteristics Participants (n=12) Mean age, years (range) 50 (40–72) Educational Level   High school or less 3 (25.0)   College 7 (58.3)   Graduate 2 (16.7) Marital Status (percent)   Single 3(25.0)   Married 4(33.0)   Widow/Divorced 5(42.0) Annual Income (%)   $0-$24,999 4(33.0)   $25,000-$49,000 5(42.0)   ≥$50,000 3(33.0) Mean years (range) since diagnosis of breast cancer 8.7 (2–26) Table 4 Strategies and participant app component preferences App Component Strategy Quotes WCRF/AICR cancer prevention guidelines Educational materials “Because as a cancer survivor, I think we need that kind of education. Because there are so many programs teaching how to work out; but as breast cancer survivors, need special education that will not traumatize our bodies.” Food and exercise diary Dietary intake and physical activity progress “I want an app to chart or journal what I have been eating and doing—to give me ideas and guidance.” Email reminders Feedback Instructions “I like email reminders and tips for eating every day, reminding me to drink water, to eat fruits and vegetables.” Body mass index calculator Self-monitoring of body weight “The best app is one that calculates how fat you are. Having a BMI calculator is a must!’ Healthy weight range Self-monitoring of body weight Goal setting “Healthy weight in my community is different from for other women. I need to know how much I should weight to prevent my breast cancer from coming back—I need a goal.” Energy/calorie requirement calculator Self-monitoring of portion control Goal setting “Apps that have restaurant information and the calculation of the calories for any of the foods you eat are best.” Food group recommendations Self-monitoring of fruits and vegetables; red and processed meats; whole grains “It would help if the app allowed me to track my progress—tell me if I am eating the right number of foods to prevent cancer.” Recipes—pictures of foods, YouTube videos Educational materials Instructions “The ones (apps) with links to other websites was very good.” Exercise—instructions, YouTube videos Educational materials Instructions “More information, especially about physical activity.” Tracking negative thoughts/stress Stress reduction “I learned how to release my stress, to breathe that’s a good thing and also not to stress our body, to go further than we can. We need to include this in the app.” Links to FaceBook, Twitter Social support Feedback “Another thing I like is when you can talk with certain people—people going through what you are going though; experts like nutritionists and fitness experts.” Internet website links Resources “If it’s too much information, someone like me who is new to using apps, if it’s too technical, the user will give up.” It’s important to make the app ‘user-friendly’ and link out to other resources.” Reminders to log food and activity Self-monitoring Goal setting “The best app is one that communicates with you—that sends you reminders to track your progress.” Flags for lapses in diet and physical activity goal adherence Self-monitoring Goal setting “Once I set a goal, I usually stick to it. But some things that I think that I am doing right, I am not really doing right. I need someone to tell me.” ==== Refs References American Cancer Society Cancer Facts & Figures for African Americans 2016–2018 2016 Accessed May 24, 2016 Atlanta American Cancer Society Available online: http://www.cancer.org/acs/groups/content/@editorial/documents/document/acspc-047403.pdf Ajzen I The theory of planned behavior Organ Behav Hum Perf 1991 12 1 50 2 179 211 Ansa B Yoo W Whitehead M Coughlin S Smith S Beliefs and behaviors about breast cancer recurrence risk reduction among African American breast cancer survivors Int J Environ Res Public Health 2015 12 23 13 1 46 Bartholomew LK Parcel GS Kok G Intervention mapping: a process for developing theory and evidence-based health education programs Health Educ Behav 1998 10 1 25 5 545 563 9768376 Carlsen B Glenton C What about N? A methodological study of sample-size reporting in focus group studies BMC Med Res Methodol 2011 3 11 11 1 1 21208427 Centers for Disease Control and Prevention (CDC) National Center for Chronic Disease Prevention and Health Promotion, Division of Population Health BRFSS Prevalence & Trends Data 2015 Accessed May 26, 2016 Available online: URL: http://wwwdev.cdc.gov/brfss/brfssprevalence/ Claridy M Thompson W Ansa B Damus F Alema-Mensah E Smith S Comparison of Health-Related Quality of Life Among African American Female Breast Cancer Survivors, Survivors of Other Cancers, and Those Without Cancer Prev Chronic Dis in press Corbie-Smith G Akers A Blumenthal C Council B Wynn M Muhammad M Stith D Intervention mapping as a participatory approach to developing an HIV prevention intervention in rural African American communities AIDS Educ Prev 2010 6 22 3 184 202 20528128 Corbin J Strauss A Basics of qualitative research: techniques and procedures for developing grounded theory 2008 London Sage Friedenreich CM The role of physical activity in breast cancer etiology Semin Oncol 2010 6 37 3 297 302 20709211 Glanz K Rimer BK Viswanath K Health behavior and health education: Theory, research, and practice 2008 San Francisco Jossey-Bass Inc. Pub 45 65 Glasgow RE Marcus AC Bull SS Wilson KM Disseminating effective cancer screening interventions Cancer 2004 9 1 101 S5 1239 1250 15316911 Hastert TA Beresford SA Patterson RE Kristal AR White E Adherence to WCRF/AICR cancer prevention recommendations and risk of post-menopausal breast cancer Cancer Epidemiol Biomark Prev 2013 22 9 1498 1508 Howell A Anderson AS Clarke RB Duffy SW Evans DG Garcia-Closas M Gescher AJ Key TJ Saxton JM Harvie MN Risk determination and prevention of breast cancer Breast Cancer Res 2014 9 28 16 5 446 25467785 Knobf MT Thompson AS Fennie K Erdos D The effect of a community-based exercise intervention on symptoms and quality of life Cancer Nurs 2014 3 37 2 E43 E50 23519041 Kroeze W Werkman A Brug J A systematic review of randomized trials on the effectiveness of computer-tailored education on physical activity and dietary behaviors Ann Behav Med 2006 6 1 31 3 205 223 16700634 Kushi LH Doyle C McCullough M Rock CL Demark-Wahnefried W Bandera EV Gapstur S Patel AV Andrews K Gansler T American Cancer Society guidelines on nutrition and physical activity for cancer prevention CA Cancer J Clin 2012 1 1 62 1 30 67 22237782 Lyons EJ Baranowski T Basen-Engquist KM Lewis ZH Swartz MC Jennings K Volpi E Testing the effects of narrative and play on physical activity among breast cancer survivors using mobile apps: study protocol for a randomized controlled trial BMC Cancer 2016 3 9 16 1 1 McEachan RR Lawton RJ Jackson C Conner M Lunt J Evidence, theory and context: using intervention mapping to develop a worksite physical activity intervention BMC Public Health 2008 9 22 8 1 1 18173844 McCullough ML Patel AV Kushi LH Patel R Willett WC Doyle C Thun MJ Gapstur SM Following cancer prevention guidelines reduces risk of cancer, cardiovascular disease, and all-cause mortality Cancer Epidemiol Biomarkers Prev 2011 6 1 20 6 1089 1097 21467238 National Center for Health Statistics (NCHS) Health, United States, 2015: With Special Feature on Racial and Ethnic Health Disparities 2016 Accessed on May 24, 2016 Hyattsville, MD Available online: http://www.cdc.gov/nchs/data/hus/hus15.pdf NVIVO 10 for Windows aspx accessed April 18, 2016 URL: http://www.qsrinternational.com/products_nvivo.aspx , [WebCite Cache ID 6aDk3Ljxm] Pellegrini CA Pfammatter AF Conroy DE Spring B Smartphone applications to support weight loss: current perspectives Adv Health Care Technol 2015 7 1 13 22 26236766 Pew Research Center [homepage on the Internet] Anderson M The Demographics of Device Ownership 2015 Accessed May 24, 2016 Pew Research Center Available from: http://www.pewinternet.org/2015/10/29/the-demographics-of-device-ownership/ Schreier M Qualitative content analysis in practice 2012 USA Sage Publications Ltd Smith SA Claridy MD Whitehead MS Sheats JQ Yoo W Alema-Mensah E Ansa BE Braithwaite RL Factors associated with body mass index among African American breast cancer survivors J Ga Public Health Assoc 2016 Winter 5 3 259 265 27019873 Smith SA Whitehead MS Sheats JQ Ansa BE Coughlin SS Blumenthal DS Community-based participatory research principles for the African American community J Ga Public Health Assoc 2015 5 1 52 26336653 Smith SA Whitehead MS Sheats J Mastromonico J Yoo W Coughlin SS A Community-Engaged Approach to Developing a Mobile Cancer Prevention App: The mCPA Study Protocol JMIR Res Protocols 2016 1 5 1 Strickler HD Obesity and its impact on breast cancer: tumor incidence, recurrence, survival, and possible interventions Am Soc Clin Oncol 2013 52 59 World Cancer Research Fund/American Institute for Cancer Research Food, nutrition, physical activity, and the prevention of cancer: a global perspective 2007 Washington, DC AICR

PMC005xxxxxx/PMC5003318.txt

==== Front J Dev BiolJ Dev BioljdbJournal of Developmental Biology2221-3759MDPI 10.3390/jdb4020016jdb-04-00016ReviewMechanisms of Specificity for Hox Factor Activity Zandvakili Arya 12Gebelein Brian 3*Zappavigna Vincenzo Academic EditorConway Simon J. Academic Editor1 Molecular and Developmental Biology Graduate Program, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH 45229, USA2 Medical-Scientist Training Program, University of Cincinnati College of Medicine, Cincinnati, OH 45267, USA; zandvaaa@mail.uc.edu3 Division of Developmental Biology, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH 45229, USA* Correspondence: brian.gebelein@cchmc.org; Tel.: +1-513-636-336609 5 2016 6 2016 4 2 1605 4 2016 04 5 2016 © 2016 by the authors.2016Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).Metazoans encode clusters of paralogous Hox genes that are critical for proper development of the body plan. However, there are a number of unresolved issues regarding how paralogous Hox factors achieve specificity to control distinct cell fates. First, how do Hox paralogs, which have very similar DNA binding preferences in vitro, drive different transcriptional programs in vivo? Second, the number of potential Hox binding sites within the genome is vast compared to the number of sites bound. Hence, what determines where in the genome Hox factors bind? Third, what determines whether a Hox factor will activate or repress a specific target gene? Here, we review the current evidence that is beginning to shed light onto these questions. In particular, we highlight how cooperative interactions with other transcription factors (especially PBC and HMP proteins) and the sequences of cis-regulatory modules provide a basis for the mechanisms of Hox specificity. We conclude by integrating a number of the concepts described throughout the review in a case study of a highly interrogated Drosophila cis-regulatory module named “The Distal-less Conserved Regulatory Element” (DCRE). Hoxtranscription factorcis-regulatory modulesDNA binding specificity ==== Body 1. Introduction Hox proteins are a family of homeodomain-containing Transcription Factors (TFs) that are critical for regulating developmental processes in metazoans [1,2]. The stark homeotic changes occurring in Hox mutant animals, such as the classic antenna-to-leg transformations in Drosophila, testify to the importance of Hox factors for proper development [3]. In addition, more recent studies have demonstrated that Hox factors not only drive stereotypic developmental programs, but also have a role in maintaining differentiated cell populations [4,5]. Given that Hox factors are conserved from C. elegans to H. sapiens, a fundamental understanding of how Hox factors function will yield significant insights into both the development and evolution of body plans. The majority of metazoan genomes encode clusters of paralogous Hox genes (Figure 1A). While invertebrates generally have one Hox cluster, vertebrates have multiple Hox clusters owing to duplications of the entire cluster during evolution [6]. The order of Hox genes within the clusters typically correlates with their expression pattern along the Anterior-Posterior (A-P) axis of the embryo [7,8,9]. Genes at the 3′ end of a cluster are expressed in anterior regions of an embryo, whereas genes toward the 5′ end are expressed in progressively more posterior regions of the embryo. Thus, the differential expression of Hox genes is a key step in the specification of distinct cell fates along the A-P axis. At the sequence level, Hox proteins share two stereotypic domains: a conserved homeodomain and a conserved Hexapeptide (HX) motif that is located N-terminal to the homeodomain (Figure 1B). Hox factors utilize their homeodomains to directly bind DNA, and they share very similar binding preferences in vitro as monomers [21,22,23]. The HX motif mediates direct interactions with another family of TFs (PBC proteins), and it is separated from the homeodomain by a flexible and highly variable linker region [12,13]. Outside of the homeodomain and the HX motif, Hox protein sequences diverge substantially; and while we largely do not understand their functions, these non-conserved regions contain residues that can be post-translationally modified and/or have been implicated in protein-protein interactions and the regulation of transcriptional outputs [24,25,26,27]. Hox factors specify cell-fates based on their ability to interact with Cis-Regulatory Modules (CRMs) and to regulate transcription [28,29]. CRMs provide a DNA platform to organize interactions between Hox factors and other proteins. Although Hox interactions with DNA and other proteins have been studied extensively, we still lack a comprehensive understanding of how specificity is achieved to accurately recognize and regulate target genes required to direct specific cell fates. In this review, we explore three questions related to Hox specificity: First, how do paralogous Hox proteins drive different cell fates even though they utilize conserved homeodomains with highly similar DNA binding preferences? Second, metazoan genomes contain a large number of potential Hox binding sites, yet only a subset of these sites are bound at any one time. What factors determine which sequences are bound and regulated by Hox proteins, whereas other sequences containing Hox binding sites remain unbound? Third, once bound, how does an individual Hox factor activate some target genes and repress others? Below, we present emerging evidence that provides new insight into the mechanisms that contribute to the specificity of Hox action. 2. Differentiation of Hox Paralog Activities A fundamental problem in the study of TF specificity is that most TFs are members of large protein families that have highly similar DNA binding properties yet distinct in vivo functions [30]. The Hox family of TFs is an exemplar of this problem. While paralogous Hox factors bind highly similar DNA sequences, genetic loss- and gain-of-function studies demonstrate that Hox factors control diverse cell fates along the A-P axis of metazoans [1,31]. Furthermore, in C. elegans, it has been demonstrated that paralogous Hox proteins have highly divergent genomic binding patterns in vivo [32]. However, it is not immediately clear whether this difference in in vivo function is due to differences in Hox paralogs or due to the fact that Hox paralogs are functioning in different cellular contexts. Several studies have controlled for cellular context and have demonstrated that paralogous Hox proteins have different in vivo activities within the same cell types. First, over- or under-expressing specific Hox paralogs within the same cell types can result in different phenotypes [33,34,35]. For instance, in C. elegans, when the Hox genes egl-5 and lin-39 are expressed under the control of regulatory elements from the mab-5 locus (a different Hox gene), they do not rescue the mab-5 mutant phenotype [35]. Second, the misexpression of different Hox genes results in distinct changes in global gene expression patterns. For example, six Drosophila Hox genes were individually expressed in a ubiquitous pattern with the same Gal4 driver line in Drosophila embryos, and RNA was isolated to compare changes in gene expression [36]. Of the genes that changed in expression, the majority (nearly 70%) changed in response to a single Hox factor, while only 1.3% of the genes changed in response to all six Hox factors [36]. Third, a growing number of Hox-regulated CRMs has been identified, and many are regulated by only one or a small subset of paralogous Hox factors when tested in the same cellular contexts [37,38,39,40,41]. While these studies all support the notion that Hox factors largely control distinct cell fates, it has been found that, at least in some contexts, Hox factors can produce very similar phenotypes when expressed in the same cell types. For example, in Drosophila, all Hox factors (except abdominal-B) can substitute for Labial (Lab) in specifying tritocerebral neuromere [42]. Furthermore, in vertebrates, Hox mutant phenotypes become more severe when multiple Hox genes with overlapping expression are mutated, suggesting that Hox genes have some functional redundancy [9,43]. Thus, each Hox factor is likely to regulate both common and paralog-specific target genes and cell fates in vivo. Here, we focus on how the cooperation with other sequence-specific TFs aids in the differentiation of the binding preferences and activities of Hox paralogs. We describe how interactions between Hox factors and the PBC family of homeodomain proteins uncover the importance of latent DNA binding specificity and low affinity binding sites in generating Hox-specific outputs. We subsequently explore how the HMP family of homeodomain proteins selectively interacts with posterior Hox factors. In addition, HMP TFs also directly interact with the PBC proteins and thereby form higher-order TF complexes with Hox factors that can enhance both DNA binding affinity and specificity. Next, we provide a few examples of the diversity of other TFs that interact with subsets of Hox factors, either directly or via nearby binding sites encoded within CRMs. Finally, we describe how post-translational modifications differentially affect the activities of paralogous Hox proteins. A summary of these mechanisms is provided in Figure 2. 2.1. Interactions with PBC Factors Reveal Hox Latent Specificity First identified in 1990, PBC proteins were nearly simultaneously discovered as modulators of Hox function in Drosophila and proto-oncogenes in mammals [52,53,54]. The PBC family includes Extradenticle (Exd) in Drosophila, CEH-20 and CEH-40 in C. elegans and Pbx factors in mammals; and all PBC proteins contain a highly-conserved homeodomain that differs from a canonical homeodomain by the addition of a Three Amino-acid Loop Extension (TALE) motif between helix 1 and helix 2 of the homeodomain [55] (Figure 3A). It was recognized early on that Hox and PBC proteins bind DNA cooperatively [56,57,58] and that PBC proteins were essential for Hox function [59,60,61]. Subsequent structural studies determined that Hox-PBC interactions on DNA were mediated via insertion of the Hox HX motif into a hydrophobic pocket of the PBC homeodomain that is composed of residues from the TALE motif, helix 1 and helix 3 [12,13] (Figure 3B). The Hox HX motif is located N-terminal to the homeodomain and contains a highly-conserved Y/F-P/D-W-M sequence (Figure 1B), where the W residue is critical for making hydrophobic interactions with the PBC TALE motif [12]. While the majority of Hox factors have a defined HX motif, the posterior Abdominal-B (Abd-B) or Hox paralog group 9–13 factors only rely upon a conserved W residue to mediate this interaction [62]. Importantly, the interaction between Hox and PBC factors occurs through nearby DNA binding sites for each factor and, thereby, results in both enhanced DNA binding specificity and affinity [23,56,63]. In addition to enhancing DNA binding affinity, interactions between TFs can result in changes in binding preferences via allosteric changes in conformation of the TFs and/or the DNA [71,72,73]. The concept of revealing differences in binding preferences between similar TFs is known as latent specificity. The latent specificity mechanism is well-established for interactions between Hox and PBC proteins. Soon after the discovery of PBC proteins, biochemical evidence demonstrated that PBC proteins modify the DNA binding specificity of Drosophila and mammalian Hox paralogs [56,57,74,75]. This concept was most comprehensively tested in vitro by a 2011 SELEX-seq (Systematic Evolution of Ligands by EXponential enrichment followed by sequencing) study, which demonstrated that Drosophila Exd-Hox heterodimers have a greater diversity of binding site preferences than Hox monomers [23]. Specifically, it was found that interactions between Hox factors and Exd result in the selection of distinct sequences within six core nucleotides (A5YNNAY10) of the Hox binding site (note, the combined PBC-Hox site typically consists of 12 nucleotides, nnTGAYnnAYnn). Interestingly, the binding preference for Exd-Hox dimers segregated with the expression of Hox factors along the A-P axis; thus, the preference for a core motif can be grouped into distinct clusters of anterior (Labial and Proboscipedia), middle (Deformed and Sex combs reduced) and posterior Hox factors (Antennapedia, Ultrabithorax, Abdominal-A and Abdominal-B). These preferences also segregated with the predicted width of the DNA minor groove within the Exd-Hox core-motif. Specifically, it was found that anterior Hox factors select sequences with a narrow minor groove in position A9Y10, while posterior Hox factors select sequences with wider minor grooves [23,76]. In addition to differentiating preferences within the Exd-Hox core-motif, there were distinct preferences between Hox factors for sequences flanking the Hox half of the Exd-Hox motif, especially for anterior Hox factors [23]. However, there is an exception to the rule of Exd differentiating preferences; interaction with Exd actually made the preferences of Ultrabithorax (Ubx) and Abd-B more similar [23]. Structural studies have revealed that interactions with Exd on DNA allow protein sequences that are more variable between Hox paralogs (namely the linker region and N-terminal arm of the homeodomain) to make a greater contribution to DNA binding. Comparing crystal structures of Exd and the Sex Combs Reduced (Scr) Hox factor bound to either a selective sequence specifically bound by Scr (fkh250) or a non-selective consensus Exd-Hox site bound by many Hox factors (fkh250con) revealed significant differences in the mechanism of DNA binding [71]. On the selective sequence, the Scr N-terminal arm is structured and inserted into the minor groove, whereas these same amino acids are disordered and make a minimal contribution to Scr binding on the fkh250con sequence [71]. Selectivity was found to depend upon two residues—Histidine-12 (His-12, in the Linker Region) and Arginine 3 (Arg3, in the N-terminal arm)—that insert into the minor groove of fkh250. His-12 and Arg3 are highly conserved among Scr homologs, suggesting that these residues are important to Scr function. Since His-12 and Arg3 do not make direct hydrogen bonds with bases, it was proposed that these two residues recognize the shape of the DNA. Indeed, a recent SELEX-seq study using Hox-Exd dimers demonstrated that mutating His-12 and Arg3 in Scr results in less selectivity for sequences that are predicted to have narrow minor grooves at position A8Y9 [76]. Furthermore, mutating the linker region and the N-terminal arm of the homeodomain in Antennapedia (Antp) to match those of Scr not only changed Exd-Antp binding preferences to be very similar to those of Exd-Scr, but it also allowed the mutant Antp protein to activate the Scr-specific enhancer (fkh250) in vivo [76]. This SELEX-seq study is also notable for demonstrating that using both DNA shape-readout, as well as DNA base-readout greatly improves the predictions of sequence-specific affinity of Hox-Exd complexes. Importantly, the concept of latent specificity may be broadly applicable to the study of TFs in general. A recent large-scale SELEX-based screen of cooperative DNA binding between TF pairs identified additional examples of latent specificity. For instance, HoxB2 was found to diversify the binding preferences of a subset of ETS factors (E-26 transcription factors) [45]. Interestingly, this screen demonstrated that of the 315 TF-TF heterodimers that were analyzed, 2/3 of the heterodimeric binding preferences were substantially different than the monomeric binding preferences of the individual TFs. However, in many cases, when a set of paralogous TFs interacted with the same partner, the resultant binding preference was very similar, suggesting that latent specificity is unlikely to be the sole source of distinguishing the activities of paralogous TFs. 2.2. Low-Affinity Binding Sites Distinguish Hox-PBC Dimers The latent specificity mechanism demonstrates that adjacent Hox-PBC binding sites in CRMs can contribute to paralog-specific binding of CRMs. However, many of the highest affinity sites identified from these assays can be bound by several paralogous Hox factors, suggesting that high-affinity sites may be regulated by many Hox factors. More recently, a study proposed that CRMs distinguish between Hox paralogs via low-affinity binding sites [41]. The idea behind this proposal is that TFs within a family have very similar DNA binding domains that vary only in sites that make weak interactions with DNA. Therefore, the variations in DNA-binding domains only make minor contributions to the overall binding energy when binding high affinity sites. By contrast, subtle differences in DNA binding domains can make a larger contribution to the overall binding energy when interacting with low-affinity sites. The binding of the fkh250 sequence by Scr described above is an example of how a lower affinity site depends on Scr-specific amino acid contacts for DNA binding, whereas a higher affinity site (fkh250con) only requires homeodomain amino acids that are conserved in many Hox paralogs [71]. In addition, Crocker et al. identified a cluster of low-affinity Ubx-Exd binding sites within enhancers in the Drosophila shavenbaby locus, and they found that replacing these sites with higher affinity Hox-Exd sites allowed the activation of the enhancer by other Hox factors in reporter assays [41]. Interestingly, it is not uncommon to find functional, low-affinity binding sites within developmentally-important enhancers [41,77,78,79,80], suggesting that low affinity might provide a selective advantage over high-affinity binding sites. However, the prevalence of low-affinity binding sites in developmentally-important enhancers may also be a function of the fact that for any given TF, a greater number of sequences can yield low-affinity binding sites than high affinity binding sites. Crocker et al. provide a recent review of low-affinity binding sites and their potential functions [81]. 2.3. Additional PBC Interaction Surfaces Differentiate Hox Paralogs While it has been widely demonstrated that the Hox HX motif mediates Hox-PBC interactions, results from several biochemical experiments suggest that it is not the sole mechanism for Hox-PBC interactions. In fact, mutating or removing the HX motif in Lab, Ubx or Abdominal-A (Abd-A) does not abolish the ability of these Hox factors to interact with Exd on DNA [15,56,82,83,84,85,86,87]. Furthermore, in vivo Exd-dependent activities for Lab, Ubx and Abd-A were largely not abolished by HX mutations [15,57,83,84,85,86,87]. In addition, recent in vivo fluorescence complementation assays (Bimolecular Fluorescence (BiFC)) have demonstrated that the majority of Hox factors are not completely dependent on the HX motif to form complexes with Exd [88]. These findings indicate that Hox factors can use alternative mechanisms to interact with PBC factors to regulate gene expression. An alternative PBC interaction motif has been identified in the Drosophila Ubx and Abd-A Hox factors. This motif is conserved among arthropod and cnidarian Ubx and Abd-A homologs and, hence, has been named the UbdA motif [18]. The UbdA motif is located immediately C-terminal to the homeodomain. Biochemical and in vivo BiFC assays have demonstrated that the UbdA motif contributes to interactions with Exd [15,84,86,88]. Moreover, mutation of the UbdA motif in Abd-A and Ubx results in decreased Exd-dependent activities in vivo [15,18,84,86]. More recently, a crystal structure of Ubx interacting with Exd via its UbdA domain has been described. Foos et al. crystalized two types of Ubx/Exd/DNA complexes, one using Ubx proteins with both the HX and UbdA motif and one with only the UbdA motif [18]. In both complexes, the homeodomains of Ubx and Exd were situated on opposite sides of the DNA. When both the HX and UbdA motifs were present in Ubx, only the HX motif made contacts Exd, whereas the UbdA motif was mostly disordered. By contrast, if only the UbdA only is present in Ubx, the UbdA forms a coiled extension off of the homeodomain’s third helix and makes contact with the loop region between the second and third helices of the Exd homeodomain (Figure 3B). The UbdA-Exd interaction occurs on the opposite side of the DNA relative to the HX-Exd interaction. Based on these structures, it is not clear what regulates whether the HX and/or UbdA mediates interactions between Ubx/Abd-A and PBC proteins. However, Saadaoui et al. (2011) demonstrated that Ubx isoforms with longer linker region lengths are more resistant to HX mutations in regulating a CRM containing adjacent Exd-Ubx binding sites [15]. Given the topology, this may be due to the fact that the longer linker region length gives flexibility to Ubx and Abd-A proteins to mediate contacts on both sides of the DNA via the HX and UbdA motifs. 2.4. Interactions with HMP Proteins Differentiate Hox Paralogs In addition to PBC factors, another group of TALE homeodomain proteins directly associates with Hox factors on DNA and contributes to Hox paralog specificity. The HMP family of proteins includes Homothorax (Hth) in Drosophila, UNC-62 in C. elegans and Meis and Prep (pKnox) in vertebrates. Hth was originally identified as a gene important to larval cuticle development in Drosophila, while Meis was originally identified as a proto-oncogene in mammals [89,90]. These genes were later shown to interact with PBC and Hox proteins and thereby contribute to Hox-mediated outcomes [26,44,91]. The vertebrate HMP proteins, Meis1A and 1B, have been shown to interact with Hox factors from multiple paralog groups (murine Hox paralog groups 2, 4, 5, 8, 9, 10, 11, 12 and 13) using yeast two-hybrid assays [26]. However, DNA binding assays indicate that only posterior Hox factors (Abd-B family members in particular) bind cooperatively with HMP proteins to DNA containing adjacent Hox-HMP binding sites [26,44,45]. In Drosophila, the Ubx and Abd-A Hox factors have also been shown to directly cooperate with Hth by binding and regulating CRMs containing adjacent Hox-Hth binding sites [40,46]. These findings suggest that the arrangement of adjacent Hox-HMP binding sites may contribute to paralog specificity by favoring posterior Hox factors over anterior ones. However, the crystal structures of Hox-HMP-DNA complexes have not been solved, and therefore, the mechanism and specificity of interaction between Hox and HMP factors remains unclear. In addition to interacting with Hox factors, the TALE PBC and HMP proteins heterodimerize through highly-conserved domains found N-terminal to their homeodomains. Importantly, PBC-HMP heterodimer formation increases the nuclear import of PBC, which is largely cytoplasmic in the absence of HMP proteins, and enhances the stability of HMP proteins [91,92,93,94,95]. More recently, it has been shown that the Hox-PBC-HMP interactions are ancient and evolutionarily conserved among cnidarians and bilaterians and that Hox, PBC and HMP from different species can interact with one another and substitute for one another in functional assays [96]. Molecularly, Hox-PBC-HMP interactions result in the formation of higher order complexes on DNA [97,98,99,100]. In fact, several CRMs containing binding sites for all three factors have been identified, especially in the vertebrate hindbrain and in Drosophila [40,46,58,64,65,66,67,69] (Figure 3C). In general, these CRMs contain either an adjacent PBC/Hox site with a nearby HMP site or an adjacent HMP/Hox site with a nearby PBC site. Analysis of binding site patterns in these CRMs reveals a variety of spacing and orientations of Hox, PBC and HMP binding sites that are sufficient to mediate the formation of Hox/PBC/HMP complexes on DNA [101] (Figure 3C). Moreover, one recent study found that hindbrain enhancers in zebrafish are enriched for HMP binding sites within 50 bp of combined Hox/PBC sites [102]. While this study did not conclusively demonstrate that these CRMs were directly regulated by Hox/PBC/HMP complexes, it does illustrate the potential flexibility of spacing for interacting HMP, Hox and PBC binding sites. In addition, the fact that certain interactions between Hox and PBC/HMP are paralog specific, i.e., the UbdA interaction motif in Ubx/Abd-A, the non-conserved HX motif in Abd-B homologs, the cooperative DNA binding between HMP proteins and posterior Hox factors, suggests the possibility that particular Hox-PBC-HMP binding site arrangements are paralog specific. Future studies focused on analyzing a large number of Hox-regulated CRMs will be needed to determine if specific arrangements of binding sites are an important contributor to Hox paralog specificity. 2.5. Hox Paralogs Have Different Partner Preferences As described for the PBC, HMP and Hox factors, interactions between TFs can yield cooperative complex formation on DNA, and the binding of each TF contributes to the overall binding affinity and specificity. In addition to PBC and HMP proteins, other sequence-specific TFs have been shown to interact with Hox factors, and several of these interactions are paralog specific. Thus, a CRM may encode Hox-paralog-specific activities by including a nearby binding site for a paralog-specific interacting TF. Large-scale screens have identified additional TFs that interact with specific Hox factors [45,103]. For instance, Baëza et al. tested the ability of five Drosophila Hox factors (Scr, Antp, Ubx, Abd-A and Abd-B) to interact with 35 other TFs using an in vivo BiFC assay [103]. This study demonstrated that each tested Hox factor interacts with a distinct combination of the 35 TFs. On average, there was a 59% pairwise similarity between the sets of TFs that interacted with each Hox factor. Presumably these differences in affinity for other TFs will contribute to the differential binding of Hox factors to genomic sequences. However, there is currently limited data showing that the interactions between the Hox factors and the 35 other TFs occur on DNA, much less affect DNA binding specificity. The data above suggest that Hox-regulated CRMs will require additional TF binding sites to yield paralog-specific outputs. This model is supported by examples of specific CRMs [40,46]. For instance, a rhomboid (rho) CRM (RhoA) in Drosophila contains binding sites for both an abdominal Hox complex composed of Exd/Hth/Abd-A, as well as the Pax2 transcription factor [47]. Importantly, mutations in either the Hox site or the Pax2 site disrupt activation, demonstrating that both the abdominal Hox complex and Pax2 protein need to bind the CRM for proper output. Moreover, the RhoA CRM is only activated by one specific abdominal Hox factor (Abd-A) that interacts with Pax2, whereas a thoracic Hox factor, Antp, does not interact with Pax2 and is unable to stimulate this CRM. This difference in ability to form complexes with Pax2 is thought to contribute to the paralog-specific activity of RhoA. Moreover, the interaction between specific posterior Hox factors and Pax2 may extend to vertebrates, as well. HoxA11 has been shown to interact with Pax2, and binding sites for both factors are required for the expression of a Six2 CRM in the developing mouse kidney [104]. 2.6. Post-Translational Modifications Differentially Affect Hox Paralogs Several different enzymes post-translationally modify Hox factors [48,49,50,51,105,106,107,108,109,110,111]. Here, we review those modifications that are known to differentially affect Hox paralogs. Poly(ADP)-Ribose Polymerase-1 (PARP-1) was shown to poly(ADP)-ribosylate several mammalian Hox factors (HoxA5, HoxA7, HoxB6, HoxB7, HoxC6, HoxC8), but this modification only reduces the in vitro DNA binding and transcriptional activity of HoxA7 and HoxB7 [48]. This study demonstrated that HoxA7 and HoxB7 have a long glutamate-rich repeat in their C-terminal motif that is necessary for poly(ADP)-ribosylation and that the addition of a long glutamate-rich repeat in HoxB6 was sufficient to allow poly(ADP)-ribosylation to disrupt HoxB6 activity in vitro [48]. Similarly, it has been shown that phosphorylation by Casein Kinase II (CKII) differentially affects Hox paralogs. Phosphorylation of the HoxA9 homeodomain by CKII decreases DNA binding and thereby diminishes the ability of HoxA9 to keep cultured hematopoietic progenitors from differentiating [49,50]. By contrast, HoxB7 appears to gain activity after phosphorylation. Two CKII phosphorylation sites (S132 and T203) were shown to flank the HoxB7 homeodomain, and mutating these sites to alanine inhibited the ability of HoxB7 to maintain hematopoietic cell culture in an undifferentiated state [51]. However, in both of these cases, it has not yet been delineated what mechanisms signal for the Hox paralogs to be post-translationally modified by these enzymes. An intriguing possibility is that post-translational modifications may occur at CRMs. Previous studies have shown that post-translational modifiers, such as PARP and CKII, can be recruited to CRMs and regulate gene expression [112,113,114], but it is unclear whether post-translational modifications of Hox factors is a CRM-dependent process or a global mechanism of regulating Hox factor activity. If the former is true, recruitment of post-translational modifiers to DNA could potentially contribute to paralog-specific activity of individual CRMs as well. 3. Target Accessibility In order for Hox factors to regulate gene expression, they must gain access to target CRMs. There is substantial evidence that the binding patterns of most TFs are primarily determined by nucleosome positioning [115]. Are Hox factors like most TFs in that they only bind genomic sequences found in open chromatin or do they have pioneer activity capable of interacting with DNA wrapped in nucleosomes? Here, we present recent evidence that the Hox factor binding may be largely constrained to open chromatin and provide examples demonstrating that Hox factors rely on other TFs to open chromatin prior to Hox binding. If Hox factors significantly alter the genomic chromatin landscape, one would predict that serially homologous tissues that express different Hox factors along the body plan would differ in their chromatin accessibility profiles. In Drosophila, the wing and haltere are serially homologous appendages with the wing arising from the second thoracic segment (T2) and the haltere developing from the third thoracic segment (T3) [116]. Previous studies demonstrated that the Hox factor, Ubx, is necessary to specify the haltere. In fact, genetic removal of Ubx can transform the haltere into wing tissue, whereas the misexpression of Ubx in the second thoracic segment transforms the wing into a haltere-like structure [7,117]. Comparisons of genome accessibility between the haltere and wing discs, however, revealed a largely similar chromatin landscape [118]. This finding suggests that the expression of Ubx does not alter cell fates by global changes in chromatin structure, but by acting upon accessible regulatory elements that are present in both the wing and haltere imaginal discs. As a more direct test of the ability of Hox factors to alter genome accessibility, recent studies in cell culture revealed that Hox factors may differ in their ability to bind closed versus open chromatin. Chromatin-immunoprecipitation of Hox factors followed by genomic sequencing (ChIP-seq) in Drosophila cell culture found that two Hox factors are largely constrained by nucleosome positioning [119]. In this study, ChIP was performed for transiently transfected Ubx or Abd-A in Drosophila Kc167 embryonic cell lines, which revealed that more than 94% of Ubx and Abd-A ChIP peaks occurred within pre-existing DNaseI hypersensitive regions of the genome. The findings for Ubx are consistent with the global accessibility profiles found in the haltere vs. wing described above [118]. In contrast to Ubx and Abd-A, however, a substantial proportion (25%) of Abd-B-associated ChIP peaks in Drosophila cell culture were located in previously closed chromatin [119]. These studies suggest that Hox factors may fundamentally differ in their ability to associate with nucleosome-bound genomic regions. The finding that at least a subset of Hox factors bind predominantly open chromatin leads to an interesting question: how do Hox factors gain access to CRMs that exist in closed genomic states? At least part of the answer appears to be through interactions with PBC and HMP proteins. As described earlier, heterodimerization of PBC and HMP allows translocation of the PBC proteins from the cytoplasm to the nucleus [91]. Since Kc167 cells lack Hth expression, Exd remains cytoplasmic, and thus, the above ChIP-seq studies performed on Ubx and Abd-A were done in the absence of PBC and HMP proteins. To determine what impact Exd and Hth expression has on Hox binding profiles, the authors performed complimentary ChIP-seq studies in cells co-transfected with Hth [119]. These experiments revealed that the proportion of Ubx ChIP peaks in DNaseI insensitive regions expanded from 5% to 17%, and the total number of Ubx ChIP peaks doubled. Moreover, the additional sites that were bound by Ubx were enriched for PBC and HMP binding motifs. Altogether these studies suggest that by itself, Ubx does not substantially change nucleosome positioning; and that PBC and HMP proteins enhance Ubx binding to a greater number of sites, including those in closed chromatin. Mechanistic support for the idea that PBC and HMP proteins can change the chromatin status of Hox targets prior to Hox binding comes from recent studies in zebrafish [120]. A complex of HoxB1b-Pbx-Prep/Meis activates hoxb1a transcription in the early zebrafish embryo [66]. Maternally-loaded Pbx and Prep bind the hoxb1a locus and cause histone acetylation of the locus prior to the expression of HoxB1b or transcription of hoxb1a [121]. Additionally, Pbx and Prep were shown to recruit RNA polymerase II to the hoxb1a promoter and maintain RNA polymerase II in a paused state. Slightly later in embryogenesis, HoxB1b is expressed and binds the hoxb1a locus, where it recruits factors that phosphorylate RNA polymerase II to promote transcriptional elongation [120]. Given the role of histone acetylation in changing nucleosome positioning, these studies suggest that PBC and HMP open chromatin and allow Hox factors to gain access to targets [122]. In total, these studies demonstrate that PBC and HMP proteins can contribute to Hox genomic binding patterns (Figure 4). Hence, PBC and HMP binding sites in Hox-regulated CRMs may provide Hox factors greater access to CRMs. Consistent with this idea, several genome-wide binding studies demonstrate substantial overlap between the binding of Hox factors and PBC/HMP proteins [123,124,125]. However, these studies did not determine if the PBC/HMP factors were bound to genomic regions prior to Hox factor recruitment versus direct cooperative binding of PBC/HMP/Hox complexes to DNA. Thus, future studies on the potential role of PBC/HMP proteins as pioneer factors are needed to determine if PBC/HMP proteins broadly define Hox target selection via the opening of chromatin prior to Hox factor recruitment. 4. Effect on Transcription Like many TFs, individual Hox factors can both activate and repress target expression [38,126]. Although, Hox factors have been shown to recruit and/or interact with factors known to affect transcriptional outcomes, i.e., chromatin remodelers, the mediator complex and the general transcriptional machinery [16,17,124,127,128,129,130,131], it is largely unclear what determines whether a Hox factor will act as an activator or a repressor on a given target gene. One example that explores the mechanism of Hox-mediated repression versus activation is the mammalian osteocalcin promoter [128,132,133]. The osteocalcin promoter contains adjacent Hox-Pbx binding sites, and in pre-osteoblasts, Pbx1 complexes with HoxA10 and recruits HDACs to the osteocalcin promoter, which represses gene expression. However, as pre-osteoblasts differentiate into osteoblasts, Pbx1 expression decreases, and HoxA10 then recruits CBP/p300 to acetylate osteocalcin and activate gene expression. These data suggest that the same PBC-Hox binding site can function in either gene activation or repression dependent on the presence of a PBC protein. The binding of other TFs besides PBC and HMP can also influence whether a Hox factor activates or represses a target gene. For example, the Abd-A Hox factor can both activate the expression of rhomboid in abdominal sensory cells and repress the expression of Distal-less (Dll) in the abdominal ectoderm [40,46]. As described above, Abd-A activates the rhoA CRM via interactions with Pax2, and the RhoA CRM contains nearby binding sites for both of these factors [47]. However, Abd-A represses Dll expression via multiple Hox/PBC and Hox/HMP binding sites, and this repression activity requires the nearby binding sites for two additional transcription factors, Sloppy-paired (Slp, a FoxG homolog) and Engrailed (En), which are known to recruit the Groucho co-repressor protein [46,134,135,136]. Hence, the decision for this Hox factor to activate versus repress is dependent on which additional TF sites are located nearby. Since we know a great deal about how the Hox factors regulate Dll expression, we use this example as a case study below to highlight different mechanisms of Hox specificity. 5. Case Study: The Distal-less Conserved Regulatory Element A number of the concepts underlying Hox specificity are represented in a single case study of how different Hox factors bind to and regulate the DCRE, a Hox-regulated CRM that controls Distal-less (Dll) expression in Drosophila embryos. Dll expression is essential for the specification of appendages, such as legs from the thorax [137]. Therefore, the viability of Drosophila depends on activating Dll expression in thoracic segments and repressing its expression in the abdomen. Several Dll CRMs have been identified, and the DMX is the best-characterized regulatory element that is active in the early leg precursor cells [46,70,138]. Initial studies suggested the DMX is composed of separable activation (DMEact) and repression elements (DCRE) [39,46]. More recent studies paint a more complex picture with the DMEact and the DCRE contributing to both thoracic activation and abdominal repression [70]. For the purpose of this review, we will focus on how the DCRE utilizes several different Hox/PBC and Hox/HMP sites, as well as additional TF binding sites to contribute to paralog-specific activation versus paralog-specific repression. Functional studies have revealed that the DCRE contains two Hox/PBC and one Hox/HMP site that are bound and regulated differentially by the Antp, Ubx, Abd-A and Abd-B Hox factors (Figure 5A,B) [46,70,138,139]. In the thorax, Antp cooperates with Exd/Hth to stimulate enhancer activity via binding the two Hox/PBC sites of the DCRE [70]. However, the Hox/HMP sites are not required for this activity, indicating that Antp may only specifically work on Hox/PBC sites and not Hox/HMP sites. While genetic and reporter studies have revealed that Antp is necessary for enhancing thoracic gene expression, it is currently unclear how Antp activates gene expression and whether additional binding sites in the DCRE are required for this activity. In the abdomen, the abdominal Hox factors (Ubx, Abd-A and Abd-B) repress gene expression through the DCRE using two different cell-specific mechanisms. The Drosophila embryo is segmented, and each segment is composed of distinct anterior and posterior compartment cell types [140]. Mutagenesis studies on the DCRE revealed that mutations in the Hox/PBC and Hox/HMP sites resulted in a loss in repression activity in both cell types [46,70]. In contrast, a subset of mutations outside of the Hox/PBC and Hox/HMP sites revealed that distinct TF binding sites were required for anterior- versus posterior-compartment-mediated repression [46]. In anterior compartment cells, Slp binding sites are necessary to repress gene expression, whereas in posterior compartment cells, an En binding site is needed for repression [46]. Hence, both Slp and En are necessary for complete repression in the abdomen, because Slp expression is limited to the anterior compartment of embryonic segments, while En expression in limited to the posterior compartment of embryonic segments [141,142]. Slp and En both directly interact with Groucho, a histone deacetylase, and recruit it to CRMs [134,135,136]. Thus, it is thought that abdominal Hox factors promote a repressive chromatin state on the DCRE. Moreover, chromatin studies of the Dll locus demonstrate that in the thorax, the DMEact and Dll promoter physically communicate (i.e., DNA looping), but DCRE-mediated repression in the abdomen disrupts this communication [143] (Figure 4B). These data help to demonstrate how the same Hox-regulated CRM can mediate multiple outputs depending on the cellular context. Our advanced understanding of how the DCRE mediates three distinct functions (thoracic activation vs. anterior and posterior compartment repression) raises multiple questions in regards to Hox specificity. First, since Slp and En are expressed in both the abdomen and the thorax, so why does Antp not repress gene expression via the DCRE? Different preferences for TF interactors provide insight into this question: abdominal Hox factors have been shown to form complexes with En on DNA and to interact with Slp via fluorescent complementation assays, but Antp fails to do so in the same assays [46,103]. Hence, without being able to interact with Slp an En, Antp is unable to repress gene expression via the DCRE. Second, how do the specific sequences found within the two Hox/PBC and one Hox/HMP binding sites contribute to paralog specificity? As described above, the Hox/HMP site only contributes to abdominal repression and not to thoracic activation [70]. This finding is consistent with previous studies that suggest that only posterior Hox factors mediate strong interactions with HMP proteins on Hox/HMP sites [26,44,45]. In addition, the analysis of the two Hox/PBC sites suggests that both are non-optimal sequences with either an abnormal spacing (inclusion of an additional nucleotide between the PBC and Hox site) or with several mismatches from consensus PBC/Hox sites [23]. Moreover, Electromobility Shift Assays (EMSAs) demonstrate that Abd-A forms stronger complexes with Exd and Hth on the DCRE sequence than does Antp [70]. This may be due to the fact that the DCRE contains adjacent Hox-Hth binding sites that favor more posterior Hox factors [26,44,45]. Moreover, the molecular basis of this difference in affinity may be related to the presence of the UbdA motif in Abd-A (and Ubx), whereas Antp lacks this motif. For example, mutations in the UbdA motif in Ubx disrupt DCRE binding and repression more than mutations in the HX motif [86]. These findings suggest that the DCRE sequence achieves paralog specificity by favoring Hox paralogs that contain the UbdA motif (Ubx and Abd-A, not Antp). Third, does the arrangement of Hox-PBC-HMP binding sites contribute to Hox paralog specificity? It has been previously suggested that the arrangement of Hox-PBC-HMP binding sites may contribute to the activity of Hox-regulated CRM independent of paralog specificity [144]. To test this idea, we recently altered the configuration of Hox sites within the DCRE by swapping them with the Exd/Hth/Hox site found within the RhoA CRM [70]. As described, Abd-A/Exd/Hth complexes repress gene expression via the DCRE and activate gene expression via the RhoA. We found that this hybrid CRM is able to mediate gene repression equivalently to wild-type DCRE in the anterior compartments of the abdomen. These results demonstrate that, at least in this case, the Hox-PBC-HMP binding site arrangement does not determine the activity of the CRM. Fourth, does post-translational modifications of Hox factors alter their ability to regulate the DCRE? In Drosophila, Ubx inhibits limb development in the abdomen by repressing expression of Dll via the DCRE [46,138]. However, Ubx orthologs in other arthropods have been shown to lack this repression activity [145,146]. In fact, Ubx is expressed in cells that give rise to legs along the trunk in Artemia (brine shrimp) [147,148]. In contrast to Drosophila Ubx, Artemia Ubx can be phosphorylated by CKII at serine and threonine residues in its C-terminal region [106]. Mutating phosphorylation sites in Artermia Ubx results in an increased ability to repress Dll when misexpressed in Drosophila embryos. Moreover, mutating Drosophila Ubx to add phosphorylation sites to its C-terminal region reduces the ability of the mutant Ubx to repress Dll expression [108]. These studies suggest that an evolutionary divergence of legs along the trunk of Artemia and Drosophila may be in part due to divergence in the ability of Ubx to be phosphorylated in these two species. Fifth, do all of the Hox factors that regulate Dll expression require the function of Exd and Hth? Exd and Hth cooperate with Antp, as well as Ubx/Abd-A to mediate gene activation and repression, respectively [46]. However, Exd and Hth have been found to antagonize Abd-B-mediated repression at the DCRE [139]. In fact, Abd-B represses Hth expression in the most posterior abdominal segments during embryogenesis. Since Hth is necessary for nuclear localization of Exd, the repression of Hth expression by Abd-B results in a PBC/HMP-free region. Hence, Abd-B creates an Exd/Hth-free area and represses the DCRE in an Exd/Hth independent manner [91,139]. 6. Conclusions Throughout this review, we have described multiple mechanisms by which Hox paralogs may differentiate their binding specificities and activities. These include: (1) latent specificity revealed by PBC proteins; (2) utilization of low affinity sites; (3) multiple interaction surfaces for PBC proteins; (4) preference of HMP proteins to cooperatively bind DNA with posterior Hox paralogs; (5) differential affinity for other TFs; and (6) differential regulation by post-translational modifiers (Figure 2). Additionally, we reviewed how interactions with other TFs regulate genomic binding patterns of Hox factors, as well as their effect on transcription. While we focused on Hox factors in this review, many of the same problems of specificity presented here apply to other families of transcription factors. However, it is unclear whether the different TF families primarily rely on similar strategies or distinct strategies to overcome these problems of specificity. Elucidating the molecular mechanisms of how TFs bind specific targets and produce specific regulatory outcomes will be an important step in our understanding of how eukaryotes produce robust and specific gene expression patterns throughout development. Acknowledgments The authors wish to acknowledge the National Institutes of Health (NIH) grant (GM079428) for funding this work. Conflicts of Interest The authors declare no conflict of interest. The funding sponsors had no role in the design of the study; in the collection, analyses or interpretation of data; in the writing of the manuscript; nor in the decision to publish the results. Abbreviations The following abbreviations are used in this manuscript: CRMcis-regulatory module LabLabial ScrSex-combs Reduced AntpAntennapedia UbxUltrabithorax Abd-AAbdominal-A Abd-BAbdominal-B TALEThree Amino-acid Loop Extension ExdExtradenticle HthHomothorax Figure 1 Schematic of Hox gene locus and Hox proteins. (A) Schematic of Hox gene clusters in C. elegans, D. melanogaster and M. musculus. Genes with similar colors are thought to derive from a common ancestor [10]; (B) Schematic of a Hox protein with regions labeled and described in boxes (bottom) [11,12,13,14,15,16,17,18]. Per-amino-acid conservation score across Drosophila Hox protein sequences demonstrates that the HX and homeodomain regions are the most conserved regions across paralogous Hox proteins. Multiple sequence alignment was produced via ClustalΩ (top) [19,20]. Note: The size of each Hox protein region is not to scale. Figure 2 Summary of the mechanisms of Hox paralog specificity. (A) A single Hox site is unlikely to differentiate between two different Hox paralogs; (B,C) Low-affinity Hox-PBC binding sites are more likely to differentiate between Hox paralogs than high affinity Hox-PBC binding sites [23,41]; (D,E) All Hox paralogs can typically bind adjacent Hox-PBC sites, but adjacent Hox-HMP sites have a preference for posterior Hox paralogs [26,44,45]; (F) Interactions with a nearby binding site for another TF can cause a CRM to be paralog specific [46,47]; (G) The same post-translational modifications (gray circle) can affect paralogous Hox proteins differently [48,49,50,51]. “P” (PBC protein) and “H” (HMP protein). Figure 3 Interaction between Hox factors and PBC/HMP proteins. (A) Names of PBC and HMP homeodomain proteins in C. elegans, Drosophila and vertebrates; (B) Motifs in Hox factors used to mediate interactions with PBC and HMP proteins (top). Yeast-2-hybrid data suggest sequences N-terminal to the homeodomain mediate interactions with homothorax [26]. Hox proteins can mediate interactions with PBC proteins via the HX motif or, in the case of non-vertebrate Abd-A and Ubx homologs, the UbdA motif. Structural panels: Structures produced by Foos et al., 2015, demonstrating both HX and UbdA interaction modes between Drosophila Ubx and Exd. PBC protein in green; Ubx protein in purple; HX motif in red; UbdA motif in yellow [18]. When bound to a canonical Hox-PBC binding site, Hox and PBC proteins bind on opposite sides of the DNA. Left structural panel: The HX motif (red) mediates interactions with a hydrophobic pocket formed by helix 1, helix 3 and the TALE motif of the Exd homeodomain. Right structural panel: The UbdA motif mediates interactions with helix 3 of the PBC homeodomain. UbdA can either be unstructured or form an α-helical extension of the third helix of the Hox homeodomain. Note: The Ubx protein fragment in the left panel contains both the HX and UbdA motifs, while the Ubx protein fragment in the right panel only contains the UbdA motif. (C) Examples of CRMs with Hox-PBC-HMP binding sites, demonstrating variations in the order, orientations and spacing of the binding sites. The space between HMP and Hox-PBC sites is indicated. EVIII [64]; Lab 48/95 [65]; R3-PM2 [66]; PP2 [67]; PM-PH2 [68]; PHP1 [69]; RhoA [40]; DCRE [70]. Figure 4 PBC and HMP can provide Hox factors access to targets. (A) Evidence suggests that Hox factors are typically limited to binding DNA in open chromatin regions [119]. (B,C) Expression of PBC and HMP protein has been shown to allow binding of Hox proteins to previously closed chromatin [119]. Prior to Hox binding, HMP and PBC proteins can bind DNA to promote chromatin modifications associated with chromatin opening, such as acetylation [120]. Figure 5 Regulation of Distal-less (Dll) gene expression by the DCRE in Drosophila. Dll is expressed in the Drosophila embryonic thorax, but not the abdomen. (A) Dll expression is controlled by two elements: the DMEact (an activator) and the DCRE (a mixed repressor and activator). DMEact is sufficient to drive expression in both the abdomen and the thorax, whereas the DCRE inhibits abdominal expression and enhances DMEact driven thoracic expression [70]; (B) Sequence of the DCRE, with known binding sites labeled. Slp and En binding sites are necessary for repressive DCRE activity in the abdomen [46,70]; (C) In the thorax, the chromatin structure of the Dll locus places the DMEact in proximity to the Dll promoter, whereas DNA looping between the DMEact regulatory region and the Dll promoter is not observed in abdominal segments [143]. These findings are consistent with results demonstrating that the DCRE “boosts” thoracic DMEact activation of Dll transcription via a mechanism dependent on Antp (the thoracic Hox factor) [70]. In contrast, the DCRE represses Dll transcription via mechanisms dependent on the abdominal Hox factors (Abd-A, Ubx and Abd-B) in the abdomen [46,70,138,139]. However, it is currently unclear if the thoracic and/or abdominal Hox factors directly regulate DNA looping between the distal leg enhancer and the proximal Dll promoter region. ==== Refs References 1. McGinnis W. Krumlauf R. Homeobox genes and axial patterning Cell 1992 68 283 302 10.1016/0092-8674(92)90471-N 1346368 2. Cerdá-Esteban N. Spagnoli F.M. Glimpse into Hox and tale regulation of cell differentiation and reprogramming Dev. Dyn. 2014 243 76 87 10.1002/dvdy.24075 24123411 3. Schneuwly S. Klemenz R. Gehring W.J. Redesigning the body plan of Drosophila by ectopic expression of the homoeotic gene Antennapedia Nature 1987 325 816 818 10.1038/325816a0 3821869 4. Wang K.C. Helms J.A. Chang H.Y. Regeneration, repair and remembering identity: The three Rs of Hox gene expression Trends Cell Biol. 2009 19 268 275 10.1016/j.tcb.2009.03.007 19428253 5. Friedrich J. Sorge S. Bujupi F. Eichenlaub M.P. Schulz N.G. Wittbrodt J. Lohmann I. Hox Function is Required for the Development and Maintenance of the Drosophila Feeding Motor Unit Cell Rep. 2016 14 850 860 10.1016/j.celrep.2015.12.077 26776518 6. Pascual-Anaya J. D’Aniello S. Kuratani S. Garcia-Fernàndez J. Evolution of Hoxgene clusters in deuterostomes BMC Dev. Biol. 2013 13 26 10.1186/1471-213X-13-26 23819519 7. Lewis E.B. A gene complex controlling segmentation in Drosophila Nature 1978 276 565 570 10.1038/276565a0 103000 8. Monteiro A.S. Ferrier D.E.K. Hox genes are not always Colinear Int. J. Biol. Sci. 2006 2 95 103 10.7150/ijbs.2.95 16763668 9. Wellik D.M. Hox Genes and Vertebrate Axial Pattern Curr. Top. Dev. Biol. 2009 88 257 278 19651308 10. Swalla B.J. Building divergent body plans with similar genetic pathways Heredity 2006 97 235 243 10.1038/sj.hdy.6800872 16868565 11. Bondos S.E. Swint-Kruse L. Matthews K.S. Flexibility and Disorder in Gene Regulation: LacI/GalR and Hox Proteins J. Biol. Chem. 2015 290 24669 24677 10.1074/jbc.R115.685032 26342073 12. Passner J.M. Ryoo H.D. Shen L. Mann R.S. Aggarwal A.K. Structure of a DNA-bound Ultrabithorax–Extradenticle homeodomain complex Nature 1999 397 714 719 10.1038/17833 10067897 13. Piper D.E. Batchelor A.H. Chang C.-P. Cleary M.L. Wolberger C. Structure of a HoxB1–Pbx1 Heterodimer Bound to DNA: Role of the Hexapeptide and a Fourth Homeodomain Helix in Complex Formation Cell 1999 96 587 597 10.1016/S0092-8674(00)80662-5 10052460 14. Papadopoulos D.K. Reséndez-Pérez D. Cárdenas-Chávez D.L. Villanueva-Segura K. Canales-del-Castillo R. Felix D.A. Fünfschilling R. Gehring W.J. Functional synthetic Antennapedia genes and the dual roles of YPWM motif and linker size in transcriptional activation and repression Proc. Natl. Acad. Sci. USA 2011 108 11959 11964 10.1073/pnas.1108686108 21712439 15. Saadaoui M. Merabet S. Litim-Mecheri I. Arbeille E. Sambrani N. Damen W. Brena C. Pradel J. Graba Y. Selection of distinct Hox-Extradenticle interaction modes fine-tunes Hox protein activity Proc. Natl. Acad. Sci. USA 2011 108 2276 2281 10.1073/pnas.1006964108 21262810 16. Prince F. Katsuyama T. Oshima Y. Plaza S. Resendez-Perez D. Berry M. Kurata S. Gehring W.J. The YPWM motif links Antennapedia to the basal transcriptional machinery Development 2008 135 1669 1679 10.1242/dev.018028 18367556 17. Boube M. Hudry B. Immarigeon C. Carrier Y. Bernat-Fabre S. Merabet S. Graba Y. Bourbon H.-M. Cribbs D.L. Drosophila melanogaster Hox transcription factors access the RNA polymerase II machinery through direct homeodomain binding to a conserved motif of mediator subunit Med19 PLoS Genet. 2014 10 e1004303 10.1371/journal.pgen.1004303 24786462 18. Foos N. Maurel-Zaffran C. Maté M.J. Vincentelli R. Hainaut M. Berenger H. Pradel J. Saurin A.J. Ortiz-Lombardía M. Graba Y. A Flexible Extension of the Drosophila Ultrabithorax Homeodomain Defines a Novel Hox/PBC Interaction Mode Structure 2015 23 270 279 10.1016/j.str.2014.12.011 25651060 19. Livingstone C.D. Barton G.J. Protein sequence alignments: A strategy for the hierarchical analysis of residue conservation Comput. Appl. Biosci. 1993 9 745 756 10.1093/bioinformatics/9.6.745 8143162 20. Sievers F. Wilm A. Dineen D. Gibson T.J. Karplus K. Li W. Lopez R. McWilliam H. Remmert M. Söding J. Fast, scalable generation of high-quality protein multiple sequence alignments using Clustal Omega Mol. Syst. Biol. 2011 7 10.1038/msb.2011.75 21988835 21. Noyes M.B. Christensen R.G. Wakabayashi A. Stormo G.D. Brodsky M.H. Wolfe S.A. Analysis of homeodomain specificities allows the family-wide prediction of preferred recognition sites Cell 2008 133 1277 1289 10.1016/j.cell.2008.05.023 18585360 22. Berger M.F. Badis G. Gehrke A.R. Talukder S. Philippakis A.A. Peña-Castillo L. Alleyne T.M. Mnaimneh S. Botvinnik O.B. Chan E.T. Variation in Homeodomain DNA Binding Revealed by High-Resolution Analysis of Sequence Preferences Cell 2008 133 1266 1276 10.1016/j.cell.2008.05.024 18585359 23. Slattery M. Riley T. Liu P. Abe N. Gomez-Alcala P. Dror I. Zhou T. Rohs R. Honig B. Bussemaker H.J. Cofactor Binding Evokes Latent Differences in DNA Binding Specificity between Hox Proteins Cell 2011 147 1270 1282 10.1016/j.cell.2011.10.053 22153072 24. Zhao J.J. Lazzarini R.A. Pick L. Functional dissection of the mouse Hox-a5 gene EMBO J. 1996 15 1313 1322 8635464 25. Tan X.-X. Bondos S. Li L. Matthews K.S. Transcription Activation by Ultrabithorax Ib Protein Requires a Predicted α-Helical Region Biochemistry 2002 41 2774 2785 10.1021/bi011967y 11851425 26. Williams T.M. Williams M.E. Innis J.W. Range of HOX/TALE superclass associations and protein domain requirements for HOXA13:MEIS interaction Dev. Biol. 2005 277 457 471 10.1016/j.ydbio.2004.10.004 15617687 27. Hsiao H.-C. Gonzalez K.L. Catanese D.J. Jr. Jordy K.E. Matthews K.S. Bondos S.E. The Intrinsically Disordered Regions of the Drosophila melanogaster Hox Protein Ultrabithorax Select Interacting Proteins Based on Partner Topology PLoS ONE 2014 9 e108217 10.1371/journal.pone.0108217 25286318 28. Regulski M. Harding K. Kostriken R. Karch F. Levine M. McGinnis W. Homeo box genes of the Antennapedia and Bithorax Complexes of Drosophila Cell 1985 43 71 80 10.1016/0092-8674(85)90013-3 2416463 29. Rezsohazy R. Non-transcriptional interactions of Hox proteins: inventory, facts, and future directions Dev. Dyn. 2014 243 117 131 10.1002/dvdy.24060 24115586 30. Weirauch M.T. Yang A. Albu M. Cote A.G. Montenegro-Montero A. Drewe P. Najafabadi H.S. Lambert S.A. Mann I. Cook K. Determination and Inference of Eukaryotic Transcription Factor Sequence Specificity Cell 2014 158 1431 1443 10.1016/j.cell.2014.08.009 25215497 31. Martin A. Serano J.M. Jarvis E. Bruce H.S. Wang J. Ray S. Barker C.A. O’Connell L.C. Patel N.H. CRISPR/Cas9 Mutagenesis Reveals Versatile Roles of Hox Genes in Crustacean Limb Specification and Evolution Curr. Biol. 2016 26 14 26 10.1016/j.cub.2015.11.021 26687626 32. Niu W. Lu Z.J. Zhong M. Sarov M. Murray J.I. Brdlik C.M. Janette J. Chen C. Alves P. Preston E. Diverse transcription factor binding features revealed by genome-wide ChIP-seq in C. elegans Genome Res. 2011 21 245 254 10.1101/gr.114587.110 21177963 33. Bach C. Buhl S. Mueller D. García-Cuéllar M.-P. Maethner E. Slany R.K. Leukemogenic transformation by HOXA cluster genes Blood 2010 115 2910 2918 10.1182/blood-2009-04-216606 20130239 34. Misra M. Sours E. Lance-Jones C. Hox transcription factors influence motoneuron identity through the integrated actions of both homeodomain and non-homeodomain regions Dev. Dyn. 2012 241 718 731 10.1002/dvdy.23763 22411553 35. Gutierrez A. Knoch L. Witte H. Sommer R.J. Functional specificity of the nematode Hox gene mab-5 Development 2003 130 983 993 10.1242/dev.00320 12538523 36. Hueber S.D. Bezdan D. Henz S.R. Blank M. Wu H. Lohmann I. Comparative analysis of Hox downstream genes in Drosophila Development 2007 134 381 392 10.1242/dev.02746 17166915 37. Pearson J.C. Lemons D. McGinnis W. Modulating Hox gene functions during animal body patterning Nat. Rev. Genet. 2005 6 893 904 10.1038/nrg1726 16341070 38. Mann R.S. Lelli K.M. Joshi R. Hox Specificity : Unique Roles for Cofactors and Collaborators Curr. Top. Dev. Biol. 2009 88 63 101 19651302 39. Gebelein B. Culi J. Ryoo H.D. Zhang W. Mann R.S. Specificity of Distalless repression and limb primordia development by abdominal Hox proteins Dev. Cell 2002 3 487 498 10.1016/S1534-5807(02)00257-5 12408801 40. Li-Kroeger D. Witt L. Grimes H. Cook T.A. Gebelein B. Hox and Senseless Antagonism Functions as a Molecular Switch to Regulate EGF Secretion in the Drosophila PNS Dev. Cell 2008 15 298 308 10.1016/j.devcel.2008.06.001 18694568 41. Crocker J. Abe N. Rinaldi L. McGregor A.P. Frankel N. Wang S. Alsawadi A. Valenti P. Plaza S. Payre F. Low affinity binding site clusters confer hox specificity and regulatory robustness Cell 2015 160 191 203 10.1016/j.cell.2014.11.041 25557079 42. Hirth F. Loop T. Egger B. Miller D.F.B. Kaufman T.C. Reichert H. Functional equivalence of Hox gene products in the specification of the tritocerebrum during embryonic brain development of Drosophila Development 2001 128 4781 4788 11731458 43. Raines A.M. Magella B. Adam M. Potter S.S. Key pathways regulated by HoxA9,10,11/HoxD9,10,11 during limb development BMC Dev. Biol. 2015 15 28 10.1186/s12861-015-0078-5 26186931 44. Shen W.F. Montgomery J.C. Rozenfeld S. Moskow J.J. Lawrence H.J. Buchberg A.M. Largman C. AbdB-like Hox proteins stabilize DNA binding by the Meis1 homeodomain proteins Mol. Cell. Biol. 1997 17 6448 6458 10.1128/MCB.17.11.6448 9343407 45. Jolma A. Yin Y. Nitta K.R. Dave K. Popov A. Taipale M. Enge M. Kivioja T. Morgunova E. Taipale J. DNA-dependent formation of transcription factor pairs alters their binding specificity Nature 2015 527 384 388 10.1038/nature15518 26550823 46. Gebelein B. McKay D.J. Mann R.S. Direct integration of Hox and segmentation gene inputs during Drosophila development Nature 2004 431 653 659 10.1038/nature02946 15470419 47. Li-Kroeger D. Cook T.A. Gebelein B. Integration of an abdominal Hox complex with Pax2 yields cell-specific EGF secretion from Drosophila sensory precursor cells Development 2012 139 1611 1619 10.1242/dev.077842 22438572 48. Wu X. Ellmann S. Rubin E. Gil M. Jin K. Han L. Chen H. Kwon E.M. Guo J. Ha H.C. ADP ribosylation by PARP-1 suppresses HOXB7 transcriptional activity PLoS ONE 2012 7 e40644 10.1371/journal.pone.0040644 22844406 49. Eklund E.A. Jalava A. Kakar R. Tyrosine Phosphorylation of HoxA10 Decreases DNA Binding and Transcriptional Repression during Interferon γ-induced Differentiation of Myeloid Leukemia Cell Lines J. Biol. Chem. 2000 275 20117 20126 10766757 50. Vijapurkar U. Fischbach N. Shen W. Brandts C. Stokoe D. Lawrence H.J. Largman C. Protein kinase C-mediated phosphorylation of the leukemia-associated HOXA9 protein impairs its DNA binding ability and induces myeloid differentiation Mol. Cell. Biol. 2004 24 3827 3837 10.1128/MCB.24.9.3827-3837.2004 15082777 51. Yaron Y. McAdara J.K. Lynch M. Hughes E. Gasson J.C. Identification of novel functional regions important for the activity of HOXB7 in mammalian cells J. Immunol. 2001 166 5058 5067 10.4049/jimmunol.166.8.5058 11290787 52. Peifer M. Wieschaus E. Mutations in the Drosophila gene extradenticle affect the way specific homeo domain proteins regulate segmental identity Genes Dev. 1990 4 1209 1223 10.1101/gad.4.7.1209 1976570 53. Nourse J. Mellentin J.D. Galili N. Wilkinson J. Stanbridge E. Smith S.D. Cleary M.L. Chromosomal translocation t(1;19) results in synthesis of a homeobox fusion mRNA that codes for a potential chimeric transcription factor Cell 1990 60 535 545 10.1016/0092-8674(90)90657-Z 1967982 54. Kamps M.P. Murre C. Sun X. Baltimore D. A new homeobox gene contributes the DNA binding domain of the t(1;19) translocation protein in pre-B all Cell 1990 60 547 555 10.1016/0092-8674(90)90658-2 1967983 55. Bürglin T.R. Analysis of TALE superclass homeobox genes (MEIS, PBC, KNOX, Iroquois, TGIF) reveals a novel domain conserved between plants and animals Nucleic Acids Res. 1997 25 4173 4180 10.1093/nar/25.21.4173 9336443 56. Chan S.-K. Jaffe L. Capovilla M. Botas J. Mann R.S. The DNA binding specificity of ultrabithorax is modulated by cooperative interactions with extradenticle, another homeoprotein Cell 1994 78 603 615 10.1016/0092-8674(94)90525-8 7915199 57. Chang C.P. Brocchieri L. Shen W.F. Largman C. Cleary M.L. Pbx modulation of Hox homeodomain amino-terminal arms establishes different DNA-binding specificities across the Hox locus Mol. Cell. Biol. 1996 16 1734 1745 10.1128/MCB.16.4.1734 8657149 58. Pöpperl H. Bienz M. Studer M. Chan S.K. Aparicio S. Brenner S. Mann R.S. Krumlauf R. Segmental expression of Hoxb-1 is controlled by a highly conserved autoregulatory loop dependent upon exd/pbx Cell 1995 81 1031 1042 10.1016/S0092-8674(05)80008-X 7600572 59. Rauskolb C. Peifer M. Wieschaus E. Extradenticle, a regulator of homeotic gene activity, is a homolog of the homeobox-containing human proto-oncogene pbx1 Cell 1993 74 1101 1112 10.1016/0092-8674(93)90731-5 8104703 60. Rauskolb C. Wieschaus E. Coordinate regulation of downstream genes by extradenticle and the homeotic selector proteins EMBO J. 1994 13 3561 3569 7914871 61. Sun B. Hursh D.A. Jackson D. Beachy P.A. Ultrabithorax protein is necessary but not sufficient for full activation of decapentaplegic expression in the visceral mesoderm EMBO J. 1995 14 520 535 7859741 62. Izpisúa-Belmonte J.C. Falkenstein H. Dollé P. Renucci A. Duboule D. Murine genes related to the Drosophila AbdB homeotic genes are sequentially expressed during development of the posterior part of the body EMBO J. 1991 10 2279 2289 1676674 63. Van Dijk M.A. Murre C. Extradenticle Raises the DNA binding specificity of homeotic selector gene products Cell 1994 78 617 624 10.1016/0092-8674(94)90526-6 7915200 64. Ebner A. Cabernard C. Affolter M. Merabet S. Recognition of distinct target sites by a unique Labial/Extradenticle/Homothorax complex Development 2005 132 1591 1600 10.1242/dev.01721 15753213 65. Ryoo H.D. Marty T. Casares F. Affolter M. Mann R.S. Regulation of Hox target genes by a DNA bound Homothorax/Hox/Extradenticle complex Development 1999 126 5137 5148 10529430 66. Ferretti E. Cambronero F. Tümpel S. Longobardi E. Wiedemann L.M. Blasi F. Krumlauf R. Hoxb1 Enhancer and Control of Rhombomere 4 Expression: Complex Interplay between PREP1-PBX1-HOXB1 Binding Sites Mol. Cell. Biol. 2005 25 8541 8552 10.1128/MCB.25.19.8541-8552.2005 16166636 67. Ferretti E. Marshall H. Popperl H. Maconochie M. Krumlauf R. Blasi F. Segmental expression of Hoxb2 in r4 requires two separate sites that integrate cooperative interactions between Prep1, Pbx and Hox proteins Development 2000 127 155 166 10654609 68. Tümpel S. Cambronero F. Ferretti E. Blasi F. Wiedemann L.M. Krumlauf R. Expression of Hoxa2 in rhombomere 4 is regulated by a conserved cross-regulatory mechanism dependent upon Hoxb1 Dev. Biol. 2007 302 646 660 10.1016/j.ydbio.2006.10.029 17113575 69. Manzanares M. Bel-Vialar S. Ariza-McNaughton L. Ferretti E. Marshall H. Maconochie M.M. Blasi F. Krumlauf R. Independent regulation of initiation and maintenance phases of Hoxa3 expression in the vertebrate hindbrain involve auto- and cross-regulatory mechanisms Development 2001 128 3595 3607 11566863 70. Uhl J.D. Zandvakili A. Gebelein B. A Hox Transcription Factor Collective Binds a Highly Conserved Distal-less cis -Regulatory Module to Generate Robust Transcriptional Outcomes PLoS Genet 2016 12 e1005981 10.1371/journal.pgen.1005981 27058369 71. Joshi R. Passner J.M. Rohs R. Jain R. Sosinsky A. Crickmore M.A. Jacob V. Aggarwal A.K. Honig B. Mann R.S. Functional Specificity of a Hox Protein Mediated by the Recognition of Minor Groove Structure Cell 2007 131 530 543 10.1016/j.cell.2007.09.024 17981120 72. Kim S. Broströmer E. Xing D. Jin J. Chong S. Ge H. Wang S. Gu C. Yang L. Gao Y.Q. Probing Allostery Through DNA Science 2013 339 816 819 10.1126/science.1229223 23413354 73. Narasimhan K. Pillay S. Huang Y.-H. Jayabal S. Udayasuryan B. Veerapandian V. Kolatkar P. Cojocaru V. Pervushin K. Jauch R. DNA-mediated cooperativity facilitates the co-selection of cryptic enhancer sequences by SOX2 and PAX6 transcription factors Nucleic Acids Res. 2015 43 1513 1528 10.1093/nar/gku1390 25578969 74. Chang C.P. Shen W.F. Rozenfeld S. Lawrence H.J. Largman C. Cleary M.L. Pbx proteins display hexapeptide-dependent cooperative DNA binding with a subset of Hox proteins Genes Dev. 1995 9 663 674 10.1101/gad.9.6.663 7729685 75. Lu Q. Kamps M.P. Heterodimerization of Hox proteins with Pbx1 and oncoprotein E2a-Pbx1 generates unique DNA-binding specifities at nucleotides predicted to contact the N-terminal arm of the Hox homeodomain--demonstration of Hox-dependent targeting of E2a-Pbx1 in vivo Oncogene 1997 14 75 83 10.1038/sj.onc.1200799 9010234 76. Abe N. Dror I. Yang L. Slattery M. Zhou T. Bussemaker H.J. Rohs R. Mann R.S. Deconvolving the recognition of DNA shape from sequence Cell 2015 161 307 318 10.1016/j.cell.2015.02.008 25843630 77. Tanay A. Extensive low-affinity transcriptional interactions in the yeast genome Genome Res. 2006 16 962 972 10.1101/gr.5113606 16809671 78. Scardigli R. Bäumer N. Gruss P. Guillemot F. Roux I.L. Direct and concentration-dependent regulation of the proneural gene Neurogenin2 by Pax6 Development 2003 130 3269 3281 10.1242/dev.00539 12783797 79. Ramos A.I. Barolo S. Low-affinity transcription factor binding sites shape morphogen responses and enhancer evolution Phil. Trans. R. Soc. B 2013 368 20130018 10.1098/rstb.2013.0018 24218631 80. Evans N.C. Swanson C.I. Barolo S. Sparkling insights into enhancer structure, function, and evolution Curr. Top. Dev. Biol. 2012 98 97 120 22305160 81. Crocker J. Noon E.P.-B. Stern D.L. The Soft Touch: Low-Affinity Transcription Factor Binding Sites in Development and Evolution Curr. Top. Dev. Biol. 2016 17 455 469 82. Chan S.K. Pöpperl H. Krumlauf R. Mann R.S. An extradenticle-induced conformational change in a HOX protein overcomes an inhibitory function of the conserved hexapeptide motif EMBO J. 1996 15 2476 2487 8665855 83. Galant R. Walsh C.M. Carroll S.B. Hox repression of a target gene: Extradenticle-independent, additive action through multiple monomer binding sites Development 2002 129 3115 3126 12070087 84. Lelli K.M. Noro B. Mann R.S. Variable motif utilization in homeotic selector (Hox)–cofactor complex formation controls specificity Proc. Natl. Acad. Sci. USA 2011 108 21122 21127 10.1073/pnas.1114118109 22160705 85. Merabet S. Kambris Z. Capovilla M. Bérenger H. Pradel J. Graba Y. The Hexapeptide and Linker Regions of the AbdA Hox Protein Regulate Its Activating and Repressive Functions Dev. Cell 2003 4 761 768 10.1016/S1534-5807(03)00126-6 12737810 86. Merabet S. Saadaoui M. Sambrani N. Hudry B. Pradel J. Affolter M. Graba Y. A unique Extradenticle recruitment mode in the Drosophila Hox protein Ultrabithorax Proc. Natl. Acad. Sci. USA 2007 104 16946 16951 10.1073/pnas.0705832104 17942685 87. Tour E. Hittinger C.T. McGinnis W. Evolutionarily conserved domains required for activation and repression functions of the Drosophila Hox protein Ultrabithorax Development 2005 132 5271 5281 10.1242/dev.02138 16284118 88. Hudry B. Remacle S. Delfini M.-C. Rezsohazy R. Graba Y. Merabet S. Hox Proteins Display a Common and Ancestral Ability to Diversify Their Interaction Mode with the PBC Class Cofactors PLoS Biol. 2012 10 e1001351 10.1371/journal.pbio.1001351 22745600 89. Jürgens G. Wieschaus E. Nüsslein-Volhard C. Kluding H. Mutations affecting the pattern of the larval cuticle inDrosophila melanogaster Wilhelm Rouxs Arch. Dev. Biol. 1984 193 283 295 10.1007/BF00848157 90. Moskow J.J. Bullrich F. Huebner K. Daar I.O. Buchberg A.M. Meis1, a PBX1-related homeobox gene involved in myeloid leukemia in BXH-2 mice Mol. Cell. Biol. 1995 15 5434 5443 10.1128/MCB.15.10.5434 7565694 91. Rieckhof G.E. Casares F. Ryoo H.D. Abu-Shaar M. Mann R.S. Nuclear Translocation of Extradenticle Requires homothorax, which Encodes an Extradenticle-Related Homeodomain Protein Cell 1997 91 171 183 10.1016/S0092-8674(00)80400-6 9346235 92. Abu-Shaar M. Ryoo H.D. Mann R.S. Control of the nuclear localization of extradenticle by competing nuclear import and export signals Genes Dev. 1999 13 935 945 10.1101/gad.13.8.935 10215621 93. Berthelsen J. Kilstrup-Nielsen C. Blasi F. Mavilio F. Zappavigna V. The subcellular localization of PBX1 and EXD proteins depends on nuclear import and export signals and is modulated by association with PREP1 and HTH Genes Dev. 1999 13 946 953 10.1101/gad.13.8.946 10215622 94. Jaw T.J. You L.R. Knoepfler P.S. Yao L.C. Pai C.Y. Tang C.Y. Chang L.P. Berthelsen J. Blasi F. Kamps M.P. Direct interaction of two homeoproteins, homothorax and extradenticle, is essential for EXD nuclear localization and function Mech. Dev. 2000 91 279 291 10.1016/S0925-4773(99)00316-0 10704852 95. Pai C.Y. Kuo T.S. Jaw T.J. Kurant E. Chen C.T. Bessarab D.A. Salzberg A. Sun Y.H. The Homothorax homeoprotein activates the nuclear localization of another homeoprotein, extradenticle, and suppresses eye development in Drosophila Genes Dev. 1998 12 435 446 10.1101/gad.12.3.435 9450936 96. Hudry B. Thomas-Chollier M. Volovik Y. Duffraisse M. Dard A. Frank D. Technau U. Merabet S. Molecular insights into the origin of the Hox-TALE patterning system eLife 2014 3 e01939 10.7554/eLife.01939 24642410 97. Berthelsen J. Zappavigna V. Ferretti E. Mavilio F. Blasi F. The novel homeoprotein Prep1 modulates Pbx–Hox protein cooperativity EMBO J. 1998 17 1434 1445 10.1093/emboj/17.5.1434 9482740 98. Jacobs Y. Schnabel C.A. Cleary M.L. Trimeric Association of Hox and TALE Homeodomain Proteins Mediates Hoxb2 Hindbrain Enhancer Activity Mol. Cell. Biol. 1999 19 5134 5142 10.1128/MCB.19.7.5134 10373562 99. Shanmugam K. Green N.C. Rambaldi I. Saragovi H.U. Featherstone M.S. PBX and MEIS as Non-DNA-Binding Partners in Trimeric Complexes with HOX Proteins Mol. Cell. Biol. 1999 19 7577 7588 10.1128/MCB.19.11.7577 10523646 100. Shen W.F. Rozenfeld S. Kwong A. Kömüves L.G. Lawrence H.J. Largman C. HOXA9 forms triple complexes with PBX2 and MEIS1 in myeloid cells Mol. Cell. Biol. 1999 19 3051 3061 10.1128/MCB.19.4.3051 10082572 101. Uhl J.D. Cook T.A. Gebelein B. Comparing anterior and posterior Hox complex formation reveals guidelines for predicting cis -regulatory elements Dev. Biol. 2010 343 154 166 10.1016/j.ydbio.2010.04.004 20398649 102. Grice J. Noyvert B. Doglio L. Elgar G. A Simple Predictive Enhancer Syntax for Hindbrain Patterning Is Conserved in Vertebrate Genomes PLoS ONE 2015 10 e0130413 10.1371/journal.pone.0130413 26131856 103. Baëza M. Viala S. Heim M. Dard A. Hudry B. Duffraisse M. Rogulja-Ortmann A. Brun C. Merabet S. Inhibitory activities of short linear motifs underlie Hox interactome specificity in vivo eLife 2015 4 e06034 10.7554/eLife.06034 25869471 104. Gong K.-Q. Yallowitz A.R. Sun H. Dressler G.R. Wellik D.M. A Hox-Eya-Pax complex regulates early kidney developmental gene expression Mol. Cell. Biol. 2007 27 7661 7668 10.1128/MCB.00465-07 17785448 105. Gavis E.R. Hogness D.S. Phosphorylation, expression and function of the Ultrabithorax protein family in Drosophila melanogaster Development 1991 112 1077 1093 1682129 106. Hsia C.C. Paré A.C. Hannon M. Ronshaugen M. McGinnis W. Silencing of an abdominal Hox gene during early development is correlated with limb development in a crustacean trunk Evol. Dev. 2010 12 131 143 10.1111/j.1525-142X.2010.00399.x 20433454 107. Jaffe L. Ryoo H.D. Mann R.S. A role for phosphorylation by casein kinase II in modulating Antennapedia activity in Drosophila Genes Dev. 1997 11 1327 1340 10.1101/gad.11.10.1327 9171376 108. Taghli-Lamallem O. Hsia C. Ronshaugen M. McGinnis W. Context-dependent regulation of Hox protein functions by CK2 phosphorylation sites Dev. Genes Evol. 2008 218 321 332 10.1007/s00427-008-0224-1 18504607 109. Eklund E.A. Goldenberg I. Lu Y. Andrejic J. Kakar R. SHP1 protein-tyrosine phosphatase regulates HoxA10 DNA binding and transcriptional repression activity in undifferentiated myeloid cells J. Biol. Chem. 2002 277 36878 36888 10.1074/jbc.M203917200 12145285 110. Lindsey S. Huang W. Wang H. Horvath E. Zhu C. Eklund E.A. Activation of SHP2 protein-tyrosine phosphatase increases HoxA10-induced repression of the genes encoding gp91(PHOX) and p67(PHOX) J. Biol. Chem. 2007 282 2237 2249 10.1074/jbc.M608642200 17138561 111. Wang H. Lindsey S. Konieczna I. Bei L. Horvath E. Huang W. Saberwal G. Eklund E.A. Constitutively active SHP2 cooperates with HoxA10 overexpression to induce acute myeloid leukemia J. Biol. Chem. 2009 284 2549 2567 10.1074/jbc.M804704200 19022774 112. Basnet H. Su X.B. Tan Y. Meisenhelder J. Merkurjev D. Ohgi K.A. Hunter T. Pillus L. Rosenfeld M.G. Tyrosine phosphorylation of histone H2A by CK2 regulates transcriptional elongation Nature 2014 516 267 271 10.1038/nature13736 25252977 113. Kraus W.L. Hottiger M.O. PARP-1 and gene regulation: Progress and puzzles Mol. Aspects Med. 2013 34 1109 1123 10.1016/j.mam.2013.01.005 23357755 114. Wang S. Jones K.A. CK2 Controls the Recruitment of Wnt Regulators to Target Genes in Vivo Curr. Biol. 2006 16 2239 2244 10.1016/j.cub.2006.09.034 17113388 115. Biggin M.D. Animal Transcription Networks as Highly Connected, Quantitative Continua Dev. Cell 2011 21 611 626 10.1016/j.devcel.2011.09.008 22014521 116. Carroll S.B. Grenier J.K. Weatherbee S.D. From DNA to Diversity: Molecular Genetics and the Evolution of Animal Design 2nd ed. Blackwell Pub Malden, MA, USA 2005 117. Casares F. Calleja M. Sánchez-Herrero E. Functional similarity in appendage specification by the Ultrabithorax and abdominal-A Drosophila HOX genes EMBO J. 1996 15 3934 3942 8670898 118. McKay D.J. Lieb J.D. A Common Set of DNA Regulatory Elements Shapes Drosophila Appendages Dev. Cell 2013 27 306 318 10.1016/j.devcel.2013.10.009 24229644 119. Beh C.Y. El-Sharnouby S. Chatzipli A. Russell S. Choo S.W. White R. Roles of cofactors and chromatin accessibility in Hox protein target specificity Epigenetics Chromatin 2016 9 10.1186/s13072-015-0049-x 26753000 120. Choe S.-K. Ladam F. Sagerström C.G. TALE factors poise promoters for activation by Hox proteins Dev. Cell 2014 28 203 211 10.1016/j.devcel.2013.12.011 24480644 121. Choe S.-K. Sagerström C.G. Variable Meis-dependence among paralog group-1 Hox proteins Biochem. Biophys. Res. Commun. 2005 331 1384 1391 10.1016/j.bbrc.2005.04.063 15883028 122. Shogren-Knaak M. Ishii H. Sun J.-M. Pazin M.J. Davie J.R. Peterson C.L. Histone H4-K16 Acetylation Controls Chromatin Structure and Protein Interactions Science 2006 311 844 847 10.1126/science.1124000 16469925 123. Amin S. Donaldson I.J. Zannino D.A. Hensman J. Rattray M. Losa M. Spitz F. Ladam F. Sagerström C. Bobola N. Hoxa2 Selectively Enhances Meis Binding to Change a Branchial Arch Ground State Dev. Cell 2015 32 265 277 10.1016/j.devcel.2014.12.024 25640223 124. Huang Y. Sitwala K. Bronstein J. Sanders D. Dandekar M. Collins C. Robertson G. MacDonald J. Cezard T. Bilenky M. Identification and characterization of Hoxa9 binding sites in hematopoietic cells Blood 2012 119 388 398 10.1182/blood-2011-03-341081 22072553 125. Choo S.W. White R. Russell S. Genome-Wide Analysis of the Binding of the Hox Protein Ultrabithorax and the Hox Cofactor Homothorax in Drosophila PLoS ONE 2011 6 e14778 10.1371/journal.pone.0014778 21483667 126. Ladam F. Sagerström C.G. Hox regulation of transcription: More complex(es) Dev. Dyn. 2014 243 4 15 10.1002/dvdy.23997 23765878 127. Collins C. Wang J. Miao H. Bronstein J. Nawer H. Xu T. Figueroa M. Muntean A.G. Hess J.L. C/EBPα is an essential collaborator in Hoxa9/Meis1-mediated leukemogenesis Proc. Natl. Acad. Sci. USA 2014 111 9899 9904 10.1073/pnas.1402238111 24958854 128. Gordon J.A.R. Hassan M.Q. Saini S. Montecino M. van Wijnen A.J. Stein G.S. Stein J.L. Lian J.B. Pbx1 represses osteoblastogenesis by blocking Hoxa10-mediated recruitment of chromatin remodeling factors Mol. Cell. Biol. 2010 30 3531 3541 10.1128/MCB.00889-09 20439491 129. Kim H.-H. Park S.J. Han J.-H. Pathak C. Cheong H.-K. Lee B.-J. Structural insight into the interaction between the Hox and HMGB1 and understanding of the HMGB1-enhancing effect of Hox-DNA binding Biochim. Biophys. Acta 2015 1854 449 459 10.1016/j.bbapap.2015.02.009 25707357 130. Lu Y. Goldenberg I. Bei L. Andrejic J. Eklund E.A. HoxA10 represses gene transcription in undifferentiated myeloid cells by interaction with histone deacetylase 2 J. Biol. Chem. 2003 278 47792 47802 10.1074/jbc.M305885200 14512427 131. Zappavigna V. Falciola L. Helmer-Citterich M. Mavilio F. Bianchi M.E. HMG1 interacts with HOX proteins and enhances their DNA binding and transcriptional activation EMBO J. 1996 15 4981 4991 8890171 132. Hassan M.Q. Tare R. Lee S.H. Mandeville M. Weiner B. Montecino M. van Wijnen A.J. Stein J.L. Stein G.S. Lian J.B. HOXA10 controls osteoblastogenesis by directly activating bone regulatory and phenotypic genes Mol. Cell. Biol. 2007 27 3337 3352 10.1128/MCB.01544-06 17325044 133. Gordon J.A.R. Hassan M.Q. Koss M. Montecino M. Selleri L. van Wijnen A.J. Stein J.L. Stein G.S. Lian J.B. Epigenetic regulation of early osteogenesis and mineralized tissue formation by a HOXA10-PBX1-associated complex Cells Tissues Organs 2011 194 146 150 10.1159/000324790 21597276 134. Andrioli L.P. Oberstein A.L. Corado M.S.G. Yu D. Small S. Groucho-dependent repression by sloppy-paired 1 differentially positions anterior pair-rule stripes in the Drosophila embryo Dev. Biol. 2004 276 541 551 10.1016/j.ydbio.2004.09.025 15581884 135. Roth M. Bonev B. Lindsay J. Lea R. Panagiotaki N. Houart C. Papalopulu N. FoxG1 and TLE2 act cooperatively to regulate ventral telencephalon formation Development 2010 137 1553 1562 10.1242/dev.044909 20356955 136. Verginelli F. Perin A. Dali R. Fung K.H. Lo R. Longatti P. Guiot M.-C. del Maestro R.F. Rossi S. di Porzio U. Transcription factors FOXG1 and Groucho/TLE promote glioblastoma growth Nat. Commun. 2013 4 2956 10.1038/ncomms3956 24356439 137. Cohen S.M. Brönner G. Küttner F. Jürgens G. Jäckle H. Distal-less encodes a homoeodomain protein required for limb development in Drosophila Nature 1989 338 432 434 10.1038/338432a0 2564639 138. Vachon G. Cohen B. Pfeifle C. McGuffin M.E. Botas J. Cohen S.M. Homeotic genes of the bithorax complex repress limb development in the abdomen of the Drosophila embryo through the target gene Distal-less Cell 1992 71 437 450 10.1016/0092-8674(92)90513-C 1358457 139. Sambrani N. Hudry B. Maurel-Zaffran C. Zouaz A. Mishra R. Merabet S. Graba Y. Distinct Molecular Strategies for Hox-Mediated Limb Suppression in Drosophila : From Cooperativity to Dispensability/Antagonism in TALE Partnership PLoS Genet 2013 9 e1003307 10.1371/journal.pgen.1003307 23505377 140. Lawrence P.A. The present status of the parasegment Development 1988 104 61 65 2472942 141. Cadigan K.M. Grossniklaus U. Gehring W.J. Localized expression of sloppy paired protein maintains the polarity of Drosophila parasegments Genes Dev. 1994 8 899 913 10.1101/gad.8.8.899 7926775 142. Grossniklaus U. Pearson R.K. Gehring W.J. The Drosophila sloppy paired locus encodes two proteins involved in segmentation that show homology to mammalian transcription factors Genes Dev. 1992 6 1030 1051 10.1101/gad.6.6.1030 1317319 143. Agelopoulos M. McKay D.J. Mann R.S. Developmental regulation of chromatin conformation by Hox proteins in Drosophila Cell Rep. 2012 1 350 359 10.1016/j.celrep.2012.03.003 22523743 144. Merabet S. Hudry B. Hox transcriptional specificity despite a single class of cofactors: are flexible interaction modes the key? Plasticity in Hox/PBC interaction modes as a common molecular strategy for shaping Hox transcriptional activities BioEssays 2013 35 88 92 10.1002/bies.201200146 23255231 145. Grenier J.K. Carroll S.B. Functional evolution of the Ultrabithorax protein Proc. Natl. Acad. Sci. USA 2000 97 704 709 10.1073/pnas.97.2.704 10639143 146. Ronshaugen M. McGinnis N. McGinnis W. Hox protein mutation and macroevolution of the insect body plan Nature 2002 415 914 917 10.1038/nature716 11859370 147. Averof M. Akam M. Hox genes and the diversification of insect and crustacean body plans Nature 1995 376 420 423 10.1038/376420a0 7630416 148. Averof M. Patel N.H. Crustacean appendage evolution associated with changes in Hox gene expression Nature 1997 388 682 686 9262403

PMC005xxxxxx/PMC5003336.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757127410.1371/journal.pone.0161062PONE-D-16-01565Research ArticleMedicine and Health SciencesMental Health and PsychiatryMood DisordersDepressionBiology and Life SciencesPsychologyEmotionsSocial SciencesPsychologyEmotionsMedicine and Health SciencesMental Health and PsychiatryPsychological StressBiology and Life SciencesPsychologyPsychological StressSocial SciencesPsychologyPsychological StressPeople and PlacesPopulation GroupingsAge GroupsAdolescentsBiology and Life SciencesNeuroscienceCognitive ScienceCognitionMedicine and Health SciencesMental Health and PsychiatryMedicine and Health SciencesHealth CarePatientsOutpatientsBiology and Life SciencesPsychologyPsychometricsSocial SciencesPsychologyPsychometricsAdolescent Depression and Negative Life Events, the Mediating Role of Cognitive Emotion Regulation Adolescent Depression and Mediation by Cognitive Emotion RegulationStikkelbroek Yvonne 1*Bodden Denise H. M. 12Kleinjan Marloes 3Reijnders Mirjam 1van Baar Anneloes L. 11 Child and Adolescent Studies, Utrecht University, PO Box 80.140, NL-3508 TC, Utrecht, The Netherlands2 Department of Developmental Psychology, Radboud University, Nijmegen, The Netherlands3 Trimbos Institute (Netherlands Institute of Mental Health and Addiction), Utrecht, The Netherlandsvan Amelsvoort Therese EditorMaastricht University, NETHERLANDSCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: YS. Data curation: MR YS. Formal analysis: MK MR. Funding acquisition: DHMB YS. Investigation: YS DHMB. Methodology: MK. Project administration: DHMB. Supervision: ALVB DHMB. Validation: YS DHMB MK MR ALVB. Visualization: MR. Writing – original draft: YS. Writing – review & editing: YS DHMB MK MR ALVB. * E-mail: Y.Stikkelbroek@uu.nl29 8 2016 2016 11 8 e016106213 1 2016 29 7 2016 © 2016 Stikkelbroek et al2016Stikkelbroek et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Background Depression during adolescence is a serious mental health problem. Difficulties in regulating evoked emotions after stressful life events are considered to lead to depression. This study examined if depressive symptoms were mediated by various cognitive emotion regulation strategies after stressful life events, more specifically, the loss of a loved one, health threats or relational challenges. Methods We used a sample of 398 adolescents (Mage = 16.94, SD = 2.90), including 52 depressed outpatients, who all reported stressful life event(s). Path analyses in Mplus were used to test mediation, for the whole sample as well as separately for participants scoring high versus low on depression, using multigroup analyses. Results Health threats and relational challenging stressful life events were associated with depressive symptoms, while loss was not. More frequent use of maladaptive strategies was related to more depressive symptoms. More frequent use of adaptive strategies was related to less depressive symptoms. Specific life events were associated with specific emotion regulation strategies. The relationship between challenging, stressful life events and depressive symptoms in the whole group was mediated by maladaptive strategies (self-blame, catastrophizing and rumination). No mediation effect was found for adaptive strategies. Conclusion The association between relational challenging, stressful life events and depressive symptoms was mediated by maladaptive, cognitive emotion regulation strategies. Zonmw- the Dutch nonprofit Organisation for Health Research and Development,grant number 80-82435-98-10117Bodden Denise H. M. This work was supported by Zonmw- the Dutch nonprofit Organisation for Health Research and Development, grant number 80-82435-98-10117 (http://www.zonmw.nl/). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll data files are available from the DANS database (see URL below). (http://dx.doi.org/10.17026/dans-288-7dh6)Data Availability All data files are available from the DANS database (see URL below). (http://dx.doi.org/10.17026/dans-288-7dh6) ==== Body Introduction Depression during adolescence is a serious problem because of its high prevalence [1–3], considerable burden of disease [4], suicide risk [5], other comorbid psychiatric disorders [6] and the high risk of recurrence [5,7]. Although knowledge about the etiology of depression has increased in the last decade, it is still difficult to explain and predict who becomes depressed, because of the many factors involved [8]. A well-established predictor of depressive symptoms is the experience of a stressful life event [9], like a romantic break up. An intriguing question is what mechanisms are involved that may lead different types of stressful life events to contribute to the development of depressive symptoms and ultimately depression. Difficulty in emotion regulation is considered to contribute to depression [10–12]. Emotion regulation is seen as a potential mediator of depressive symptoms after stressful life events have occurred [13]. From a clinical viewpoint it is necessary to establish which factors actually mediate the occurrence of depressive symptoms after specific types of stressful life events, as such factors could then be addressed effectively in prevention and treatment efforts. Adolescence is a challenging developmental phase with many physical and psychological changes, which may generate stress. Stressors are recognized as very important in the etiology and maintenance of internalizing problems [14]. Three different types of stressful life events have been identified, namely loss, health threats and relational challenges [15,16]. Stressful life events are associated with a larger increase in depressive symptoms than other types of stress, such as academic stress [17–22]. In children, negative life events were found to be a significant predictor of depression, putting children at risk for future depressive episodes [23,24]. Stressful life events during childhood have repeatedly been found to be associated with an increased risk of developing mental disorders during adulthood [25–27]. Furthermore, stressors can be followed by depressive symptoms, and depressive symptoms in themselves can generate stressful life events, resulting in a reciprocal relationship [14,28,29]. Garnefski and colleagues [15] found no difference in type of stressful life event and the association with depressive symptoms in adolescents aged 14 to 18. For clinically depressed adolescents, however, it is unclear as yet whether this is also true. The dual process model of cognitive vulnerability to depression hypothesizes that associative thought processing (automatic processing) induces depressive symptoms when no correction occurs by explicitly reflective processing [30]. This is especially the case when the associative processing is negatively biased with thoughts about oneself. Furthermore, life stress appears to deplete cognitive resources, which are necessary for reflective processes to correct associative processing [30]. This dual process of self-referent association and cognitive reflection is considered to be of importance for the regulation of emotions. Difficulty in regulating evoked emotions is often thought to lead to depression [11,12,31]. In adults, emotion regulation plays a central role in the etiology and maintenance of clinical levels of psychopathology [32–34]. Emotion regulation is a complex process with (un-)conscious, cognitive, and self-regulatory components. Emotion regulation strategies can be adaptive or maladaptive. Maladaptive emotion regulation has repeatedly been linked to various mental disorders, including the onset of depressive symptoms [35–39]. Maladaptive emotion regulation is also a risk factor for the recurrence of depression in adults [40]. Furthermore, findings suggest that the strength of the relationship between maladaptive emotion regulation strategies and psychopathology may be a function of clinical severity [41]. Some specific maladaptive cognitive emotion regulation strategies such as self-blame, rumination and catastrophizing, are associated with higher levels of depressive symptoms in adolescence, while adaptive strategies such as positive reappraisal, positive refocusing and putting things into a broader perspective, are associated with less depressive symptoms [15,42]. Only a few studies have established the relationship between stressful life events and emotion (dys)regulation in adults [10] and adolescents [13]. The type of life event influences the use of specific cognitive emotion regulation strategies, with health threat being associated with self-blame and relational challenge with other-blame in adolescence [15]. As pointed out, stressful life events have an impact on emotion regulation, which in turn influences the degree of depressive symptoms. Therefore, emotion regulation can be considered a mediator in the relationship between stressful life events and depressive symptoms [43,44]. Studies on emotion regulation as a mediator of depressive symptoms in adolescents are scarce. One study reported that emotion regulation strategies were found to mediate the relationship between interpersonal stress and depressive symptoms in undergraduate students [13]. It is unclear if this mediation is also present when adolescents have high levels of depressive symptoms, or whether the use of specific emotion regulation strategies mediates the relation between certain stressful life events and depressive symptoms. The aim of this study is to examine if the relation between stressful life events and depressive symptoms is mediated by cognitive emotion regulation strategies and whether these potentially mediating effects differ per type of stressful life event (Fig 1). We used a multi-sample approach, including a community sample and clinically referred depressed outpatients. Mediation was studied for the whole sample, as well as for the depressed and non-depressed adolescents separately. We expected to find that [1] the relationship between stressful life events (loss, health threat or relational challenge) and depressive symptoms is mediated by cognitive emotion regulation strategies; [2] the model for mediation would be specific for type of stressful life event: self-blame would be important after health threat and other-blame after relational stress, and no mediation would be present after loss; [3] the model would also be specific in that the pathway from stressful life events via cognitive emotion regulation to depressive symptoms would be more pronounced in adolescents with high levels of depressive symptoms compared to adolescents with low levels of depressive symptoms. 10.1371/journal.pone.0161062.g001Fig 1 Mediation model being tested; cognitive emotion regulation strategies as mediator of depressive symptoms after a stressful life event. Methods Participants In a total of 653 adolescents (of whom 84 were outpatients), those who rated no life event as stressful (n = 35) were excluded. Adolescents in the community sample (N = 569) receiving psychological treatment or social counseling (n = 40) were also excluded because of a potential effect on emotion regulation strategies. If answers on gender (n = 3), age (n = 4), life events or scores for whole scales were missing, cases were also excluded from the analyses (n = 173), which amounted to 39.2% from the community and 38.1% from the outpatient sample. Attrition analyses showed that the excluded adolescents (n = 255, Mage = 15.85, 31.8% boys) did not significantly differ from the included adolescents on the relevant scales, namely age, gender, depressive symptoms, life events and cognitive emotion regulation strategies. The result was a final sample of 398 adolescents aged 11 to 22 years (68.8% girls, 92.5% Dutch) including 52 (13%) clinical depressed outpatients. This sample was divided into two groups with low or high scores on depression, based on the official Dutch cut-off point (<12 vs. ≥12) for clinical depression on the Children’s Depression Inventory 2 (CDI-2). These groups will be referred to as non-depressed and depressed. The non-depressed group consisted of 289 adolescents (66.8% girls, Mage = 16.94, 97.2% enrolled in education), whereas the depressed group consisted of 109 adolescents (74.3% girls, Mage = 17.09, 49.5% enrolled in education). Data collection Data collection of patients was approved by the independent Medical Ethics Committee (METC) of the Utrecht Medical Centre at Utrecht University. Two samples were used in this study. The first sample consisted of 569 adolescents (age range 11–21) from the general population who were recruited by Master's degree students in schools and sports associations across The Netherlands. The adolescents were asked to participate in the research. After written informed consent forms were obtained from the participants and their parents, a self-report questionnaire was completed. The second sample participated in an effectiveness trial comparing Cognitive Behavioral Therapy to usual care (for more information, see [45]). 84 Clinically referred adolescents (age range 12–22) suffering from depression were recruited from 14 mental health care institutions across the Netherlands between 2011 and 2014. A psychologist informed adolescents and parents about the study and when both gave written informed consent, self-report questionnaires were completed pre-treatment using online or paper-and-pencil questionnaires. Measures Depressive symptoms The degree of depressive symptoms was measured with the Child Depression Inventory-2 [46,47], a revision of the CDI [48,49]. The CDI-2 is a self-report questionnaire for children (7 to 17 years) that reflects affective, behavioral, and cognitive symptoms of depression. Each of the 28 items offers 3 assertions: non-depressed (score 0, e.g., “I am sad once in a while”); mildly depressed (score 1, e.g., “I am sad many times.”); and clearly depressed (score 2, e.g., “I am sad all the time.”). The participants had to choose the assertion that applied most during the past two weeks. Total scores could range from 0 to 56, and a score of 12 or above (based on the CDI) is considered a clinically relevant score [49]. In this sample, 27.39% (n = 109) had a clinically relevant score. Reliability was very good for both the official target population of adolescents aged up to 17 years (n = 213, α = .93), and for adolescents older than 17 (n = 185, α = .91). Cognitive emotion regulation strategies Cognitive emotion regulation strategies were investigated with the Cognitive Emotion Regulation Questionnaire [50]. The CERQ consists of 36 items, reflecting 9 conceptually distinct adaptive or maladaptive strategies. Items refer to what someone thinks in response to a life event. The four maladaptive subscales are: Self-blame (thoughts of putting the blame of what you have experienced on yourself), Other-blame (thoughts of putting the blame of what you have experienced on others), Catastrophizing (thoughts of explicitly emphasizing the terror of an experience) and Rumination (thinking about the feelings and thoughts associated with the negative event). The five adaptive subscales are: Putting into perspective (thoughts of playing down the seriousness of the event or emphasizing the relativity when comparing it to other events); Positive refocusing (thinking about joyful and pleasant issues instead of thinking about the actual event); Positive reappraisal (thoughts of attaching a positive meaning to the event in terms of personal growth); Acceptance (thoughts of accepting what you have experienced and resigning to yourself what has happened); and Refocus on planning (thinking about what steps to take and how to handle the negative event). Each subscale contains 4 items measured on a 5-point Likert scale ranging from 1 (almost never) to 5 (almost always), with a higher score indicating more use of the specific adaptive strategy. Because extreme scores on Acceptance could be maladaptive the score was transformed into a categorical variable high (score above 14), low (score less than 10) and medium [36]. Research on the CERQ subscales indicated that internal consistencies were good, ranging from .67 to .81, and good validity [51]. In the present study, alphas ranged from good .70 to very good .82. Stressful life events For this study, we constructed the Life Event Scale [52], a 23 item self-report questionnaire about three types of life events, based on the distinction of psychological stress made by Lazarus [16]; Loss (1 item: death of a loved one including pets), Health threat (8 items: serious (mental) illness, suicide attempt, sexual abuse, psychological abuse, alcohol or drug abuse, crime and accidents concerning the self (not for suicide attempt), parent, sibling or friends), and Relational (or situational) challenges (14 items: parental divorce, step-parents, moving, changing schools, romantic break-up, police contact (parent or self), redundancy (parent or self), pregnancy, school failure, being bullied, conflict with parents or friends, being expelled from school). Participants were asked if they had experienced the life event, yes or no. If yes, respondents were asked to rate how stressful the event was from not stressful (0) to very stressful (3). Only adolescents with at least a score of two were included. The amount of Health threats and Relational challenges were both summed into a single score, both items were highly positively skewed. Loss was dichotomous. Data analytic strategy Missing data were imputed using Relative Mean Substitution [53]. Descriptive statistics were calculated, and Pearson correlations were computed for all variables included in the models. Multivariate Analysis of Variance (MANOVA) was used to test for differences in depressive symptoms and cognitive emotion regulation strategies based on gender or on condition. To examine the correlations between the three stressful life event variables, the nine proposed mediators and the outcome measure of depression, we applied path analyses using the software package MPLUS 7 [54]. Models were tested first for the whole sample and thereafter separately for participants scoring high versus low on depression using multigroup analyses. Within the models, the correlation between the three stressful life event variables was taken into account, as well as the correlation between the proposed mediators. The comparative fit index (CFI, preferably .95 or higher), the root mean square error of approximation (RMSEA, preferably .08 or lower), and the standardized root mean square residual (SRMR, preferably .09 or lower) served as model fit indices [55]. To examine the hypothesized mediation of cognitive emotion regulation strategies in the association between the type of stressful life events and depression, we used the Model Indirect approach using MPLUS 7 with a bootstrap procedure to ensure the accuracy and robustness of the analyses and to estimate Type I errors. To assess the possible moderating effect of the level of depression, multi-group analyses were conducted within MPLUS 7. This was done by testing whether the model fit (Δχ²) was significantly better for the model in which the paths of interest were allowed to differ between non-depressed and depressed, compared to the model in which the paths of interest were constrained to be equal [56]. Next, differences between both groups for the relations between model variables were tested per direct path, also using the chi-square difference test. This was done by constraining each path of interest separately while all other paths were unconstrained, and comparing this model to the model in which the path of interest, as well as all other paths, was unconstrained. To test the differences in the indirect effects between both groups, the MODEL TEST command was used. This command permitted testing of linear restrictions on the parameters using the Wald chi-square test [54]. Results The percentage of adolescents with high scores (≥12) on depressive symptoms was 27.39%. The percentages of life events reported by the total sample (N = 398) adolescents were 24.87% for loss; 33.17% for health threats, and 60.55% for relational challenges. A MANOVA on the total sample showed gender differences in cognitive emotion regulation strategies (F(9,388) = 2.754, p = .004), specifically girls scored lower on Other blame than boys (F(1,396) = 6.642, p = .010) and girls scored higher on Rumination than boys (F(1,396) = 8.548, p = .004). This analysis was repeated for both subgroups and also showed gender differences in the non-depressed group (F(9,279) = 2.441, p = .011), specifically girls scored lower on Other blame than boys (F(1,287) = 10.872, p = .001). An ANOVA showed no gender differences in depressive symptoms in the total group, nor in the subgroups. A MANOVA using both the total sample and the subgroups showed no differences in cognitive emotion regulation strategies between adolescents who did or did not experience loss. An ANOVA using both the total sample and the subgroups showed no differences in depressive symptoms between adolescents who did or did not experience loss. Another MANOVA showed group differences in the three types of stressful life events, which occurred more often in the depressed subgroup (F(3,394) = 24.410, p < .001), specifically for health threats (F(1,396) = 12.419, p < .001) and relational challenges (F(1,396) = 72.475, p < .001). Associations among variables Correlations, means and standard deviations for the total sample and all model variables are reported in Table 1, and separately for depressed and non-depressed adolescents in Table 2. 10.1371/journal.pone.0161062.t001Table 1 Correlations Total Group (N = 398).     1 2 3 4 5 6 7 8 9 10 11 12 13 M SD 1 Depressive Symptoms r .505 .120 .430 .319 -.275 -.304 -.413 .007 -.278 .049 .254 .454 9.608 8.854 p .000 .016 .000 .000 .000 .000 .000 .888 .000 .326 .000 .000 2 Self-blame r .222 .433 .497 .119 -.086 -.004 .296 .166 .042 .176 .360 9.475 3.350 p .000 .000 .000 .017 .085 .929 .000 .001 .404 .000 .000 3 Other-blame r .442 .168 .068 .106 .048 .171 .124 -.128 .063 .128 6.771 2.806 p .000 .001 .174 .034 .342 .001 .013 .010 .207 .011 4 Catastrophizing r .432 -.177 -.097 -.174 .066 -.043 -.019 .106 .276 6.739 2.726 p .000 .000 .053 .000 .190 .390 .709 .035 .000 5 Rumination r .056 -.028 .161 .241 .323 .145 .112 -.141 10.477 3.782 p .262 .574 .001 .000 .000 .004 .026 .005 6 Putting into Perspective r .494 .641 .452 .480 .003 -.060 -.141 12.269 3.927 p .000 .000 .000 .000 .949 .229 .005 7 Positive Refocusing r .545 .330 .425 -.032 -.065 -.109 11.912 3.870 p .000 .000 .000 .522 .198 .030 8 Positive Reappraisal r .400 .715 -.023 -.043 -.076 12.779 3.890 p .000 .000 .648 .395 .132 9 Acceptance r .369 .021 .048 .131 1.015 .737 p .000 .674 .345 .009 10 Refocus on Planning r -.053 -.039 -.011 12.799 3.612 p .296 .441 .827 11 Age r .121 .143 16.937 2.895 p .016 .004 12 Stressful Health Threats r .434 .538 .919 p .000 13 Stressful Relational Challenges r 1.294 1.491   p                               Significant results are printed in bold. 10.1371/journal.pone.0161062.t002Table 2 Correlations Depressed Group (N = 109) and Non-depressed Group (N = 289).     1 2 3 4 5 6 7 8 9 10 11 12 13 M SD 1 Depressive Symptoms r .372 -.079 .210 .212 -.218 -.279 -.340 .061 -.244 -.059 .368 .240 21.743 7.721 p .000 .416 .028 .027 .023 .003 .000 .530 .011 .545 .000 .012 2 Self-blame r .181 .095 .384 .506 .046 -.155 -.140 .158 -.017 .062 .309 .271 11.881 3.656 p .002 .328 .000 .000 .633 .107 .146 .100 .863 .520 .001 .004 3 Other-blame r .078 .228 .489 .201 .031 .122 .110 .108 .136 -.110 -.160 .- 038 7.440 3.512 p .186 .000 .000 .036 .749 .207 .256 .264 .158 .254 .096 .696 4 Catastrophizing r .172 .257 .383 .464 -.326 -.202 -.272 -.006 -.157 -.073 .009 .156 8.513 3.219 p .003 .000 .000 .000 .001 .035 .004 .950 .103 .452 .927 .106 5 Rumination r .070 .389 .099 .303 -.050 -.124 -.013 -.038 .114 .055 .050 .155 12.330 3.687 p .237 .000 .092 .000 .607 .200 .897 .695 .238 .572 .606 .107 6 Putting into Perspective r -.044 .382 .147 .048 .220 .489 .650 .470 .465 .161 .025 -.145 10.651 3.857 p .458 .000 .012 .413 .000 .000 .000 .000 .000 .094 .793 .133 7 Positive Refocusing r -.063 .141 .167 .141 .133 .445 .572 .256 .468 -.055 .002 -.115 10.275 4.098 p .282 .017 .005 .016 .024 .000 .000 .007 .000 .572 .982 .234 8 Positive Reappraisal r -.220 .343 .101 .092 .421 .593 .470 .395 .713 .113 -.056 -.048 10.661 3.945 p .000 .000 .086 .118 .000 .000 .000 .000 .000 .240 .564 .620 9 Acceptance r -.064 .412 .200 .109 .358 .470 .382 .443 .289 .017 .091 .114 1.028 .726 p .276 .000 .001 .064 .000 .000 .000 .000 .002 .859 .348 .236 10 Refocus on Planning r -.178 .454 .173 .158 .528 .445 .361 .695 .413 .136 -.092 -.078 11.551 3.463 p .002 .000 .003 .007 .000 .000 .000 .000 .000 .159 .340 .418 11 Age r -.030 .172 -.031 -.162 .270 .166 -.073 .228 .198 .195 .092 .090 16.530 2.519 p .616 .003 .601 .006 .000 .005 .219 .000 .001 .001 .342 .352 12 Stressful Health Threats r .024 -.014 .154 .062 .070 -.038 -.033 .056 .026 .041 .161 .435 .798 1.070 p .688 .815 .009 .297 .237 .520 .581 .347 .658 .493 .006 .000 13 Stressful Relational Challenges r .287 .195 .150 .129 .290 .004 .057 .130 .153 .163 .245 .384 2.248 1.701 p .000 .001 .011 .028 .000 .942 .337 .027 .009 .005 .000 .000 M 5.031 8.568 6.519 6.069 9.779 12.879 12.529 13.578 1.010 13.270 17.090 .439 .934   SD   2.990 2.728 2.626 2.172 3.581 3.784 3.599 3.561 .743 3.560 3.015 .836 1.227     The depressed sample is displayed in the top half of the table and the non-depressed sample is displayed in the bottom half; significant results are printed in bold. Within the total sample, age did not correlate significantly with depressive symptoms. Health threatening and relational challenging stressful life events showed weak to moderate correlations with depressive symptoms. Except for acceptance, all cognitive emotion regulation strategies, correlated weakly to strongly with the number of depressive symptoms. Weak to moderate correlations were found for relational challenging stressful life events and seven out of nine cognitive emotion regulation strategies; self-blame, other-blame, catastrophizing, rumination, putting into perspective, positive refocusing and acceptance. Within the non-depressed group, age did not correlate significantly with depressive symptoms. Relational challenging stressful life events were weakly correlated with more depressive symptoms. Weak correlations between relational challenging stressful life events and several strategies were found, namely; self-blame, other-blame, catastrophizing, rumination, positive reappraisal, acceptance and refocus on planning. Of these strategies, only self-blame, catastrophizing, positive reappraisal and refocus on planning in turn correlated weakly with depressive symptoms. Stressful health threatening life events correlated weakly with other-blame, which did not significantly correlate with depressive symptoms. In the depressed group, age did not correlate significantly with depressive symptoms. Relational challenging and also health threatening stressful life events showed a weak to moderate correlation with the degree of depressive symptoms, as well as with self-blame. All strategies, except for other-blame and acceptance, were weak to moderately correlated with depression. Path analyses The model fit indices for the mediation model within the whole sample were satisfactory (CFI = .95, RMSEA = .09, SRMR = .05). Multi-group analysis was used to test differences in depression level. The Chi-squared difference test indicated that the model differed for non-depressed and depressed participants (Δχ² (80) = 158.45, p < .001). The results of the model are described for participants with high and low depression scores separately below (see Fig 2 for the model and Fig 3 for the indirect effects). Gender was included in the model, but showed no direct significant effects on depressive symptoms or on any cognitive emotion regulation strategy. 10.1371/journal.pone.0161062.g002Fig 2 Direct Effects for the Total Sample. Standardized estimates of the direct effects on the cognitive emotion regulation and depressive symptoms. Only significant effects (* p < .05, ** p < .01, *** p < .001) within the total sample (N = 398) are shown. 10.1371/journal.pone.0161062.g003Fig 3 Mediation Model. Standardized estimates of the indirect effects on the depressive symptoms (N = 398). Only significant effects are shown (* p < .05, *** p < .001). Estimates apply to the total sample (N = 398). Direct associations between type of life events and depression Within the total sample and within the non-depressed subgroup, only relational challenging stressful life events were associated with depressive symptoms (β = .231, p < .001 and β = 338, p < .001, respectively). The depressed group showed a significant association of health threatening stressful life events with depressive symptoms (β = .285, p = .003). This association was also significantly different compared to the non-depressed group (Δχ² (1) = 11.55, p < .001). Loss was not associated with depressive symptoms in the total sample, or in either subgroup. Direct associations between type of life events and cognitive emotion regulation strategies Within the total sample, relational challenging stressful life events were associated with various strategies, namely self-blame, other-blame, catastrophizing, rumination, putting into perspective and acceptance (Fig 2). Within the non-depressed group, direct associations between relational challenging stressful life events and other-blame (β = .122, p = .039), positive reappraisal (β = .130, p = .034) and refocus on planning (β = .168, p = .004). Within the depressed group, direct associations between relational challenging stressful life events and self-blame (β = .195, p = .049), catastrophizing (β = .209, p = .041) and rumination (β = .240, p = .011) were found. Health threatening stressful life events were not significantly associated with any cognitive emotion regulation strategy within the whole sample, nor within the subgroups. Loss was negatively associated with catastrophizing within the non-depressed group (β = -.136, p = .015). Direct associations between cognitive emotion regulation strategies and depression Within the whole sample more depressive symptoms were significantly associated with: self-blame, catastrophizing, rumination, positive reappraisal and refocus on planning (Fig 2). In the non-depressed group, self-blame (β = .252, p < .001), less positive reappraisal (β = -.269, p = .003) and less refocusing on planning (β = -.245, p = .004) were significantly associated with depressive symptoms. In the depressed group, no significant associations between degree of depressive symptoms and use of any cognitive emotion regulation strategy was found. The correlations between strategies and depressive symptoms did not differ significantly between depressed and non-depressed subgroup. Acceptance was not significantly associated with the degree of depressive symptoms within either subgroup. Indirect associations Within the whole sample, significant indirect paths from relational challenging stressful life events to depressive symptoms were found via self-blame, catastrophizing and rumination (Fig 3). Indirect paths were found in the non-depressed group, from relational challenging stressful life events to depressive symptoms via self-blame and refocusing on planning. Within the depressed group, no indirect paths were identified. However, the Wald test of parameter constraints showed no significant differences between the non-depressed and depressed group for these indirect paths. Discussion The findings of this study support the general hypothesis that certain stressful life events were related to the level of depressive symptoms and that this relationship was mediated by maladaptive cognitive emotion regulation strategies. These findings are important for clinical practice to increase understanding of the association between specific types of stressful life events and the use of cognitive emotion regulation strategies. In addition, the results confirm the dual process model on associative and reflective processing in adolescents and depression [30]. The important role of self-referent association, maladaptive cognitive emotion regulation strategy namely self-blame, as a mediator of depressive symptoms was confirmed. The main results can be summarized as follows: (I) Health threatening and relational challenging stressful life events were associated with depressive symptoms, while stressful loss related life events were not; (II) more frequent use of maladaptive cognitive emotion regulation strategies was related to more depressive symptoms; (III) more frequent use of adaptive cognitive emotion regulation strategies was related to less depressive symptoms; (IV) specific life events were associated with specific emotion regulation strategies; (V) only the association between relational challenging stressful life events and depressive symptoms was mediated by maladaptive cognitive emotion regulation strategies (self-blame, catastrophizing and rumination); and (VI) adaptive cognitive emotion regulation strategies were not identified as mediators in the total sample. These results deserve some further interpretation and reflection. The relationship between cognitive emotion regulation strategies and depressive symptoms was established in our study. More use of maladaptive (self-blame, catastrophizing, rumination) and less use of adaptive cognitive emotion regulation adaptive strategies (positive reappraisal, refocus on planning) were significantly associated with more depressive symptoms in the whole sample. This finding is in line with earlier findings that maladaptive emotion regulation was linked to the onset of depressive symptoms [35–39]. Maladaptive cognitive emotion regulation could representative of the self-referent association of the dual process model and adaptive cognitive reflection is considered to be of importance for the regulation of emotions. Deficient emotion regulation is also a risk factor for recurrence of depression [40]. Our findings suggest that using maladaptive strategies can be more harmful, than the absence of using adaptive strategies. These findings are relevant for clinical practice to enhance prevention and treatment in order to detect and address specific mechanisms at work in depression. One specific and adaptive cognitive emotion regulation strategy, acceptance, was not significantly associated with depressive symptoms, as was also reported in a meta-analysis band colleagues [41]. Increase of acceptance of a problem or risk is a common objective in various treatments such as Mindfulness, and Acceptance and Commitment Therapy [57]. In acceptance-based treatments, acceptance is promoted in order to reduce experiential avoidance [58,59]. However, the role of acceptance might differ during the process of handling stressful life events and therefore it might also have been disguised in our study. A high score on acceptance immediately after a stressful life event could, for instance, be related to learned helplessness. Timing in relation to the occurrence of stressful life events should be taken into account in future research on acceptance. In addition, acceptance may be more of an end state that is based upon other regulation strategies, instead of reflecting an active and dynamic cognitive emotion regulation strategy. Stressful loss was not found to be associated with higher levels of depressive symptoms. This finding that the death of a loved one was generally not associated with elevated levels of depressive symptoms should be interpreted with caution. Earlier studies found that a substantial number of 75% to 80% of children do not develop mental health problems after the death of a parent or sibling, but these studies also found a significant increase in internalizing symptoms in these cases [60–64]. Another study found a significant increase in depression in the second year after bereavement [65]. In our study however, lapse of time after loss could not be accounted for. In a community sample, the comparison between family bereaved and non-bereaved showed a robust difference in internalizing problems by the age of 19 [24]. No long-term effects could be assessed in the current study. Furthermore, a broad definition of loss was used, which included pets. The kind of loss might also be of importance for the impact on mental health. Health threatening stressful life events were only associated with depressive symptoms in the depressed group and not in the non-depressed group. Although the level of depressive symptoms was high in the depressed adolescents, health threats still accounted for substantially more depressive symptoms. These findings are in line with a review on depressive symptoms in epileptic youth, showing an elevated risk for depression in this specific group with health problems [66]. So it seems that health threats are particularly important for depressive symptoms in adolescents. An association between relational challenging stressful life events and depressive symptoms was established for the total sample, and seen in both the depressed as well as the non-depressed group. This association was not found in earlier research conducted with secondary school students [67]. This discrepancy can be explained by the use of a larger multi-group sample in this study with higher levels of depressive symptoms, which made detection of the association possible. This shows that the use of multiple samples is indeed important in future research on cognitive emotion regulation [41]. Our findings confirm the existence of a specific association between relational challenges and cognitive emotion regulation. Loss or health threatening stressful life events showed no specific association with any of the cognitive emotion regulation strategies within the whole group, suggesting that the type of stressful event influenced the use of specific cognitive emotion regulation. Relational challenging stressful life events were associated with maladaptive strategies as well as with two adaptive strategies: putting experiences into perspective and acceptance. The non-depressed group showed a significant association between relational challenging stressful life events and three adaptive strategies, namely positive reappraisal, refocus on planning, and acceptance. In the depressed group, this association was not found, suggesting that non-depressed and depressed adolescents differ in their use of maladaptive and adaptive cognitive emotion regulation strategies. However, no significant difference between groups in the strength of this association was found. These results must thus be interpreted with caution, and studies with larger groups are needed to rule out power issues in interpreting these differences. The mediating role of maladaptive cognitive emotion regulation strategies was established, Self-blame, catastrophizing and rumination could be identified as mediators between stressful relational challenging life events and depressive symptoms. However, depressed and non-depressed adolescents did not differ significantly in these mediation relationships, which may be due to the size of the depressed sample. Still, this is an important finding, which could be useful for clinical practice. Experiencing relational challenging stressful life events and blaming oneself, emphasizing the terror of experiences or dwelling on feelings and thoughts about the events, may put adolescents at risk for depressive symptoms. Strengths and limitations This study is innovative for several reasons. First, according to the literature, testing mediational models on the etiology of adolescent depression was needed [14]. Second, the use of the multi-sample approach, including severely depressed adolescent patients, is scarce and should be used more often [68], as it is essential for the study of psychopathology [14,41]. Third, a rigorous criterion for life events was used, namely stressful life events reported by the adolescent as upsetting. Previous research used the number of life events as a variable, without the upsetting criterion, thereby ignoring whether the life events actually impact the life of participants in a negative fashion. Fourth, not only being upset but also the type of life event was taken into account as an important variable affecting depressive symptoms, as well as the cognitive emotion regulation strategies. Fifth, to our knowledge, we were the first to test the mediating role of cognitive emotion regulation in the association between stressful life events and depressive symptoms in adolescents and to test differences in mediation between low and high levels of depressive symptoms. This study also has several limitations. First, cross-sectional data were used and therefore no temporal conclusions could be drawn. Future research is needed to test mediational models in a longitudinal design. Second, the profile of different mediators within one individual could not be taken into account. For example, the mediation of depressive symptoms by rumination may be more prominent if the use of self-blame is high and positive refocusing is low. Even more useful for clinical practice would be the identification of patterns in how the various emotion regulation strategies are used [69]. Third, the relationship between associative (maladaptive strategies) and reflective processing (adaptive strategies) could not be taken into account, while this may be a function of clinical severity [41]. Further research on specific correlations between cognitive emotion regulation strategies is needed and of importance to determine which strategy should be addressed in interventions. Fourth, a lack of power could underlie the fact that no mediating paths were found in the depressed group, and that no significant differences in the strength of the mediational paths between the non-depressed and depressed adolescents were found. A larger sample than N = 109 is needed to identify possible mediators in the depressed group. Fifth, the time between the last stressful life event and the measurement of depressive symptoms was not taken into account. In a longitudinal design this variable should be included because depressive symptoms could increase instantly or gradually after some time, through use of specific cognitive regulation strategies [70]. Despite the mentioned limitations, our study contributes to current scientific knowledge by showing that depressive symptoms are mediated by maladaptive cognitive emotion regulation strategies (self-blame, rumination and catastrophizing) uniquely after stressful relational challenging life events. Mediation was not found after losing a loved one or experiencing a health threat. Adaptive cognitive emotion regulation strategies, for instance acceptance, were not identified as mediators. These findings are important for clinical practice. Use of specific maladaptive cognitive regulation strategies after relationally challenging stressful life events can aggravate depressive symptoms. To prevent depression after negative life events, maladaptive cognitive emotion regulation strategies should be reduced in adolescents. This research is part of the Adolescent Depression study of the University of Utrecht. We would like to thank Maria Kovacs and MHS for their generosity in letting us use the CDI-2. We are grateful for the contribution of the Mental Health Institutions and their professionals, as without them data collection would be impossible: Accare, Altrecht, Ambulatorium, Bascule, Curium, GGZ-centraal, Herlaarhof, Lentis, Orbis, Perspectief, Praktijk Appelboom, Traverse, and Triversum. We also thank the students for collecting data in the community sample. This research is funded by Zonmw- the Dutch Organisation for Health Research and Development, grant number 80-82435-98-10117. ==== Refs References 1 Cohen P , Cohen J , Kasen S , Velez CN , Hartmark C , Johnson J , et al An Epidemiological Study of Disorders in Late Childhood and Adolescence—I. Age‐and Gender‐Specific Prevalence . Journal of Child Psychology and Psychiatry 1993 ;34 (6 ):851 –867 . 8408371 2 Fergusson DM , Horwood LJ , Lynskey MT . Prevalence and comorbidity of DSM-III-R diagnoses in a birth cohort of 15 year olds . Journal of the American Academy of Child and Adolescent Psychiatry 1993 ;32 (6 ):1127 –1134 . 8282656 3 Hankin BL , Abramson LY , Moffitt TE , Silva PA , McGee R , Angell KE . Development of depression from preadolescence to young adulthood: emerging gender differences in a 10-year longitudinal study . J Abnorm Psychol 1998 ;107 (1 ):128 –140 . 9505045 4 Hoeymans N , Gommer A , Poos M editors. Welke verschillen zijn er tussen leeftijdsgroepen: Volksgezondheid Toekomst Verkenning, Nationaal Kompas Volksgezondheid Bilthoven : RIVM ; 2006 . 5 Ryan ND . Treatment of depression in children and adolescents . The Lancet 2005 ;366 (9489 ):933 –940 . 6 Herbert J , Goodyer I , Altham P , Pearson J , Secher S , Shiers H . Adrenal secretion and major depression in 8-to 16-year-olds, II. Influence of co-morbidity at presentation . Psychol Med 1996 ;26 (2 ):257 –263 . 8685282 7 Curry J , Silva S , Rohde P , Ginsburg G , Kratochvil C , Simons A , et al Recovery and recurrence following treatment for adolescent major depression . Arch Gen Psychiatry 2011 ;68 (3 ):263 –269 . 10.1001/archgenpsychiatry.2010.150 21041606 8 Hankin BL . Future directions in vulnerability to depression among youth: Integrating risk factors and processes across multiple levels of analysis . Journal of Clinical Child and Adolescent Psychology 2012 ;41 (5 ):695 –718 . 10.1080/15374416.2012.711708 22900513 9 Kendler KS , Karkowski LM , Prescott CA . Causal relationship between stressful life events and the onset of major depression . Am J Psychiatry 1999 ;156 (6 ):837 –841 . 10360120 10 Abravanel BT , Sinha R . Emotion dysregulation mediates the relationship between lifetime cumulative adversity and depressive symptomatology . J Psychiatr Res 2014 ;61 :89 –96 . 10.1016/j.jpsychires.2014.11.012 25528603 11 Campbell-Sills L , Barlow DH . Incorporating emotion regulation into conceptualizations and treatments of anxiety and mood disorders In: Gross JJ , editor. Handbook of emotion regulation New York : Guilford Press ; 2007 p. 542 –559 . 12 Mennin DS , Holaway RM , Fresco DM , Moore MT , Heimberg RG . Delineating components of emotion and its dysregulation in anxiety and mood psychopathology . Behavior Therapy 2007 ;38 (3 ):284 –302 . 17697853 13 Moriya J , Takahashi Y . Depression and interpersonal stress: The mediating role of emotion regulation . Motiv Emotion 2013 ;37 (3 ):600 –608 . 14 Grant KE , Compas BE , Thurm AE , McMahon SD , Gipson PY , Campbell AJ , et al Stressors and child and adolescent psychopathology: Evidence of moderating and mediating effects . Clin Psychol Rev 2006 ;26 (3 ):257 –283 . 16364522 15 Garnefski N , Boon S , Kraaij V . Relationships between cognitive strategies of adolescents and depressive symptomatology across different types of life event . Journal of Youth and Adolescence 2003 ;32 (6 ):401 –408 . 16 Lazarus RS . Stress and emotion: A new synthesis New York : Springer Publishing Company ; 2006 . 17 Charbonneau AM , Mezulis AH , Hyde JS . Stress and emotional reactivity as explanations for gender differences in adolescents’ depressive symptoms . Journal of youth and adolescence 2009 ;38 (8 ):1050 –1058 . 10.1007/s10964-009-9398-8 19636770 18 Flynn M , Rudolph KD . The contribution of deficits in emotional clarity to stress responses and depression . Journal of applied developmental psychology 2010 ;31 (4 ):291 –297 . 22021945 19 Flynn M , Kecmanovic J , Alloy LB . An examination of integrated cognitive-interpersonal vulnerability to depression: The role of rumination, perceived social support, and interpersonal stress generation . Cognitive Therapy and Research 2010 ;34 (5 ):456 –466 . 25429169 20 Hammen C . Stress generation in depression: Reflections on origins, research, and future directions . J Clin Psychol 2006 ;62 (9 ):1065 –1082 . 16810666 21 Hankin BL , Mermelstein R , Roesch L . Sex differences in adolescent depression: Stress exposure and reactivity models . Child Dev 2007 ;78 (1 ):279 –295 . 17328705 22 Rudolph KD . Developmental influences on interpersonal stress generation in depressed youth . J Abnorm Psychol 2008 ;117 (3 ):673 –679 . 10.1037/0021-843X.117.3.673 18729618 23 Nolen-Hoeksema S , Girgus JS , Seligman ME . Predictors and consequences of childhood depressive symptoms: a 5-year longitudinal study . J Abnorm Psychol 1992 ;101 (3 ]:405 –422 . 1500598 24 Stikkelbroek Y , Bodden DH , Reitz E , Vollebergh WA , van Baar AL . Mental health of adolescents before and after the death of a parent or sibling . Eur Child Adolesc Psychiatry 2015 :1 –11 . 25 Collishaw S , Pickles A , Messer J , Rutter M , Shearer C , Maughan B . Resilience to adult psychopathology following childhood maltreatment: Evidence from a community sample . Child Abuse Negl 2007 ;31 (3 ):211 –229 . 17399786 26 Comijs HC , Beekman AT , Smit F , Bremmer M , Tilburg Tv , Deeg DJ . Childhood adversity, recent life events and depression in late life . J Affect Disord 2007 ;103 (1 ):243 –246 .17291592 27 Cuijpers P , Smit F , Unger F , Stikkelbroek Y , ten Have M , de Graaf R . The disease burden of childhood adversities in adults: a population-based study . Child Abuse Negl 2011 ;35 (11 ):937 –945 . 10.1016/j.chiabu.2011.06.005 22099144 28 A multi-wave longitudinal study of the transactional relations between depressive symptoms, hopelessness, and stress in high-risk youth. Paper accepted for presentation at the annual convention of the Canadian Association for Child and Adolescent Psychiatry. Depression From Childhood Through Adolescence and Adulthood; 2004. 29 Waaktaar T , Borge AIH , Fundingsrud HP , Christie HJ , Torgersen S . The role of stressful life events in the development of depressive symptoms in adolescence—A longitudinal community study . J Adolesc 2004 ;27 (2 ):153 –163 . 15023515 30 Beevers CG . Cognitive vulnerability to depression: A dual process model . Clin Psychol Rev 2005 ;25 (7 ):975 –1002 . 15905008 31 Gross JJ , Muñoz RF . Emotion regulation and mental health . Clinical psychology: Science and practice 1995 ;2 (2 ):151 –164 . 32 Berenbaum H , Raghavan C , Le H , Vernon LL , Gomez JJ . A taxonomy of emotional disturbances . Clinical Psychology: Science and Practice 2003 ;10 (2 ):206 –226 . 33 Greenberg LS . Emotion-focused therapy: Coaching clients to work through their feelings Washington DC : American Psychological Association ; 2002 . 34 Mennin D , Farach F . Emotion and evolving treatments for adult psychopathology . Clinical Psychology: Science and Practice 2007 ;14 (4 ):329 –352 . 35 Ehring T , Fischer S , Schnülle J , Bösterling A , Tuschen-Caffier B . Characteristics of emotion regulation in recovered depressed versus never depressed individuals . Personality and Individual Differences 2008 ;44 (7 ):1574 –1584 . 36 Garnefski N , Kraaij V . Relationships between cognitive emotion regulation strategies and depressive symptoms: A comparative study of five specific samples . Personality and Individual differences 2006 ;40 (8 ):1659 –1669 . 37 Joormann J , D'Avanzato C . Emotion regulation in depression: Examining the role of cognitive processes: Cognition & Emotion Lecture at the 2009 ISRE Meeting . Cognition and Emotion 2010 ;24 (6 ):913 –939 . 38 Martin RC , Dahlen ER . Cognitive emotion regulation in the prediction of depression, anxiety, stress, and anger . Personality and Individual Differences 2005 ;39 (7 ):1249 –1260 . 39 Schroevers M , Kraaij V , Garnefski N . Goal disturbance, cognitive coping strategies, and psychological adjustment to different types of stressful life event . Personality and Individual Differences 2007 ;43 (2 ):413 –423 . 40 Rude S , McCarthy C . Brief report . Cognition and Emotion 2003 ;17 (5 ):799 –806 . 41 Aldao A , Nolen-Hoeksema S , Schweizer S . Emotion-regulation strategies across psychopathology: A meta-analytic review . Clin Psychol Rev 2010 ;30 (2 ):217 –237 . 10.1016/j.cpr.2009.11.004 20015584 42 Kraaij V , Garnefski N , de Wilde EJ , Dijkstra A , Gebhardt W , Maes S , et al Negative life events and depressive symptoms in late adolescence: Bonding and cognitive coping as vulnerability factors? Journal of youth and adolescence 2003 ;32 (3 ):185 –193 . 43 Goodman KL , Southam-Gerow MA . The regulating role of negative emotions in children’s coping with peer rejection . Child Psychiatry & Human Development 2010 ;41 (5 ):515 –534 .20464475 44 Kelly BM , Schwartz D , Gorman AH , Nakamoto J . Violent victimization in the community and children’s subsequent peer rejection: the mediating role of emotion dysregulation . J Abnorm Child Psychol 2008 ;36 (2 ):175 –185 . 17846882 45 Stikkelbroek Y , Bodden DH , Dekovic M , van Baar AL . Effectiveness and cost effectiveness of cognitive behavioral therapy (CBT) in clinically depressed adolescents: individual CBT versus treatment as usual (TAU) . BMC Psychiatry 2013 11 21 ;13 :314-244X-13-314. 46 Kovacs M . The child depression inventory 2 (CDI-2) Toronto, Canada : Multi Health Systems ; 2011 . 47 Bodden DH , Stikkelbroek Y , Braet C . The child depression inventory 2 (CDI-2) Utrecht : Utrecht University ; 2010 . 48 Kovacs M . Children's depression inventory: Manual North Tonawanda : Multi-Health Systems ; 1992 . 49 Braet C , Timbremont B . Children’s depression inventory [Dutch Version] Lisse : Swets & Zeitlinger ; 2002 . 50 Garnefski N , Kraaij V , Spinhoven P . Handleiding voor het gebruik van de Cognitive Emotion Regulation Questionnaire [Manual for the use of the Cognitive Emotion Regulation Questionnaire] (CERQ) Leiderdorp : Datec ; 2002 . 51 Garnefski N , Kraaij V , Spinhoven P . Negative life events, cognitive emotion regulation and emotional problems . Personality and Individual differences 2001 ;30 (8 ):1311 –1327 . 52 Bodden DH , Stikkelbroek YAJ . Life Events Scale . Utrecht : Utrecht University ; 2010 . 53 Raaijmakers QA . Effectiveness of different missing data treatments in surveys with Likert-type data: Introducing the relative mean substitution approach . Educational and Psychological Measurement 1999 ;59 (5 ):725 –748 . 54 Muthén L , Muthén B . 1998–2010 Mplus User's Guide . 7th ed. Los Angeles, CA : Muthén & Muthén ; 2010 . 55 Iacobucci D . Structural equations modeling: Fit indices, sample size, and advanced topics . Journal of Consumer Psychology 2010 ;20 (1 ):90 –98 . 56 Lammers J , Kuntsche E , Engels RC , Wiers RW , Kleinjan M . Mediational relations of substance use risk profiles, alcohol-related outcomes, and drinking motives among young adolescents in the Netherlands . Drug Alcohol Depend 2013 ;133 (2 ):571 –579 . 10.1016/j.drugalcdep.2013.07.030 23998377 57 Hofmann SG , Asmundson GJ . Acceptance and mindfulness-based therapy: New wave or old hat? Clin Psychol Rev 2008 ;28 (1 ):1 –16 . 17904260 58 Eifert GH , Forsyth JP . Acceptance and Commitment Therapy for Anxiety Disorders: A Practitioner's Treatment Guide to Using Mindfulness, Acceptance, and Values-Based Behavior Change . Oakland, California : New Harbinger Publications ; 2005 . 59 Hayes SC , Strosahl KD , Wilson KG . Acceptance and commitment therapy: An experiential approach to behavior change New York : Guilford Press ; 1999 . 60 Cerel J , Fristad MA , Verducci J , Weller RA , Weller EB . Childhood bereavement: psychopathology in the 2 years postparental death . Journal of the American Academy of Child and Adolescent Psychiatry 2006 ;45 (6 ):681 –690 . 16721318 61 Dowdney L . Annotation: Childhood bereavement following parental death . Journal of Child Psychology and Psychiatry 2000 ;41 (7 ):819 –830 . 11079425 62 Dowdney L . Children bereaved by parent or sibling death . Psychiatry 2005 ;4 (9 ):118 –122 . 63 Luecken LJ , Roubinov DS . Pathways to lifespan health following childhood parental death . Social and personality psychology compass 2012 ;6 (3 ):243 –257 . 23555319 64 Worden JW , Silverman PR . Parental death and the adjustment of school-age children . OMEGA—Journal of Death and Dying 1996 ;33 (2 ):91 –102 . 65 Brent D , Melhem N , Donohoe MB , Walker M . The incidence and course of depression in bereaved youth 21 months after the loss of a parent to suicide, accident, or sudden natural death . Am J Psychiatry 2009 ;166 (7 ):786 –794 . 10.1176/appi.ajp.2009.08081244 19411367 66 Hankin BL . Adolescent depression: Description, causes, and interventions . Epilepsy and Behavior 2006 ;8 (1 ):102 –114 . 16356779 67 Garnefski N , Legerstee J , Kraaij V , van den Kommer T , Teerds J . Cognitive coping strategies and symptoms of depression and anxiety: A comparison between adolescents and adults . J Adolesc 2002 ;25 (6 ):603 –611 . 12490178 68 Southam-Gerow MA , Kendall PC . Emotion regulation and understanding: Implications for child psychopathology and therapy . Clin Psychol Rev 2002 ;22 (2 ):189 –222 . 11806019 69 Dixon-Gordon KL , Aldao A , De Los Reyes A . Repertoires of emotion regulation: A person-centered approach to assessing emotion regulation strategies and links to psychopathology . Cognition and Emotion 2014 (ahead-of-print):1 –12 . 70 Calvete E , Orue I , Hankin BL . Transactional relationships among cognitive vulnerabilities, stressors, and depressive symptoms in adolescence . J Abnorm Child Psychol 2013 ;41 (3 ):399 –410 . 10.1007/s10802-012-9691-y 23093441

PMC005xxxxxx/PMC5003337.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757120410.1371/journal.pone.0162039PONE-D-16-21207Research ArticleBiology and Life SciencesAnatomyHeadEyesMedicine and Health SciencesAnatomyHeadEyesBiology and Life SciencesAnatomyOcular SystemEyesMedicine and Health SciencesAnatomyOcular SystemEyesMedicine and Health SciencesDiagnostic MedicineDiagnostic RadiologyMagnetic Resonance ImagingResearch and Analysis MethodsImaging TechniquesDiagnostic RadiologyMagnetic Resonance ImagingMedicine and Health SciencesRadiology and ImagingDiagnostic RadiologyMagnetic Resonance ImagingResearch and Analysis MethodsBioassays and Physiological AnalysisElectrophysiological TechniquesBrain ElectrophysiologyElectroencephalographyEvent-Related PotentialsBiology and Life SciencesPhysiologyElectrophysiologyNeurophysiologyBrain ElectrophysiologyElectroencephalographyEvent-Related PotentialsMedicine and Health SciencesPhysiologyElectrophysiologyNeurophysiologyBrain ElectrophysiologyElectroencephalographyEvent-Related PotentialsBiology and Life SciencesNeuroscienceNeurophysiologyBrain ElectrophysiologyElectroencephalographyEvent-Related PotentialsBiology and Life SciencesNeuroscienceBrain MappingElectroencephalographyEvent-Related PotentialsMedicine and Health SciencesDiagnostic MedicineClinical NeurophysiologyElectroencephalographyEvent-Related PotentialsResearch and Analysis MethodsImaging TechniquesNeuroimagingElectroencephalographyEvent-Related PotentialsBiology and Life SciencesNeuroscienceNeuroimagingElectroencephalographyEvent-Related PotentialsResearch and Analysis MethodsBioassays and Physiological AnalysisElectrophysiological TechniquesBrain ElectrophysiologyElectroencephalographyBiology and Life SciencesPhysiologyElectrophysiologyNeurophysiologyBrain ElectrophysiologyElectroencephalographyMedicine and Health SciencesPhysiologyElectrophysiologyNeurophysiologyBrain ElectrophysiologyElectroencephalographyBiology and Life SciencesNeuroscienceNeurophysiologyBrain ElectrophysiologyElectroencephalographyBiology and Life SciencesNeuroscienceBrain MappingElectroencephalographyMedicine and Health SciencesDiagnostic MedicineClinical NeurophysiologyElectroencephalographyResearch and Analysis MethodsImaging TechniquesNeuroimagingElectroencephalographyBiology and Life SciencesNeuroscienceNeuroimagingElectroencephalographyResearch and Analysis MethodsImaging TechniquesNeuroimagingBiology and Life SciencesNeuroscienceNeuroimagingBiology and Life SciencesAnatomyBrainAmygdalaMedicine and Health SciencesAnatomyBrainAmygdalaResearch and Analysis MethodsImaging TechniquesNeuroimagingComputed Axial TomographyBiology and Life SciencesNeuroscienceNeuroimagingComputed Axial TomographyMedicine and Health SciencesDiagnostic MedicineDiagnostic RadiologyTomographyComputed Axial TomographyResearch and Analysis MethodsImaging TechniquesDiagnostic RadiologyTomographyComputed Axial TomographyMedicine and Health SciencesRadiology and ImagingDiagnostic RadiologyTomographyComputed Axial TomographyBiology and Life SciencesNeuroscienceBrain MappingFunctional Magnetic Resonance ImagingMedicine and Health SciencesDiagnostic MedicineDiagnostic RadiologyMagnetic Resonance ImagingFunctional Magnetic Resonance ImagingResearch and Analysis MethodsImaging TechniquesDiagnostic RadiologyMagnetic Resonance ImagingFunctional Magnetic Resonance ImagingMedicine and Health SciencesRadiology and ImagingDiagnostic RadiologyMagnetic Resonance ImagingFunctional Magnetic Resonance ImagingResearch and Analysis MethodsImaging TechniquesNeuroimagingFunctional Magnetic Resonance ImagingBiology and Life SciencesNeuroscienceNeuroimagingFunctional Magnetic Resonance ImagingGamma Oscillations in the Temporal Pole in Response to Eyes Temporal Pole Gamma Oscillations to EyesSato Wataru 1*Kochiyama Takanori 2Uono Shota 1Matsuda Kazumi 3Usui Keiko 3Usui Naotaka 3Inoue Yushi 3Toichi Motomi 41 Department of Neurodevelopmental Psychiatry, Habilitation and Rehabilitation, Graduate School of Medicine, Kyoto University, 53 Shogoin-Kawaharacho, Sakyo-ku, Kyoto, 606–8507, Japan2 Brain Activity Imaging Center, Advanced Telecommunications Research Institute International, 2-2-2 Hikaridai, Seika-cho, Soraku-gun, Kyoto, 619–0288, Japan3 National Epilepsy Center, Shizuoka Institute of Epilepsy and Neurological Disorders, Urushiyama 886, Shizuoka, 420–8688, Japan4 Faculty of Human Health Science, Graduate School of Medicine, Kyoto University, 53 Shogoin-Kawaharacho, Sakyo-ku, Kyoto, 606–8507, JapanPavlova Marina A. EditorUniversitatsklinikum Tubingen, GERMANYCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: WS TK YI MT. Data curation: WS SU KM KU NU YI. Formal analysis: WS TK KM. Methodology: WS TK KM KU NU YI. Writing – original draft: WS TK SU KM KU NU YI MT. Writing – review & editing: WS TK SU KM KU NU YI MT. * E-mail: sato.wataru.4v@kyoto-u.ac.jp29 8 2016 2016 11 8 e016203926 5 2016 16 8 2016 © 2016 Sato et al2016Sato et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.The eyes of an individual act as an indispensable communication medium during human social interactions. Functional neuroimaging studies have revealed that several brain regions are activated in response to eyes and eye gaze direction changes. However, it remains unclear whether the temporal pole is one of these regions. Furthermore, if the temporal pole is activated by these stimuli, the timing and manner in which it is activated also remain unclear. To investigate these issues, we analyzed intracranial electroencephalographic data from the temporal pole that were obtained during the presentation of eyes and mosaics in averted or straight directions and their directional changes. Time–frequency statistical parametric mapping analyses revealed that the bilateral temporal poles exhibited greater gamma-band activation beginning at 215 ms in response to eyes compared with mosaics, irrespective of the direction. Additionally, the right temporal pole showed greater gamma-band activation beginning at 197 ms in response to directional changes of the eyes compared with mosaics. These results suggest that gamma-band oscillations in the temporal pole were involved in the processing of the presence of eyes and changes in eye gaze direction at a relatively late temporal stage compared with the posterior cortices. The Benesse CorporationSato Wataru The JSPS Funding Program for Next Generation World-Leading ResearchersLZ008Sato Wataru This study was supported by funds from the Benesse Corporation and the JSPS Funding Program for Next Generation World-Leading Researchers (LZ008). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction It has been said that the eyes are windows to the soul [1] and thus are indispensable for human communication. Accordingly, the detection of eyes and the recognition of changes in eye gaze direction generate multiple psychological activities in the observer. For example, straight and averted eye directions trigger emotional reactions [2] and attention orienting [3], respectively, and both induce mind reading [4]. These processes play important roles in real-life social interactions [5] as well as contribute to impairments in social functioning in patients with clinical disorders [6]. Neuroscientific studies have explored the neural mechanisms underlying the processing of eye information and have identified several brain regions that are activated in response to eyes and eye gaze direction changes. Several neuroimaging studies using functional magnetic resonance imaging (fMRI) and positron emission tomography (PET) demonstrated that the posterior superior temporal sulcus (STS) and the adjacent middle and superior temporal gyri (STS region [7]) are active in response to the presence of eyes and changes in eye gaze direction, specifically in the case of averted gaze direction [8–16]. Intracranial electroencephalography (EEG) studies also reported that the STS region [17] and the inferior occipital gyrus (Sato et al., submitted) exhibit electrical activity in response to eye gaze direction changes and eyes, respectively. Additionally, neuroimaging studies revealed that other anterior regions, such as the amygdala, are also active in response to eyes and eye gaze direction changes [16,18]. Because these regions are also more active during the presentation of faces relative to non-face objects (for a review, see [19]), it is likely that the processing of eyes is accomplished as a subtype of processing for faces. These findings suggest that eye information is processed across a widespread neural network that includes the inferior occipital gyrus, STS region, and amygdala. However, it remains unclear whether there are changes in temporal pole activity in response to the presence of eyes and changes in eye gaze direction. Lesion studies in monkeys have revealed that the temporal pole plays an indispensable role during social interactions [20–23]. These findings are consistent with those of comparative anatomical studies indicating that the temporal pole exists only in primates [24,25], likely because primates are exceedingly social animals. Anatomical studies in monkeys revealed that the temporal pole receives highly processed visual signals from the posterior cortices, including the STS region, and communicates bidirectionally with the amygdala (for a review, see [26]). This suggests that the temporal pole may cooperate with these regions during the processing of eye information. Several previous neuroimaging studies in humans reported that the temporal pole is more activated in response to faces relative to other objects [27–30], suggesting that this region is involved in face processing. Although none of the aforementioned neuroimaging studies observed an activation of the temporal pole in response to eyes or eye gaze direction changes, this null finding might be accounted for by air-tissue inhomogeneity artifacts that can diminish fMRI signals in this region or by ill-defined anatomical categories (e.g., the periamygdaloid region) due to the lack of theoretical interest [31]. On the other hand, an fMRI study showed activation in the temporal pole by comparing brain activation during the observation of eyes in a mind reading task versus during the passive observation of crosshairs [32]. Another fMRI study investigated brain activation changes during the adaptation to eye gaze direction and reported reduced activity in the temporal pole in response to adapted gaze direction [33]. Based on these findings, we hypothesized that direct recording of electrical activity from the temporal pole could reveal its activation in response to the presence of eyes and changes in eye gaze direction. Furthermore, if the above hypothesis is correct, the timing and manner in which the temporal pole is activated during the processing of eye information need to be addressed. To understand the neural mechanisms, that is, the causal relationships among brain regions, the time course and frequency of brain activation should be evaluated [34,35]. To date, no electrophysiological studies, which can provide this type of information, have investigated activation in the temporal pole in response to the presence of eyes and changes in eye gaze direction. Indirect evidence from intracranial EEG recording and its event-related potential (ERP) analyses revealed temporal differences in the activation of the temporal pole and other visual areas during face processing [36]. This study showed that the temporal pole exhibits a higher ERP component that peaks at 300–400 ms in response to faces than to other objects, whereas other posterior cortices, including the inferior occipital gyrus and STS region, exhibit the ERP peaks at approximately 200 ms. Other intracranial EEG studies investigating the processing of eye information revealed that the inferior occipital gyrus (Sato et al., submitted) and STS region [17] exhibit ERPs that peak at approximately 200 ms. Collectively, these findings suggest that the temporal pole may exhibit ERPs related to eye processing during the 300–400 ms window that have a similar temporal profile to those exhibited during face processing. However, it must be noted that ERP analyses primarily detect the low-frequency components of EEG data [37]. For the comprehensive elucidation of high- and low-frequency neuronal activities at a high temporal resolution, time–frequency analyses must be conducted [38]. Previous intracranial EEG studies that performed time–frequency analyses revealed that the inferior occipital gyrus (Sato et al., submitted) and STS region [17] show gamma-band (higher than 30 Hz [39]) activation in response to eye information beginning at approximately 100 ms. These data suggest that, within the visual stream from the posterior to the anterior occipito–temporal cortices, the temporal pole may exhibit gamma-band activation during about 200–300 ms, which would be later than that observed in more posterior regions. Taking these findings together, we hypothesized that the temporal pole could show gamma-band and ERP activations beginning at approximately 200–300 and 300–400 ms, respectively, in response to the presence of eyes and changes in eye gaze direction. To test these hypotheses, we analyzed intracranial EEG data from the temporal poles of six participants undergoing presurgical assessments for epilepsy. To examine the effect of appearance of eyes, the participants were presented with visual stimuli comprising the eye region or mosaic patterns as control stimuli. To test the effect of eye gaze direction, averted and straight directions were prepared for both the eyes and mosaics. To explore the effect of changes in eye gaze direction, a second stimulus was presented 500 ms after the onset of the first stimulus with the direction of the second stimulus being different from that of the first. To examine automatic eye processing, the participants were asked to engage in dummy target detection. Time–frequency statistical parametric mapping (SPM) [40] and traditional ERP analyses were conducted. Because several lines of evidence indicate that there are functional hemispheric differences during the processing of eye information [41], the activities of the temporal poles in both hemispheres were analyzed and compared. Materials and Methods Ethics Statement This study was approved by the Ethics Committee of Shizuoka Institute of Epilepsy and Neurological Disorders, and was conducted in accordance with the approved guidelines. All participants gave written informed consent after being provided with an explanation of the experimental procedures. Participants The present study included 6 patients (5 females and 1 male; mean ± SD age, 34.5 ± 7.9 years). All participants were suffering from pharmacologically intractable focal epilepsy and underwent the implantation of intracranial electrodes as part of a presurgical evaluation. The experiment was conducted 2.0–2.8 weeks after the electrode implantation. The surgical evaluations suggested that the main epileptic foci for all the participants were outside the temporal pole; five of the foci were in the hippocampus and one was in the lateral temporal cortex. Neuropsychological assessments confirmed that all participants’ language ability and everyday memory were intact. The intelligence quotient (IQ), measured by the revised Wechsler Adult Intelligence Scale, was in the normal range in five participants, and in the mildly mentally retarded range in one participant (mean ± SD full-scale IQ: 91.8 ± 19.2; mean ± SD verbal IQ: 86.7 ± 12.0; mean ± SD performance IQ: 100.7 ± 27.3). During the experiment, no seizures were observed and all participants were mentally stable. All participants were right-handed, as assessed using the Edinburgh Handedness Inventory [42], and had normal or corrected-to-normal visual acuity. The data from different electrodes are reported elsewhere [43]. Anatomical magnetic resonance imaging (MRI) assessment Pre- and post-implantation anatomical assessments were conducted using the 1.5-T structural MRI scanning system (Signa TwinSpeed, General Electric Yokokawa) and multi-slice computed tomography (CT) scanner (Millennium VG, GE Medical Systems). For each participant, a whole brain T1-weighted MR image (matrix size = 256 × 256, field of view = 22 × 22 cm, and 76 slices resulting in voxel dimensions of 0.8594 × 0.8594 × 2.0 mm thick) and CT image (matrix size = 512 × 512, field of view = 22.8 × 22.8 cm, and 24 slices resulting in voxel dimensions of 0.4453 × 0.4453 × 5.0 mm thick) were acquired. Pre-implantation anatomical MRI and CT assessments and surgical evaluations did not reveal any structural abnormalities in the bilateral temporal poles of any participant. The intracranial electrodes were implanted using the stereotactic method [44], and the implantation sites were chosen based solely on clinical criteria. Subdural electrodes were implanted in the usual manner in both hemispheres for five participants and in the right hemisphere for one participant. Post-implantation anatomical MRI and CT assessments were conducted to confirm the positions of the electrodes of interest in the temporal pole. First, individual MRI data were segmented into gray matter, white matter, cerebrospinal fluid, skull, and scalp using the unified segmentation and normalization procedure [45] in SPM8 (http://www.fil.ion.ucl.ac.uk/spm/) implemented in MATLAB 2012b (MathWorks). Next, individual CT data were thresholded to remove the cranial content of the brain. Note that both the bony skull and the high-intensity electrode signals were preserved in this image processing. The resulting CT skull image was then co-registered to the skull image derived from MRI segmentation. To report the stereotactic coordinates of the electrode positions, individual T1-weighted MR images were normalized to a standard T1 template image defined by the Montreal Neurological Institute (MNI). The spatial transformation parameters from this normalization process were then applied to the gray matter image and the CT skull image. The CT skull image was re-thresholded and then binarized to limit the image contents to electrodes only. Finally, each electrode was well visualized in the brain surface-rendering of the gray matter image using the overlay function in MRICRON software (http://www.mccauslandcenter.sc.edu/mricro/mricron/). One of the study authors manually localized and confirmed the electrodes in the temporal pole. The mean ± SD MNI coordinates of the electrodes were as follows: right, x 33.8 ± 4.4, y 22.7 ± 3.2 z -39.3 ± 4.7; left, x -28.7 ± 5.2, y 21.2 ± 2.5, z -38.3 ± 2.7. The mean three-dimensional locations of the electrodes were projected on the MNI glass brain (SPM maximum intensity projection format; Fig 1). 10.1371/journal.pone.0162039.g001Fig 1 Location of electrodes in the temporal pole. Left) Representative anatomical magnetic resonance images. Cross hairs indicate the electrode location in the temporal pole. Right) Averaged coordinates of the electrodes in the temporal pole in the Montreal Neurological Institute space. Stimuli Fig 2 depicts the eye and mosaic stimuli. The eye stimuli were prepared from color photographs of the full-face neutral expressions of seven females and seven males who were looking either to the left or straight ahead. Only the eyes were used from the photographs, and no other facial features or eyebrows were visible in the stimuli. Mirror images of these stimuli were created using Photoshop 6.0 (Adobe). Eyes looking to the left or right were used for the averted-direction condition and eyes looking straight ahead were used for the straight-direction condition. The mean luminance of the images was kept constant using MATLAB 6.5 (MathWorks). 10.1371/journal.pone.0162039.g002Fig 2 Illustrations of the stimuli. The averted (first presentation)–straight (second presentation) direction conditions for the eyes and mosaics are shown. The mosaic stimuli were constructed from the eye stimuli. First, all of the eye stimuli were divided into small squares (10 vertical × 50 horizontal), and all squares were set to the mean luminance of pixels in each square. To construct objects conveying directional information in a manner similar to the eye stimuli, two sets of 49 small squares with the highest luminance were selected and arranged randomly to construct two large diagonally aligned squares. The squares were aligned diagonally because our preliminary experiment indicated that large squares arranged horizontally looked like eyes. The horizontal center of these large squares was comparable to the pupil positions of the eye stimuli. Other small squares were then arranged randomly in other areas. These manipulations resulted in mosaic stimuli equivalent to the corresponding original eye stimuli in terms of overall luminance and directional information, without the incorporation of eye features. Stimuli with different direction conditions were shown for the first and second stimulus presentations (i.e., averted after straight or straight after averted) to represent directional changes. Procedure The presentation of stimuli was controlled by SuperLab Pro 2.0 (Cedrus) and implemented using a Windows computer (FSA600, Teknos). Stimuli were presented on a 19-inch CRT monitor (GDM-F400, Sony) at a refresh rate of 100 Hz and a resolution of 1,024 × 768 pixels. The participants’ responses were recorded using a response box (RB-400, Cedrus). The experiments were conducted individually in a quiet room. The participants were seated comfortably with their heads supported by a chin-and-forehead rest positioned 0.57 m from the monitor. The resulting visual angle subtended by the stimulus was 1.5° vertically × 7.5° horizontally. Each stimulus was presented three times. In addition, a red cross was presented as the target in 15 trials. Thus, each participant performed a total of 183 trials: 42 trials each of averted eyes-straight eyes, straight eyes-averted eyes, averted mosaics-straight mosaics, and straight mosaics-averted mosaics, as well as 15 target trials. The stimuli were presented in a random order. In each trial, after the presentation of a cross-shaped fixation point for 500 ms, the first stimulus was presented for 500 ms in the center of the visual field. The second stimulus was then presented for 1,000 ms. In each target trial, instead of eyes or mosaic stimuli, the red cross was presented until a response was made. The participants were instructed to press a button using their right forefinger as quickly as possible after detecting the red cross. This task ensured that participants kept their attention on the stimuli, and it prevented the explicit processing of eye gaze. Performance on the target detection was perfect (correct identification rate = 100.0%), with no delay in reaction times (mean ± SD = 261.0 ± 15.6 ms). The post-hoc debriefing confirmed that the participants were not aware that the purpose of the experiment was to investigate gaze processing. The participants were also instructed not to blink while the stimuli were being presented. Inter-trial intervals varied randomly between 2,000 and 5,000 ms. To avoid habituation and drowsiness, participants were given short breaks every 45 trials. Prior to data collection, the participants were familiarized with the procedure by performing a block of 10 training trials. Data recording To examine cortical activity, intracranial EEG data were recorded using subdural platinum electrodes (2.3 mm diameter; Ad-tech). Depth platinum electrodes (0.8 mm diameter; Unique Medical) were also inserted to record subcortical activity (data not shown). Electrodes were referenced to electrodes (2.3 mm diameter; Ad-tech) that were embedded inside the scalp of the midline dorsal frontal region. Impedances were balanced and maintained below 5 kΩ. Data were amplified, filtered online (band pass: 0.5–300 Hz), and sampled at 1,000 Hz onto the hard disk drive of the EEG system (EEG-1100; Nihon Kohden). Online monitoring was conducted using a more restricted bandwidth of 0.5–120 Hz. Vertical and horizontal electrooculograms (EOGs) were simultaneously recorded using Ag/AgCl electrodes (Nihon Kohden). As in previous studies [46], off-line visual inspection confirmed that no contamination of the intracranial EEG data by EOGs occurred. Unobtrusive video recording of events was performed using the built-in video camera of the EEG system and an off-line analysis of the videos confirmed that all participants were fully engaged in the tasks. Data analysis: Preprocessing All preprocessing, time–frequency SPM analyses, and ERP analyses were performed using SPM8 (http://www.fil.ion.ucl.ac.uk/spm) implemented in MATLAB R2012b (MathWorks). Data obtained over 3,000 ms were sampled for each trial; pre-stimulus baseline data were collected for 1,000 ms, and experimental data were collected for 2,000 ms after stimulus onset at a sampling rate of 1,000 Hz. Epochs containing signals with amplitudes > ± 800 μV were excluded from the analyses and any epochs with absolute signal amplitude values > 5 SD from the mean or median signal amplitude for each electrode for each participant were rejected as artifacts. The frequencies of artifact-contaminated trials did not significantly differ across the conditions (mean ± SD = 6.20 ± 1.84%; p > 0.1, repeated-measures analysis of variance). Data analysis: Time–frequency SPM analysis Time–frequency SPM analyses [40,47,48] were performed to assess temporal pole activity. Time-frequency (power) maps were first calculated for each trial using continuous wavelet decomposition with 7-cycle Morlet wavelets during the whole epoch (-1,000–2,000 ms) and from 4 to 300 Hz, which covered theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), and gamma (30–300 Hz) bands. The time-frequency maps were then cropped to -200–500 ms within the time period evaluated to prevent edge effects of the wavelet transformation. Finally, the time-frequency maps were log-transformed and baseline-corrected separately for each frequency with respect to the mean power over the 200 ms pre-stimulus period. The time–frequency maps were then converted into two-dimensional images and entered into a general linear model (GLM) based on a fixed-effects analysis of the pooled error from all trials for all participants. Separate analyses were conducted for the first and second stimulus presentations, to test the effects of presence of eyes and changes in eye gaze direction, respectively. We set up a full factorial model including stimulus type (eyes or mosaics), stimulus direction (averted or straight), and hemisphere (right or left) as factors of interest. Corrections for non-sphericity (dependence and possible uneven variance between factor levels) were applied to ensure the assumption of an independent and identically distributed error for the GLM using the restricted maximum likelihood procedure [49]. The window of interest was restricted to the whole frequency range (4–300 Hz) during the post-stimulus period (0–500 ms) using explicit masking. Finally, time-frequency SPM{T} data were calculated for each contrast. Based on our interests, we analyzed the main effects of stimulus type (eyes versus mosaics) and interactions related to the stimulus type factor (i.e., the interactions of stimulus type × stimulus direction, stimulus type × hemisphere, and stimulus type × stimulus direction × hemisphere). For significant interactions, follow-up analyses for simple main effects of stimulus type were conducted. Significantly activated time-frequency clusters were identified if they reached an extent threshold of p < 0.05, which was family-wise error corrected for multiple comparisons over the whole time-frequency space (0–500 ms and 4–300 Hz), with a height threshold of p < 0.001 (uncorrected). Data analysis: ERP To analyze ERP, one-dimensional SPM analysis [50,51], a variant of the three-dimensional sensor-space-time SPM approach [47,48], was utilized. Because this study focused on a single electrode, the three-dimensional sensor-space-time SPM was reduced to the single-sensor-time SPM. Single trial responses from all trials for all participants were converted into one-dimensional line images after baseline correction for the -200–0 ms time period. The line images were then entered into the GLM in the same manner as for the time–frequency SPM analysis. Planned contrasts and statistical inferences were also performed in the same manner as for the time–frequency SPM analysis. Results Time–frequency SPM The time–frequency maps of temporal pole activity were analyzed using the GLM, which included the effects of stimulus type, stimulus direction, and hemisphere (Table 1). 10.1371/journal.pone.0162039.t001Table 1 Time-frequency regions showing significant temporal pole activity. Contrast Activation profile Peak Extent Time Frequency T-value Time Frequency Cluster size 　 (ms) (Hz) (ms) (Hz) (ms × Hz) The presence of eyes (first stimulus presentation)     Main effect of stimulus typea 226 121 5.36 215–236 101–150 705 267 55 3.95 248–282 52–63 249     Interaction of stimulus type × stimulus direction none     Interaction of stimulus type × hemisphereb 216 97 4.26 209–235 85–102 281     Interaction of stimulus type × stimulus direction × hemisphere none Changes in eye gaze direction (second stimulus presentation)     Main effect of stimulus typea 459 222 4.29 457–464 205–246 205     Interaction of stimulus type × stimulus direction none     Interaction of stimulus type × hemispherec 210 47 3.67 197–228 43–51 170     Interaction of stimulus type × stimulus direction × hemisphere none p < 0.05 cluster-level family-wise-error corrected. a contrast: (eyes—mosaics). b contrast: {(left eyes—left mosaics)—(right eyes—right mosaics)} inclusively masked by positive main effect of stimulus type. c contrast: {(right eyes—right mosaics)—(left eyes—left mosaics)} inclusively masked by positive main effect of stimulus type. First, temporal pole activity during the first stimulus presentation was analyzed to determine the effect of presence of eyes (Fig 3). The contrast of the main effect of stimulus type (eyes versus mosaics) revealed significant activation in the ranges of 215–236 ms and 101–150 Hz, which indicated that gamma-band activation in the bilateral temporal poles in response to eyes versus mosaics began at 215 ms. There was no significant activation for the interaction of stimulus type × stimulus direction, indicating that the above component was not specific to any gaze direction. The interaction of stimulus type × hemisphere revealed significant activation in the ranges of 209–235 ms and 85–102 Hz. Follow-up analyses indicated that this gamma-band component at a slightly lower frequency range than that of the main effect of stimulus type was evident only in the left temporal pole. There was no significant three-way interaction. 10.1371/journal.pone.0162039.g003Fig 3 Temporal pole activity in response to the presence of eyes (first stimulus presentation). Upper) Time–frequency maps. Lower) Statistical parametric maps (left) and effect sizes at the peak activation foci (right). p < 0.05 cluster-level family-wise error-corrected. EA = averted eyes; ES = straight eyes; MA = averted mosaics; MS = straight mosaics; 1 = first stimulus presentation. Next, to determine the effect of changes in eye gaze direction, temporal pole activity during the second stimulus presentation was analyzed (Fig 4). The main effect of stimulus type (eyes versus mosaics) was significant at the activation in the ranges of 457–464 ms and 205–246 Hz, which indicated that there was higher gamma-band activation in the bilateral temporal poles in response to eyes versus mosaics. There was no significant activation for the interaction of stimulus type × stimulus direction. The interaction of stimulus type × hemisphere revealed significant activation in the ranges of 197–228 ms and 43–51 Hz. Follow-up analyses revealed that this gamma-band component was evident only in the right temporal pole. There was no significant three-way interaction. 10.1371/journal.pone.0162039.g004Fig 4 Temporal pole activity in response to changes in eye gaze direction (second stimulus presentation). Upper) Time–frequency maps. Lower) Statistical parametric maps (left) and effect sizes at the peak activation foci (right). p < 0.05 cluster-level family-wise error-corrected. EA = averted eyes; ES = straight eyes; MA = averted mosaics; MS = straight mosaics; 2 = second stimulus presentation. ERP The ERP for each stimulus presentation was analyzed using the GLM in the same manner as the above time-frequency SPM analysis (Fig 5). The results showed no significant activation related to either the main effect of stimulus type or the interactions related to stimulus type. 10.1371/journal.pone.0162039.g005Fig 5 Grand-average event-related potentials in the temporal pole. EA = averted eyes; ES = straight eyes; MA = averted mosaics; MS = straight mosaics; 1 = first stimulus presentation; 2 = second stimulus presentation. Discussion The time–frequency analysis in the present study demonstrated that there was electric activation in the temporal pole in response to the presence of eyes versus mosaics and to changes in eye gaze direction versus mosaic direction. These results are inconsistent with those of several previous neuroimaging studies that did not observe temporal pole activation in response to the presence of eyes or changes in eye gaze direction (e.g., [10]). However, the discrepant findings may be accounted for by methodological differences. Whereas almost all of the previous studies assessed fMRI data, the present study analyzed electrical signals from the temporal pole. The temporal pole can show weak fMRI signals due to air-tissue inhomogeneity artifacts [52]. The previous studies may also have utilized ill-defined criteria for the temporal pole [31]. On the other hand, the present results are consistent with other neuroimaging studies showing that the temporal pole exhibits activation induced by eye information during active mind reading [32] and that there is a reduction in activation in response to adapted gaze direction [33]. However, these studies did not directly investigate neural activation in response to eyes or eye gaze direction changes. Thus, our results extend those of previous neuroimaging studies and provide the first electrophysiological evidence that the temporal pole is active in response to eyes and eye gaze direction changes. Furthermore, the time–frequency analysis in the present study provided temporal and frequency profiles for temporal pole activation in response to the presence of eyes and changes in eye gaze direction; activation was identified in the gamma-band beginning at approximately 200 ms. This temporal profile is reasonable when considering the anatomical fact that the temporal pole is located at the endpoint of the occipito–temporal visual stream [31] and the intracranial EEG findings that the inferior occipital gyrus (Sato et al., submitted) and STS region [17] exhibit gamma-band activity that begins at approximately 100 ms in response to eye information and that the temporal pole shows ERP peaks in response to faces about 100–200 ms later than the posterior cortices [7]. Similarly, the frequency profile observed in the present study is consistent with the findings of previous intracranial EEG studies showing that gamma-band activities in several brain regions, including the inferior occipital gyrus (Sato et al., submitted), STS region [17], and amygdala [43], are involved in the processing of eye information. Taken together, the present findings extend the current understanding of neural processing of eye information and indicate that the temporal pole exhibits gamma oscillations beginning at around 200 ms, which is after the activation in the posterior cortices, in response to eyes and eye gaze direction changes. Irrespective of the commonalities that can be seen in the frequency profiles of the temporal pole and amygdala during the processing of eye information, the temporal profiles and activation patterns of these two regions differ. Although the present results revealed that the temporal pole showed activation in response to the presence of eyes beginning at 215 ms, activation in the amygdala in response to the presence of eyes begins at 170 ms [43]. Whereas the temporal pole showed evident activation in response to both the presence of eyes and changes in eye gaze direction in the present study, the amygdala only shows clear activation in response to the presence of eyes [43]. These data indicate that the temporal pole conducts different types of processing of eye information at a later time point relative to the amygdala. Because the temporal pole and the amygdala have bidirectional connections [26], the temporal pole may receive inputs from the amygdala along with inputs from the posterior visual cortices. In contrast, the amygdala appears to conduct the processing of eyes prior to the receipt of visual inputs from the temporal pole using different pathways, such as the subcortical pathway via the superior colliculus and pulvinar [53,54] or the cortical pathway from the posterior regions [55,56]. The present results also revealed functional hemispheric differences such that the left and right temporal poles showed more evident gamma-band activation in response to the presence of eyes and changes in eye gaze direction, respectively. The existence of hemispheric asymmetry in the temporal pole is consistent with findings from other literatures, such as semantic cognition (for a review, see [57]). The left hemispheric dominance for processing eyes may be in line with a study of brain-damaged patients that observed a left hemispheric dominance in mind reading ability using information from eyes [58]. The right hemispheric dominance regarding eye gaze direction changes is consistent with several previous neuroimaging studies showing a right hemispheric dominance during the processing of dynamic facial signals (e.g., [59]). However, it must be noted that the electrode placement in the present study was based on anatomical information, and hence, it is possible that the electrodes in the right and left temporal poles were not comparable functionally. Further investigation is necessary to determine whether a functional hemispheric asymmetry exists with respect to temporal pole activities during the processing of eye information. In contrast to the hypotheses of the present study, the results did not reveal evident ERP or low-frequency band activation in response to the presence of eyes and changes in eye gaze direction. Although drawing of any conclusions based on a null finding should be postponed, this result suggests that the temporal pole may not process eye information using low-frequency activity. The results showing that the gamma-band activity exhibited eye information-related activity, but low-frequency band activity did not, may also have some interesting implications. It has been proposed that the brain generally uses electrical activity in low-frequency bands, such as the theta-band, for long-range inter-regional communication and that this low-frequency activity entrains local intra-regional gamma-band activation that corresponds to the computation of cell populations (e.g., [35,60]). The present findings suggest that this cross-frequency theta–gamma coupling may not be involved in the processing of eye information in the temporal pole. Instead, it may be possible that inter-regional communication between the temporal pole and other brain areas during the processing of eye information is accomplished via the same-frequency gamma–gamma coupling (cf. [61]). The present findings have several implications. First, the activation of the temporal pole and its specific temporal profile during the processing of eyes and eye gaze direction changes updates the current understanding of the spatiotemporal neural network dynamics involved in the processing of eye information. Although the temporal pole is commonly associated with the processing of social interactions [31], this area has received relatively little attention regarding the processing of eye information [62–64]. Second, the occurrence of temporal pole activation during the processing of eye information suggests that the impaired social functioning observed in monkeys following damage to this region (e.g., [20]) could at least partially be attributed to the impaired processing of eye information. Likewise, human studies have reported that the brain damage [65,66] and atrophy associated with semantic dementia [67,68] including the temporal pole induce social malfunctioning and this may be related to impaired processing of eye information. Finally, the involvement of gamma oscillations in the temporal pole during the processing of eye information corroborates previous evidence showing that the brain uses this frequency range to accomplish information processing (e.g., [69]; for a review, see [70]). A limitation of this study should be acknowledged. We asked participants to engage in a dummy task to investigate automatic activation in response to eyes and eye gaze direction changes. However, this task did not reveal any details regarding the processing of eye information associated with temporal pole activity. Different tasks might enhance or suppress activity in the temporal pole at particular temporal and frequency points. For example, a previous fMRI study has observed activation in the temporal pole during intentional mind reading based on eye information [32]. This particular mind reading task during the observation of eyes may enhance gamma-band activity beginning at about 200 ms, as we found in the present study, or it may induce additional later gamma-band activation in the temporal pole. In future studies, participants should be asked to engage in intentional cognitive processes in response to eyes or eye gaze direction changes to specify the functional correlates of temporal pole activity. In summary, the present intracranial EEG data revealed that the bilateral temporal poles exhibited a greater degree of gamma-band activation beginning at 215 ms in response to eyes compared with mosaics, irrespective of gaze direction. Additionally, the right temporal pole showed a greater degree of gamma-band activation beginning at 197 ms in response to directional changes of the eyes compared with mosaics. These results suggest that the temporal pole uses gamma oscillations to process the presence of eyes and changes in eye gaze direction at a relatively late time stage compared with brain regions in the posterior cortices, such as the inferior occipital gyrus and STS region. Supporting Information S1 Dataset Datasets of Temporal pole activity. (XLSX) Click here for additional data file. This study was supported by funds from the Benesse Corporation and the JSPS Funding Program for Next Generation World-Leading Researchers (LZ008). ==== Refs References 1 McCarty CA . Eyes are the window to the soul . Clin Exp Ophthalmology . 2011 ; 39 : 291 –292 . 2 Strom JC , Buck RW . Staring and participants' sex: physiological and subjective reactions . Pers Soc Psychol Bull . 1979 ; 5 : 114 –117 . 3 Frischen A , Bayliss AP , Tipper SP . Gaze cueing of attention: Visual attention, social cognition, and individual differences . Psychol Bull . 2007 ; 133 : 694 –724 . 17592962 4 Baron-Cohen S , Jolliffe T , Mortimore C , Robertson M . Another advanced test of theory of mind: Evidence from very high functioning adults with autism or asperger syndrome . J Child Psychol Psychiatry . 1997 ; 38 : 813 –822 . 9363580 5 Argyle M , Cook M . Gaze and mutual gaze New York : Cambridge University Press ; 1976 . 6 Nation K , Penny S . Sensitivity to eye gaze in autism: Is it normal? Is it automatic? Is it social? Dev Psychopathol . 2008 ; 20 : 79 –97 . 10.1017/S0954579408000047 18211729 7 Allison T , Puce A , McCarthy G . Social perception from visual cues: Role of the STS region . Trends Cogn Sci . 2000 ; 4 : 267 –278 . 10859571 8 Puce A , Allison T , Bentin S , Gore JC , McCarthy G . Temporal cortex activation in humans viewing eye and mouth movements . J Neurosci . 1998 ; 18 : 2188 –2199 . 9482803 9 Wicker B , Michel F , Henaff MA , Decety J . Brain regions involved in the perception of gaze: A PET study . Neuroimage . 1998 ; 8 : 221 –227 . 9740764 10 Hoffman EA , Haxby JV . Distinct representations of eye gaze and identity in the distributed human neural system for face perception . Nat Neurosci . 2000 ; 3 : 80 –84 . 10607399 11 Calder AJ , Lawrence AD , Keane J , Scott SK , Owen AM , Christoffels I , et al Reading the mind from eye gaze . Neuropsychologia . 2002 ; 40 : 1129 –1138 . 11931917 12 Hooker CI , Paller KA , Gitelman DR , Parrish TB , Mesulam MM , Reber PJ . Brain networks for analyzing eye gaze . Cogn Brain Res . 2003 ; 17 : 406 –418 . 13 Pelphrey KA , Singerman JD , Allison T , McCarthy G . Brain activation evoked by perception of gaze shifts: the influence of context . Neuropsychologia . 2003 ; 41 : 156 –170 . 12459214 14 Engell AD , Haxby JV . Facial expression and gaze-direction in human superior temporal sulcus . Neuropsychologia . 2007 ; 45 : 3234 –3241 . 17707444 15 Sato W , Kochiyama T , Uono S , Yoshikawa S . Time course of superior temporal sulcus activity in response to eye gaze: A combined fMRI and MEG study . Soc Cogn Affect Neurosci . 2008 ; 3 : 224 –232 . 10.1093/scan/nsn016 19015114 16 Sato W , Kochiyama T , Uono S , Yoshikawa S . Commonalities in the neural mechanisms underlying automatic attentional shifts by gaze, gestures, and symbols . Neuroimage . 2009 ; 45 : 984 –992 . 10.1016/j.neuroimage.2008.12.052 19167506 17 Caruana F , Cantalupo G , Lo Russo G , Mai R , Sartori I , Avanzini P . Human cortical activity evoked by gaze shift observation: an intracranial EEG study . Hum Brain Mapp . 2014 ; 35 : 1515 –1528 . 10.1002/hbm.22270 23568412 18 Kawashima R , Sugiura M , Kato T , Nakamura A , Hatano K , Ito K , et al The human amygdala plays an important role in gaze monitoring. A PET study . Brain . 1999 ; 122 : 779 –783 . 10219788 19 Haxby JV , Hoffman EA , Gobbini MI . The distributed human neural system for face perception . Trends Cogn Sci . 2000 ; 4 : 223 –233 . 10827445 20 Franzen EA , Myers RE . Neural control of social behavior: prefrontal and anterior temporal cortex . Neuropsychologia . 1973 ; 11 : 141 –157 . 4197348 21 Horel JA , Keating EG , Misantone LJ . Partial Klüver-Bucy syndrome produced by destroying temporal neocortex or amygdala . Brain Res Cogn Brain Res . 1975 ; 94 : 347 –359 . 22 Kling A , Steklis HD . A neural substrate for affiliative behavior in nonhuman primates . Brain Behav Evol . 1976 ; 13 : 216 –238 . 825190 23 Kling AS , Tachiki K , Lloyd R . Neurochemical correlates of the Klüver-Bucy syndrome by in vivo microdialysis in monkey . Behav Brain Res . 1993 ; 56 : 161 –170 . 7902099 24 Gloor P . The Temporal Lobe and Limbic System . New York : Oxford University Press ; 1997 . 25 Insausti R . Comparative neuroanatomical parcellation of the human and nonhuman primate temporal pole . J Comp Neurol . 2013 ; 521 : 4163 –4176 . 10.1002/cne.23431 23881792 26 Nakamura K , Kawashima R , Sato N , Nakamura A , Sugiura M , Kato T , et al Functional delineation of the human occipito-temporal areas related to face and scene processing: A PET study . Brain . 2000 ; 123 : 1903 –1912 . 10960054 27 Nakamura K , Kubota K . The primate temporal pole: its putative role in object recognition and memory . Behav Brain Res . 1996 ; 77 : 53 –77 . 8762159 28 Sugiura M , Kawashima R , Nakamura K , Sato N , Nakamura A , Kato T , et al Activation reduction in anterior temporal cortices during repeated recognition of faces of personal acquaintances . Neuroimage . 2001 ; 13 : 877 –890 . 11304083 29 Rajimehr R , Young JC , Tootell RB . An anterior temporal face patch in human cortex, predicted by macaque maps . Proc Natl Acad Sci USA . 2009 ; 106 : 1995 –2000 . 10.1073/pnas.0807304106 19179278 30 Nasr S , Tootell RB . Role of fusiform and anterior temporal cortical areas in facial recognition . Neuroimage . 2012 ; 63 : 1743 –1753 . 10.1016/j.neuroimage.2012.08.031 23034518 31 Olson IR , Plotzker A , Ezzyat Y . The Enigmatic temporal pole: a review of findings on social and emotional processing . 2007 ; Brain . 130 : 1718 –1731 . 17392317 32 Platek SM , Critton SR , Myers TE , Gallup GG . Contagious yawning: the role of self-awareness and mental state attribution . Brain Res Cogn Brain Res . 2003 ; 17 : 223 –227 . 12880893 33 Calder AJ , Beaver JD , Winston JS , Dolan RJ , Jenkins R , Eger E , et al Separate coding of different gaze directions in the superior temporal sulcus and inferior parietal lobule . Curr Biol . 2007 ; 17 : 20 –25 . 17208181 34 Halgren E , Baudena P , Heit G , Clarke M , Marinkovic K . Spatio-temporal stages in face and word processing. 1. Depth-recorded potentials in the human occipital, temporal and parietal lobes . J Physiol Paris . 1994 ; 88 : 1 –50 . 8019524 35 Canolty RT , Knight RT . The functional role of cross-frequency coupling . Trends Cogn Sci . 2010 ; 14 : 506 –515 . 10.1016/j.tics.2010.09.001 20932795 36 Allison T , Puce A , Spencer DD , McCarthy G . Electrophysiological studies of human face perception. I: Potentials generated in occipitotemporal cortex by face and non-face stimuli . Cereb Cortex . 1999 ; 9 : 415 –430 . 10450888 37 Edwards E , Soltani M , Deouell LY , Berger MS , Knight RT . High gamma activity in response to deviant auditory stimuli recorded directly from human cortex . J Neurophysiol . 2005 ; 94 : 4269 –4280 . 16093343 38 Makeig S , Debener S , Onton J , Delorme A . Mining event-related brain dynamics . Trends Cogn Sci . 2004 ; 8 : 204 –210 . 15120678 39 Adrian ED . Olfactory reactions in the brain of the hedgehog . J Physiol . 1942 ; 100 : 459 –473 . 16991539 40 Kilner JM , Kiebel SJ , Friston KJ . Applications of random field theory to electrophysiology . Neurosci Lett . 2005 ; 374 : 174 –178 . 15663957 41 Okada T , Sato W , Toichi M . Right hemispheric dominance in gaze-triggered reflexive shift of attention in humans . Brain Cogn . 2006 ; 62 : 128 –133 . 16730869 42 Oldfield RC . The assessment and analysis of handedness: the Edinburgh inventory . Neuropsychologia . 1971 ; 9 : 97 –113 . 5146491 43 Sato W , Kochiyama T , Uono S , Matsuda K , Usui K , Inoue Y , et al Rapid amygdala gamma oscillations in response to eye gaze . PLoS One . 2011 ; 6 : e28188 10.1371/journal.pone.0028188 22140541 44 Mihara T , Baba K . Combined use of subdural and depth electrodes. In Luders HO , Comair YG editors. Epilepsy Surgery (second ed. ). Philadelphia, PA : Lippincott Williams and Wilkins ; 2001 pp. 613 –621 45 Ashburner J , Friston KJ . Unified segmentation . Neuroimage . 2005 ; 26 : 839 –851 . 15955494 46 Lachaux JP , Rudrauf D , Kahane P . Intracranial EEG and human brain mapping . J Physiol Paris . 2003 ; 97 : 613 –628 . 15242670 47 Kilner JM , Friston KJ . Topological inference for EEG and MEG . Ann Appl Stat . 2010 ; 4 : 1272 –1290 . 48 Litvak V , Mattout J , Kiebel S , Phillips C , Henson R , Kilner J , et al EEG and MEG data analysis in SPM8 . Comput Intell Neurosci . 2011 ; 852961 10.1155/2011/852961 21437221 49 Friston KJ , Glaser DE , Henson RN , Kiebel S , Phillips C , Ashburner J . Classical and Bayesian inference in neuroimaging: Applications . Neuroimage . 2002 ; 16 : 484 –512 . 12030833 50 Oswal A , Litvak V , Sauleau P , Brown P . Beta reactivity, prospective facilitation of executive processing, and its dependence on dopaminergic therapy in Parkinson's disease . J Neurosci . 2012 ; 32 : 9909 –9916 . 10.1523/JNEUROSCI.0275-12.2012 22815506 51 Sato W , Kochiyama T , Uono S , Matsuda K , Usui K , Inoue Y , et al Temporal profile of amygdala gamma oscillations in response to faces . J Cogn Neurosci . 2012 ; 24 : 1420 –1433 . 10.1162/jocn_a_00142 21981674 52 Devlin JT , Russell RP , Davis MH , Price CJ , Wilson J , Moss HE , et al Susceptibility-induced loss of signal: comparing PET and fMRI on a semantic task . Neuroimage . 2000 ; 11 : 589 –600 . 10860788 53 Tamietto M , Pullens P , de Gelder B , Weiskrantz L , Goebel R . Subcortical connections to human amygdala and changes following destruction of the visual cortex . Curr Biol . 2012 ; 22 : 1449 –1455 . 10.1016/j.cub.2012.06.006 22748315 54 Rafal RD , Koller K , Bultitude JH , Mullins P , Ward R , Mitchell AS , Bell AH . Connectivity between the superior colliculus and the amygdala in humans and macaque monkeys: Virtual dissection with probabilistic DTI tractography . J Neurophysiol . 2015 ; 114 : 1947 –1962 . 10.1152/jn.01016.2014 26224780 55 Catani M , Jones DK , Donato R , Ffytche DH . Occipito-temporal connections in the human brain . Brain . 2003 ; 126 : 2093 –2107 . 12821517 56 Latini F . New insights in the limbic modulation of visual inputs: The role of the inferior longitudinal fasciculus and the Li-Am bundle . Neurosurg Rev . 2015 ; 38 : 179 –89 . 10.1007/s10143-014-0583-1 25323099 57 Wong C , Gallate J . The function of the anterior temporal lobe: A review of the empirical evidence . Brain Res . 2012 ; 1449 : 94 –116 . 10.1016/j.brainres.2012.02.017 22421014 58 Dal Monte O , Schintu S , Pardini M , Berti A , Wassermann EM , Grafman J , et al The left inferior frontal gyrus is crucial for reading the mind in the eyes: Brain lesion evidence . Cortex . 2014 ; 58 : 9 –17 . 10.1016/j.cortex.2014.05.002 24946302 59 Sato W , Kochiyama T , Yoshikawa S , Naito E , Matsumura M . Enhanced neural activity in response to dynamic facial expressions of emotion: An fMRI study . Brain Res Cogn Brain Res . 2004 ; 20 : 81 –91 . 15130592 60 Sato W , Kochiyama T , Uono S , Matsuda K , Usui K , Inoue Y , et al Rapid, high-frequency, and theta-coupled gamma oscillations in the inferior occipital gyrus during face processing . Cortex . 2014 ; 60 : 52 –68 . 10.1016/j.cortex.2014.02.024 24745564 61 Varela F , Lachaux JP , Rodriguez E , Martinerie J . The brainweb: Phase synchronization and large-scale integration . Nat Rev Neurosci . 2001 ; 2 : 229 –239 . 11283746 62 Emery NJ . The eyes have it: the neuroethology, function and evolution of social gaze . Neurosci Biobehav Rev . 2000 ; 24 : 581 –604 . 10940436 63 Itier RJ , Batty M . Neural bases of eye and gaze processing: the core of social cognition . Neurosci Biobehav Rev . 2009 ; 33 : 843 –863 . 10.1016/j.neubiorev.2009.02.004 19428496 64 Senju A , Johnson MH . The eye contact effect: Mechanisms and development . Trends Cogn Sci . 2009 ; 13 : 127 –134 . 10.1016/j.tics.2008.11.009 19217822 65 Lilly R , Cummings JL , Benson DF , Frankel M . The human Klüver-Bucy syndrome . Neurology . 1983 ; 33 : 1141 –1145 . 6684248 66 Ghika-Schmid F , Assal G , De Tribolet N , Regli F . Klüver-Bucy syndrome after left anterior temporal resection . Neuropsychologia . 1995 ; 33 : 101 –113 . 7731534 67 Thompson SA , Patterson K , Hodges JR . Left/right asymmetry of atrophy in semantic dementia: behavioral-cognitive implications . Neurology . 2003 ; 61 : 1196 –1203 . 14610120 68 Gorno-Tempini ML , Rankin KP , Woolley JD , Rosen HJ , Phengrasamy L , Miller BL . Cognitive and behavioral profile in a case of right anterior temporal lobe neurodegeneration . 2004 ; Cortex . 40 : 631 –644 . 15505973 69 Tallon-Baudry C , Bertrand O , Delpuech C , Pernier J . Stimulus specificity of phase-locked and non-phase-locked 40 Hz visual responses in human . J Neurosci . 1996 ; 16 : 4240 –4249 . 8753885 70 Herrmann CS , Fründ I , Lenz D . Human gamma-band activity: A review on cognitive and behavioral correlates and network models . Neurosci Biobehav Rev . 2010 ; 34 : 981 –992 . 10.1016/j.neubiorev.2009.09.001 19744515

PMC005xxxxxx/PMC5003338.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757136410.1371/journal.pone.0161601PONE-D-16-19180Research ArticleMedicine and Health SciencesSurgical and Invasive Medical ProceduresBiology and Life SciencesAnatomyDigestive SystemMouthMandibleMedicine and Health SciencesAnatomyDigestive SystemMouthMandibleBiology and Life SciencesAnatomyMusculoskeletal SystemSymphysesMedicine and Health SciencesAnatomyMusculoskeletal SystemSymphysesEngineering and TechnologyMeasurementDistance MeasurementMedicine and Health SciencesOral MedicineOrthodonticsBiology and Life SciencesAnatomyDigestive SystemTeethMolarsMedicine and Health SciencesAnatomyDigestive SystemTeethMolarsBiology and Life SciencesAnatomyHeadJawTeethMolarsMedicine and Health SciencesAnatomyHeadJawTeethMolarsMedicine and Health SciencesSurgical and Invasive Medical ProceduresPlastic Surgery and Reconstructive TechniquesPhysical SciencesMathematicsGeometrySymmetryComprehensive Analysis of Mandibular Residual Asymmetry after Bilateral Sagittal Split Ramus Osteotomy Correction of Menton Point Deviation Residual Asymmetry after Virtually Surgical Correction of Mandibular DeviationLin Han 1Zhu Ping 1Lin Qiuping 2Huang Xiaoqiong 1Xu Yue 2*Yang Xiaoping 11 Department of Oral and Maxillofacial Surgery, Guanghua School of Stomatology, Hospital of Stomatology, Sun Yat-sen University, Guangdong Provincial Key Laboratory of Stomatology, Guangzhou, China2 Department of Orthodontics, Guanghua School of Stomatology, Hospital of Stomatology, Sun Yat-sen University, Guangdong Provincial Key Laboratory of Stomatology, Guangzhou, ChinaCray James EditorMedical University of South Carolina, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: YX. Data curation: HL PZ. Formal analysis: PZ QL. Investigation: HL PZ QL XH. Methodology: YX XY. Project administration: YX. Resources: YX XY. Software: HL PZ. Validation: YX HL. Writing – original draft: HL. Writing – review & editing: YX PZ HL. * E-mail: kou9315@hotmail.com29 8 2016 2016 11 8 e016160112 5 2016 8 8 2016 © 2016 Lin et al2016Lin et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Purpose Facial asymmetry often persists even after mandibular deviation corrected by the bilateral sagittal split ramus osteotomy (BSSRO) operation, since the reference facial sagittal plane for the asymmetry analysis is usually set up before the mandibular menton (Me) point correction. Our aim is to develop a predictive and quantitative method to assess the true asymmetry of the mandible after a midline correction performed by a virtual BSSRO, and to verify its availability by evaluation of the post-surgical improvement. Patients and Methods A retrospective cohort study was conducted at the Hospital of Stomatology, Sun Yat-sen University (China) of patients with pure hemi-mandibular elongation (HE) from September 2010 through May 2014. Mandibular models were reconstructed from CBCT images of patients with pre-surgical orthodontic treatment. After mandibular de-rotation and midline alignment with virtual BSSRO, the elongation hemi-mandible was virtually mirrored along the facial sagittal plane. The residual asymmetry, defined as the superimposition and boolean operation of the mirrored elongation side on the normal side, was calculated, including the volumetric differences and the length of transversal and vertical asymmetry discrepancy. For more specific evaluation, both sides of the hemi-mandible were divided into the symphysis and parasymphysis (SP), mandibular body (MB), and mandibular angle (MA) regions. Other clinical variables include deviation of Me point, dental midline and molar relationship. The measurement of volumetric discrepancy between the two sides of post-surgical hemi-mandible were also calculated to verify the availability of virtual surgery. Paired t-tests were computed and the P value was set at .05. Results This study included 45 patients. The volume differences were 407.8±64.8 mm3, 2139.1±72.5 mm3, and 422.5±36.9 mm3; residual average transversal discrepancy, 1.9 mm, 1.0 mm, and 2.2 mm; average vertical discrepancy, 1.1 mm, 2.2 mm, and 2.2 mm (before virtual surgery). The post-surgical volumetric measurement showed no statistical differences between bilateral mandibular regions. Conclusions Mandibular asymmetry persists after Me point correction. A 3D quantification of mandibular residual asymmetry after Me point correction and mandible de-rotation with virtual BSSRO sets up a true reference mirror plane for comprehensive asymmetry assessment of bilateral mandibular structure, thereby providing an accurate guidance for orthognathic surgical planning. The author(s) received no specific funding for this work. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Facial symmetry is an essential prerequisite of a successful reconstructive and aesthetic plastic surgery. Pure hemi-mandibular elongation is often manifested as deviation of the mandibular menton (Me) point and is usually accompanied by functional abnormalities, inducing a considerable impact on the patient’s psychological wellbeing [1]. In lieu of the geometric complexity of dentition and the bony structures and soft tissue of the face, the correction of severe facial asymmetry is an extremely challenging prospect in orthognathic surgery [2]. Among the current surgical approaches, bilateral sagittal split ramus osteotomy (BSSRO) is one of the most common orthognathic procedures: it involves the rotation and movement of the osteotomized segments of the mandible with sufficient flexibility and safety of the mandibular nerve [3–5]. Orthognathic surgery is performed with correction of the mandibular rotation first, followed by marginal and surface modification [6]. Some previous reports evaluate the mandibular asymmetry by mirroring the semi-mandible with the facial mid-sagittal plane without considering secondary mandibular rotation after Me point correction, while others assess the mandibular asymmetry before the midline correction [7–9]. The comprehensive pre-surgical analysis is of significance, as it can aid in planning the deviation correction and residual asymmetry modification at the same time. As a result, it is unreasonable to use the facial sagittal plane as the reference plane for the symmetry analysis before surgically correcting the midline and de-rotating the mandible [10]. Over the past decades, the introduction of three-dimensional (3D) measurement technology has significantly improved the surgical planning of the mandibular deviation correction as well as the prediction of the treatment outcome. However, today, few statistical reports on the volume or length differences and specific locations of asymmetry on patients with mandibular deviation have been published. In most cases, the secondary mandibular morphology surgical correction after BSSRO is made on the basis of the surgeons’ clinical experience, rather than a precisely quantitative assessment of the characteristics of mandibular deviation [11]. This makes it difficult to predict ideal symmetry after the surgery. In other words, a predictively quantitative analysis of bilateral mandibular symmetry in patients with mandibular deviation after virtual BSSRO yields a much better clinical reference value than an evaluation before bony segment de-rotation in deviation modification surgery. The purpose of this study is to quantitatively predict the mandibular asymmetry in patients with mandibular deviation after Me point virtual correction. We hypothesize that the mirroring reference plane differed from the mid-sagittal plane, which causes the residual asymmetry of bilateral mandibular structure after Me point correction. The specific aims of the study are to measure the extent (volume and length) of discrepancy and specific location of mandibular asymmetry as well as verification by post-surgical outcomes. Materials and Methods Study design/sample To address the research purpose, this study is designed as a retrospective investigation. The study population is composed of patients diagnosed with pure hemi-mandibular elongation (HE) at the Hospital of Stomatology, Sun Yat-sen University, Guangzhou, China from September 2010 through May 2014. The study protocol was approved by the institutional review board at the Sun Yat-sen University and all participants signed an informed consent agreement. To be included in the study sample, patients had to meet the following inclusion criteria: (a) ANB angle of <0°, (b) dental midline deviation of ≥ 2 mm, and (c) Me point deviation of ≥ 4 mm. Patients who had a history of proven fracture, surgery, or trauma of the mandible were excluded as study subjects. Clinical examination Data on the following parameters were collected for each patient: Deviation of Me point: Deviation of the Me point was measured as the distance between the Me point and the facial midline. The facial midline was defined as the perpendicular bisector of the line drawn between the centers of the right and left pupils. Deviation of the dental midline: Deviation of the dental midlines was measured as the horizontal distance between the mesial contact points of the maxillary central incisors and mandibular central incisors. Inclination of the maxillary occlusal plane: To determine the inclination of the maxillary occlusal plane, patients were required to bite onto a tongue blade; then, the slant in the occlusal plane was detected as the angle between the blade and the inter-pupillary plane. Molar relationship. ANB angle. Virtual surgery procedure Before surgery, all of the patients were received pre-surgical orthodontic treatment. Alignment and coordination of the maxilla-mandibular dental arch width were performed to ascertain that the Me point could be adjusted to a precise position. CBCT images of patients with pre-surgical orthodontic treatment were acquired using the DCT Pro CBCT device (Vatech Co., Ltd., Hwasung, Korea), while 3D reconstruction was performed using the Mimics™ program (Materialise’s interactive medical image control system, Mimics, 14.0; Materialise, Leuven, Belgium). Virtual BSSRO was then performed on the 3D craniofacial models. The distal segments were separated from the proximal segment on both sides of the mandible, and the proximal segment was rotated and shifted backwards to achieve normal jaw relationship and midline alignment. Maxillary surgery (Lefort I) was additionally performed if an inclination of the occlusal plane was noted. After correction of the Me point, a real virtual mirroring plane (a sagittal plane passing through the facial midline and corrected Me point) for residual asymmetry measurement was created. The elongation side of the hemi-mandible was mirrored to create a mirrored model. Residual asymmetry was defined as the superimposition and boolean calculation of mirrored elongation side on the normal side (Fig 1). 10.1371/journal.pone.0161601.g001Fig 1 Procedure for measurement of the mandibular residual asymmetry: (a) Incision design and measurement before surgery; (b) Mandible segment backwards to normal maxillo-mandibular relationship; (c) Me point—facial midline alignment by mandibular de-rotation; (d) Segmentation along the facial middle sagittal plane, green color represents the elongation hemi-mandible; blue color represents the normal hemi-mandible; (e) Mirroring of elongation hemi-mandible along facial middle sagittal plane; (f) Superimposition and boolean operation of mirrored elongation hemi-mandible and normal side, pink color in the box represents the residual asymmetry between the two sides of the mandible. Residual asymmetry measurement To ensure a thorough assessment of mandibular asymmetry, both sides of therepositioned mandible is divided into three parts: the SP region comprising of the symphysis and parasymphysis, MB region comprising of the mandibular body, and the MA region comprising of the mandibular angle (the SP, MB, and MA regions are abbreviated as S, B, and A, respectively; the elongation and normal sides are represented by the subscripts e and n, respectively) [12] (Fig 2). The volume (mm3) of the residual asymmetry was measured using the Mimics™ automatic function. For measurement of transversal length discrepancy, on the elongation side, the extent of transversal length (de) is defined as the distance from the inflecting point (point I) at the site of asymmetry to the facial sagittal plane. For comparison, on the normal side, the distance (dn) is measured from reference point (R point) to the facial sagittal plane. The vertical asymmetry discrepancy is defined as the length difference between distance De (measured from point I to Frankfort horizontal (FH) plane) and distance Dn (measured from point R to FH plane) (Fig 3). 10.1371/journal.pone.0161601.g002Fig 2 Mandibular division: (a) Both sides of hemi-mandible were divided into three regions respectively after virtual Me point correction; (b) Schematic diagram of three mandibular regions: S, Symphysis and parasymphysis mental region; M, Mandibular body region; A, Mandibular angle region; the subscripts e, n represent elongation and normal side. 10.1371/journal.pone.0161601.g003Fig 3 3D measurement of the mandibular asymmetry (e.g., MB region): (a) Evaluation of the extent of transversal asymmetry (cross-sectional view). Point I represents the point of inflection in the asymmetry site on elongation side, while point R represents the contralateral reference point on the normal side. The distance from points I/R to the facial middle sagittal plane were measured for transversal asymmetry extent evaluation, indicated as de and dn respectively; (b) Evaluation of the extent of vertical asymmetry (front and lateral view). The distance from point I/R to the Frankfort horizontal (FH) plane were measured to determine the extent of vertical asymmetry, indicated as De/Dn. Post-surgical verification The availability of the virtual surgery is verified by post-surgical measurement. Prior to surgery, a computer-guided digital template for each patient was designed and fabricated to help perform the precise osteotomy and to retain the modified bony segments in the proper position of mandible in clinic as in the virtual surgical planning. CBCT images of patients were acquired six months after surgery. The reconstructed mandible model is divided into two hemi-mandibles along the facial sagittal plane, followed by volumetric measurement (S2 Fig). Clinical examination parameters are the same as mentioned above. Data Analysis The values of volume and length discrepancy of mandibular asymmetry are presented as mean ± standard deviation. The measurements were processed and analyzed using SPSS Inc. Released 2007, SPSS for Windows, Version 16.0. (SPSS Inc., Chicago, IL). The paired t-test was used to calculate the statistical significance of the difference between the volume and length of asymmetry in the SP, MB, and MA regions of the mirrored elongation and normal sides. A confidence level of a P value less than .05 was accepted as significant. Results A total of 45 pure HE patients (23 males, age, 23±4 yrs and 22 females, age, 25±5 yrs) were reviewed retrospectively. The average volumetric discrepancy after virtual Me point correction in the SP, MB, and MA regions were 407.8±64.8 mm3, 2139.1±72.5 mm3, and 422.5±36.9mm3, respectively. All with significant differences (P < 0.05), while no statistical difference was found between all of the bilateral mandibular regions in the post-surgical volumetric measurement (Table 1). Typical cases of mandibular residual asymmetry in different regions after virtual Me point correction are shown in S1 Fig. 10.1371/journal.pone.0161601.t001Table 1 The volumetric measurement of the mandibular regions after virtual Me point correction and after clinical orthognathic surgery. Virtual Me point correction group Post-surgical group SP (mm3) MB (mm3) MA (mm3) SP (mm3) MB (mm3) MA (mm3) Elongation side 3899.5±244.8 12189.2±303.1 2310.8±175.4 3479.3±189.7 10184.4±268.0 1951.2±204.1 Normal side 3491.7±190.9 10050.2±241.0 1888.3±163.3 3418.3±202.5 10076.3±313.3 1883.0±139.0 Volume difference 407.8±64.8 2139.1±72.5 422.5±36.9 61.0±247.5 108.2±405.2 68.2±231.5 P value <0.05 <0.05 <0.05 0.106 0.080 0.054 n = 45; Me, menton; SP, symphysis and parasymphysis; MB, mandibular body; MA, mandibular angle; The average transversal and vertical discrepancy of the asymmetry sites in the SP, MB, and MA regions were 1.9 mm transversally and 1.1 mm vertically, 1.0 mm transversally and 2.2 mm vertically, 2.2 mm transversally and 2.2 mm vertically, respectively, and all with significant differences (P < 0.05), as shown in Tables 2 and 3. 10.1371/journal.pone.0161601.t002Table 2 Measurement of the transversal discrepancy of asymmetry site in SP, MB, and MA regions after virtual Me point correction. SP (mm) MB (mm) MA (mm) Transversal length on elongation side (de) 6.6±2.1 7.2±1.8 9.3±2.8 Transversal length on normal side (dn) 4.7±1.2 6.2±1.8 7.0±1.9 Transversal discrepancy 1.9±1.2 1.0±0.6 2.2±1.1 P value <0.05 <0.05 <0.05 n = 45; Me, menton; SP, symphysis and parasymphysis; MB, mandibular body; MA, mandibular angle; 10.1371/journal.pone.0161601.t003Table 3 Measurement of the vertical discrepancy of asymmetry site in SP, MB, and MA regions after virtual Me point correction. SP (mm) MB (mm) MA (mm) Vertical length on elongation side (De) 79.4±3.8 77.6±2.7 72.8±2.4 Vertical length on normal side (Dn) 78.2±3.5 75.4±2.7 70.8±2.8 Vertical discrepancy 1.1±1.5 2.2±2.6 2.0±1.7 P value <0.05 <0.05 <0.05 n = 45; Me, menton; SP, symphysis and parasymphysis; MB, mandibular body; MA, mandibular angle; The results of the clinical examinations before virtual surgery and after surgery are shown in the S1 Table. The deviation of the Me point and dental midline before virtual surgery were 5.4±0.9 mm and 2.9±0.6 mm, respectively. While, the inclination of the maxillary occlusal plane was 6.1±1.3°. Among the 55 patients, 16 were positive for inclination of the occlusal plane, indicating an incidence of 29.1%. The average ANB angle was measured to be -5.1±1.7°. The post-surgical measurement results show that all of the parameters were corrected to normal range. Discussion This is the first article reporting the statistical analysis of extent (volume and length discrepancy) of residual asymmetry on patients with mandibular deviation. Due to the complexity of the mandibular U-shaped morphology, menton deviation is usually accompanied by asymmetry of other parts of the mandible [6]. A correct reference plane is the basis of a comprehensive asymmetry assessment. In our study, an actual clinical sagittal reference plane passing through the facial midline and Me point was created for quantitative assessment of bilateral mandibular asymmetry after the Me point correction with the virtual BSSRO procedure. To ensure a thorough assessment of mandibular asymmetry, the mandible is divided into three anatomical regions according to the characteristics of each region and the site of the surgical incision (Fig 2) [12]. In our previous research, volumetric measurement has been proved as a sensitive method for mandibular asymmetry assessment [13]. As shown in Table 1, the volume of the elongation side in all three mandibular regions was significantly greater than those in the contralateral side, while the volumetric discrepancy in the MB region was greatest. As the major part of the mandible contains the mandibular nerves, the MB region can pose many challenges in achieving symmetry of shape and curvature on the lower edge of the mandible. Our results showed agreement with a previous report demonstrating statistical differences of the surface distance between the ramus, mandibular body and the symphysis area on bilateral sides [14]. The major strength of this study is the revelation of the statistical analysis of residual asymmetry of the mandible, including the location and extent, based on a corrected mirrored reference plane. The differences of transversal and vertical measurement between different regions could guide the specific bony regions modification during the clinical orthognathic surgery. Putting together the results obtained for the deviation in the vertical and transversal directions in the different regions of the mandible, it could be concluded that for the SP region, the discrepancy in the transversal direction on the deviated side was a major cause of asymmetry (1.9 mm transversely and 1.1 mm vertically). This indicates that in clinical practice, for residual asymmetry in the SP region, outer bone cortex grinding [15] (for considerable transversal discrepancy) or splitting corticectomy (for minimal transversal discrepancy) would be a preferable procedure to inferior mandible border corticectomy. On the contrary, for the MB region, asymmetry is mainly caused by vertical discrepancy rather than transversal discrepancy (1.0 mm transversely and 2.2 mm vertically). Therefore, the suitable approach in deviations of the MB region would be full-thickness inferior mandible border corticectomy followed by outer cortex grinding or splitting corticectomy [11]. In the case of the MA region, both vertical and transverse discrepancy contributed to the asymmetry (2.2 mm transversely and 2.2 mm vertically). Hence, the combination of the above surgical methods would be necessary to achieve an ideal outcome in the modification of the MA region. From the result of our post-surgical measurement, it can be seen that a high degree of consistency may be found between virtual BSSRO planning and clinical post-operation outcome, confirming that the residual asymmetry analysis services as an accurate guidance for orthognathic surgery. This study contains weaknesses: one is the complexity of the length measurement index which may hinder its application to manual landmark location. The application of automatic computer measurement of distance discrepancy could make up for this drawback. Conclusion This study supports our finding that the mandibular asymmetry persists after the Me point correction. Me point correction with virtual BSSRO sets up a true reference mirrored plane for asymmetry assessment of bilateral mandibular structure, providing comprehensive planning of orthognathic surgery in patients with mandibular deviation. The variation in the location and extent of mandibular asymmetry would help orthognathic surgeons in making appropriate choices for modification operations along with the regular BSSRO, enabling correction of asymmetry in a single surgery. This would decrease the possibility of follow-up mandibular modification operations. Supporting Information S1 Fig Typical cases of the mandibular residual asymmetry in different regions. the three models in the first row represent the stage before virtual BSSRO surgery (BSSRO with/without Lefort I operation). The second row of the models represent after virtual Me point correction, and mirroring and superimposition of the hemi-mandible. Pink color represents the residual asymmetry between both sides of the mandible. (a) SP region asymmetry: asymmetry mainly exist in the SP region; (b) MB region asymmetry: asymmetry mainly exist in the MB region; (c) MA region asymmetry: asymmetry mainly exist in the MA region. (TIF) Click here for additional data file. S2 Fig Example of a typical case for evaluation method verification. (a) Three dimensional model with pre-surgical orthodontic treatment (before virtual BSSRO surgery); (b) Three dimensional model after virtual BSSRO surgery; (c) Three dimensional post-surgical model. The volumetric differences between both sides of hemi-mandible were calculated to verify the outcome consistency between surgery and virtual surgery. (TIF) Click here for additional data file. S1 Table The mean values of clinical examination after virtual Me point correction and after clinical orthognathic surgery. (DOCX) Click here for additional data file. S1 Raw measurement data (ZIP) Click here for additional data file. ==== Refs References 1 Pirttiniemi PM . Associations of mandibular and facial asymmetries—a review . Am J Orthod Dentofacial Orthop . 1994 ; 106 :191 –200 . 8059759 2 Motta AT , de Assis Ribeiro Carvalho F , Cevidanes LH , de Oliveira Almeida MA . Assessment of mandibular advancement surgery with 3D CBCT models superimposition . Dental Press J Orthod . 2010 ; 15 :45e1 –45e12 . 21738889 3 Yang HJ , Hwang SJ . Evaluation of postoperative stability after BSSRO to correct facial asymmetry depending on the amount of bone contact between the proximal and distal segment . J Craniomaxillofac Surg . 2014 ; 42 :e165 –170 . 10.1016/j.jcms.2013.07.015 24011465 4 Yang HJ , Hwang SJ . Change in condylar position in posterior bending osteotomy minimizing condylar torque in BSSRO for facial asymmetry . J Craniomaxillofac Surg . 2014 ; 42 : 325 –332 . 10.1016/j.jcms.2013.05.021 23816144 5 Xu M , Chan FC , Jin X , Xu J , Lu J , Zhang C , et al Hemimandibular hyperplasia: classification and treatment algorithm revisited . J Craniofac Surg . 2014 ; 25 :355 –358 . 10.1097/01.scs.0000436673.77122.00 24448522 6 Cevidanes Lucia H.S. , Alhadidi Abeer , Paniagua Beatriz , Styner Martin , Ludlow John , Mol Andre , et al 3D Quantification of Mandibular Asymmetry through Cone Beam Computed Tomography . Oral Surg Oral Med Oral Pathol Oral Radiol Endod . 2011 ; 111 : 757 –770 . 10.1016/j.tripleo.2011.02.002 21497527 7 Watson J , Hatamleh M , Alwahadni A , Srinivasan D . Correction of Facial and Mandibular Asymmetry Using a Computer Aided Design/Computer Aided Manufacturing Prefabricated Titanium Implant . J Craniofac Surg . 2014 ; 25 :1099 –1101 . 10.1097/SCS.0000000000000659 24777002 8 Pelo S , Correra P , Gasparini G , Marianetti TM , Cervelli D , Grippaudo C , et al Three-Dimensional Analysis and Treatment Planning of Hemimandibular Hyperplasia . J Craniofac Surg . 2011 ; 22 :2227 –2234 . 10.1097/SCS.0b013e31823200da 22134255 9 Seres L , Varga E Jr, Kocsis A , Rasko Z , Bago B , Varga E , et al Correction of a severe facial asymmetry with computerized planning and with the use of a rapid prototyped surgical template: a case report/technique article . Head Face Med . 2014 ; 10 : 27 10.1186/1746-160X-10-27 25014992 10 Cevidanes LH , Bailey LJ , Tucker GR Jr, Styner MA , Mol A , Phillips CL , et al Superimposition of 3D cone-beam CT models of orthognathic surgery patients . Dentomaxillofac Radiol . 2005 ; 34 :369 –375 . 16227481 11 Shao Z , Peng Q , Xu Y , Yu B . Combined long-curved ostectomy in the inferior mandibular border and angle of the mandible with splitting corticectomy for reduction of the lower face . Aesthetic Plast Surg . 2011 ; 35 :382 –389 . 10.1007/s00266-011-9652-9 21359995 12 Choi KY , Yang JD , Chung HY , Cho BC . Current Concepts in the Mandibular Condyle Fracture Management Part I: Overview of Condylar Fracture . Arch Plast Surg . 2012 ; 39 :291 –300 . 10.5999/aps.2012.39.4.291 22872830 13 Lin H , Zhu P , Lin Y , Wan S , Shu X , Xu Y , et al Mandibular asymmetry: a three-dimensional quantification of bilateral condyles . Head Face Med . 2013 ; 20 :42 . 14 AlHadidi A , Cevidanes LH , Mol A , Ludlow J , Styner M . Comparison of two methods for quantitative assessment of mandibular asymmetry using cone beam computed tomography image volumes . Dentomaxillofac Radiol . 2011 ; 40 :351 –357 . 10.1259/dmfr/13993523 21831974 15 Yoo HS , Choi S , Kim J . Outcome Analysis of Extended, Long, Curved Ostectomy with Outer Cortex Grinding for Prominent Mandibular Angle and Broad Chin to Achieve V-line Contouring . Arch Aesthetic Plast Surg . 2014 ; 20 :80 –84 .

PMC005xxxxxx/PMC5003339.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757139110.1371/journal.pone.0161977PONE-D-16-23684Research ArticleBiology and Life SciencesOrganismsBacteriaXanthomonasBiology and Life SciencesComputational BiologyGenome AnalysisBiology and Life SciencesGeneticsGenomicsGenome AnalysisBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesSequencing TechniquesSequence AnalysisResearch and Analysis MethodsMolecular Biology TechniquesSequencing TechniquesSequence AnalysisBiology and Life SciencesOrganismsBacteriaXanthomonasXanthomonas CampestrisBiology and life sciencesMolecular biologyMolecular biology techniquesSequencing techniquesSequence analysisDNA sequence analysisResearch and analysis methodsMolecular biology techniquesSequencing techniquesSequence analysisDNA sequence analysisMedicine and Health SciencesPathology and Laboratory MedicinePathogenesisMedicine and Health SciencesPathology and Laboratory MedicinePathogensVirulence FactorsPathogen MotilityBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesSequencing TechniquesNucleotide SequencingResearch and Analysis MethodsMolecular Biology TechniquesSequencing TechniquesNucleotide SequencingComparative Genomic and Phenotypic Characterization of Pathogenic and Non-Pathogenic Strains of Xanthomonas arboricola Reveals Insights into the Infection Process of Bacterial Spot Disease of Stone Fruits Comparative Genome Analysis Reveals Virulence Features of X. arboricola pv. pruniGarita-Cambronero Jerson 1Palacio-Bielsa Ana 2López María M. 3http://orcid.org/0000-0002-4314-857XCubero Jaime 1*1 Instituto Nacional de Investigación y Tecnología Agraria y Alimentaria (INIA), Madrid, Spain2 Centro de Investigación y Tecnología Agroalimentaria de Aragón, Instituto Agroalimentario de Aragón-IA2 - (CITA-Universidad de Zaragoza), Zaragoza, Spain3 Instituto Valenciano de Investigaciones Agrarias (IVIA), Moncada, Valencia, SpainMelcher Ulrich EditorOklahoma State University, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: JGC JC AP MML. Data curation: JGC. Formal analysis: JGC JC. Funding acquisition: JC AP MML. Investigation: JC JGC. Methodology: JGC JC. Project administration: JC. Resources: JC AP MML. Software: JGC. Supervision: JC AP MML. Validation: JGC JC AP. Visualization: JGC. Writing – original draft: JGC JC. Writing – review & editing: JGC JC AP MML. * E-mail: cubero@inia.es29 8 2016 2016 11 8 e016197713 6 2016 15 8 2016 © 2016 Garita-Cambronero et al2016Garita-Cambronero et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Xanthomonas arboricola pv. pruni is the causal agent of bacterial spot disease of stone fruits, a quarantinable pathogen in several areas worldwide, including the European Union. In order to develop efficient control methods for this disease, it is necessary to improve the understanding of the key determinants associated with host restriction, colonization and the development of pathogenesis. After an initial characterization, by multilocus sequence analysis, of 15 strains of X. arboricola isolated from Prunus, one strain did not group into the pathovar pruni or into other pathovars of this species and therefore it was identified and defined as a X. arboricola pv. pruni look-a-like. This non-pathogenic strain and two typical strains of X. arboricola pv. pruni were selected for a whole genome and phenotype comparative analysis in features associated with the pathogenesis process in Xanthomonas. Comparative analysis among these bacterial strains isolated from Prunus spp. and the inclusion of 15 publicly available genome sequences from other pathogenic and non-pathogenic strains of X. arboricola revealed variations in the phenotype associated with variations in the profiles of TonB-dependent transporters, sensors of the two-component regulatory system, methyl accepting chemotaxis proteins, components of the flagella and the type IV pilus, as well as in the repertoire of cell-wall degrading enzymes and the components of the type III secretion system and related effectors. These variations provide a global overview of those mechanisms that could be associated with the development of bacterial spot disease. Additionally, it pointed out some features that might influence the host specificity and the variable virulence observed in X. arboricola. http://dx.doi.org/10.13039/100007652Instituto Nacional de Investigación y Tecnología Agraria y AlimentariaRTA2014-0018-002-01http://orcid.org/0000-0002-4314-857XCubero Jaime http://dx.doi.org/10.13039/501100003176Ministerio de Educación, Cultura y DeporteFPU1000/2012Garita-Cambronero Jerson http://dx.doi.org/10.13039/100007652Instituto Nacional de Investigación y Tecnología Agraria y AlimentariaRTA2011-00140-C03-02http://orcid.org/0000-0002-4314-857XCubero Jaime Instituto Nacional de Investigación y Tecnología Agraria y Alimentaria, www.inia.es, Grants: RTA2011-00140-C03-02, RTA2014-00019-C02-01. Fellowship to Jerson Garita-Cambronero: Ministerio de Educación, Cultura y Deporte FPU1000/2012. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Xanthomonas arboricola [1] is a species of Gram negative, rod-shaped bacteria exclusively associated with plants. Most of the strains of this species cause diseases on several herbaceous and woody plants of agricultural interest. Beside these, some other strains have been identified as non-pathogenic, saprophytic or opportunistic pathogens. Based on the host specialization of the pathogenic strains, nine pathovars have been recently proposed [2]. X. arboricola pv. pruni, causal agent of bacterial spot disease on at least 13 species of the genus Prunus, is considered one of the most economically important pathovars within X. arboricola. This pathogen, which is classified as a quarantinable organism in the European Union, causes damages on leaves, fruits, twigs, branches and trunks of the trees [3]. Damages in 25–75% of the peach fruits in orchards in the USA have been reported [4]. Previous molecular typing analysis of a selection of strains isolated from Prunus has demonstrated the low diversity on X. arboricola pv. pruni, which forms a monophyletic group comprised of a unique clonal complex regardless of the host, the continent or the year of isolation, corresponding to a pandemic lineage that stays as a monomorphic group during evolution [2,5]. Next generation sequencing approaches have provided valuable information related to the genomics of the genus Xanthomonas, allowing understanding of the role of several pathogenic factors as well as the key determinants of bacterial adaptation and host restriction [6]. Since 2012, genome sequencing projects have started for X. arboricola, generating at least one complete genome and 15 draft genome sequences, comprising meaningful information for understanding pathogenesis in this species and for the improvement of diagnostic tools addressed to disease prevention. Comparative genomic studies in X. arboricola are attracting increased interest in this species and have started to bring important results. For instance, in a study on X. arboricola pv. juglandis, causal agent of bacterial blight on walnut (Juglans spp.), a group of non-pathogenic strains isolated from J. regia were identified and classified as phylogenetically distant from the pathogenic strains. Complete genome analysis of these atypical strains revealed differences from pathogenic strains in features related to the initial steps of bacterial infection, such as those connected to structural components of the flagellar system, non-fimbrial adhesins and chemosensors. Furthermore, the profile of type III effectors was correlated with the capacity to produce disease on walnut [7,8]. Previous studies of plant-pathogen interaction in other models such as X. citri [9,10], X. campestris [11–13] and X. oryzae [14,15], have demonstrated that the study of phenotypic and genotypic features associated with bacterial sensing and attachment, chemotaxis, motility, xanthan production, biofilm organization, the metabolism of carbon sources, as well as secretion of virulence factors, were basic for improving the knowledge of the bacterial ability to penetrate the plant tissue and to cause disease in a specific host range of plants. The general genome features of two X. arboricola pv. pruni pathogenic strains, isolated from almond (Prunus amygdalus, syn. P. dulcis) and Japanese plum (Prunus salicina), as well as one of a non-pathogenic strain isolated from Santa Lucía SL-64 rootstock (Prunus mahaleb) were described in a previous work [16,17]. Herein, our goal was to associate the genomic content of these xanthomonads with the phenotypic features of X. arboricola pv. pruni that causes bacterial spot of stone fruits. Differences in key aspects associated with bacterial sensing, motility, attachment and secretion of cell-wall degrading enzymes as well as type III effectors were determined in pathogenic and non-pathogenic strains of X. arboricola. Results Molecular characterization using multilocus sequence analysis (MLSA), genome based phylogeny and genome comparison As an initial approach to characterize a group of strains phenotypically similar to Xanthomonas isolated from Prunus in Spain, one Xanthomonas strain isolated from asymptomatic leaves of P. mahaleb (strain CITA 44), and 14 Xanthomonas strains isolated from Prunus spp. with symptoms of bacterial spot disease, were typed at four loci (dnaK, fyuA, gyrB and rpoD) in order to determine their taxonomic position [18] (Table 1). 10.1371/journal.pone.0161977.t001Table 1 Xanthomonas arboricola strains used in this study. Straina Pathovar Host Year of isolation Origin Country (province) X. arboricola CITA 44 Prunus mahaleb (rootstock Santa Lucía SL-64) 2009 Spain (Aragón) ICMP 1488¥ celebensis Musa acuminate 1960 New Zealand CFBP 1846* corylina Corylus avellana 1975 France ICMP 5726¥ corylina Corylus maxima 1939 United States ICMP 35¥ juglandis Juglans regia 1956 New Zealand IVIA 2113* juglandis Juglans regia 2010 Spain (Badajoz) CFBP 3123PT* populi Populus x euramericana cv Robusta 1979 Netherlands 10.0343* pruni Prunus laurocerasus cv. Caucasica 2010 Netherlands 10.0400* pruni Prunus laurocerasus cv. Grüner Teppich 2010 Netherlands 10.0439* pruni Prunus laurocerasus cv. Zabeliana 2010 Netherlands CFBP 3894PT* pruni Prunus salicina 1953 New Zealand CFBP 5530* pruni Prunus persica 1989 Italy CFBP 5724* pruni Prunus amygdalus Unknown United States CITA 4 pruni Prunus persica cv. Catherine 2004 Spain (Valencia) CITA 9 pruni Prunus persica cv. Merrill O'Henry 2008 Spain (Aragón) CITA 11 pruni Prunus persica cv. Richard Lady 2008 Spain (Aragón) CITA 33 pruni Prunus amygdalus cv. Guara 2009 Spain (Aragón) CITA 46 pruni Prunus persica cv. Summer Lady 2009 Spain (Navarra) CITA 70** pruni P. persica x P. amygdalus (rootstock Garnem (GxN15) 2010 Italy IVIA 2626.1 pruni Prunus salicina cv. Fortuna 2002 Spain (Badajoz) IVIA 2647-1-2 pruni Prunus salicina cv. Larry Ann 2002 Spain (Badajoz) IVIA 2647-3-1 pruni Prunus salicina cv. Friar 2002 Spain (Badajoz) IVIA 2826–1 pruni Prunus salicina cv. Anne Gold 2003 Spain (Valencia) IVIA 2832–10 pruni Prunus salicina cv. Angeleno 2003 Spain (Valencia) IVIA 3161–2 pruni Prunus amygdalus cv. Rumbeta 2006 Spain (Alicante) IVIA 3162 pruni Prunus amygdalus cv. Rumbeta 2006 Spain (Alicante) IVIA 3487–1 pruni Prunus salicina 2008 Spain (Aragón) a CITA, Centro de Investigación y Tecnología Agroalimentaria de Aragón, Zaragoza, Spain; ICMP, International Collection of Microorganisms from Plants, Auckland, New Zealand; CFBP, Collection Française de Bactéries Phytopathogénes, Angers, France; IVIA, Instituto Valenciano de Investigaciones Agrarias, Valencia, Spain. PT Pathotype strains. * Strains characterized previously as Xanthomonas arboricola and used in this study as control during the molecular characterization. ** Interception in a Spanish nursey (Aragón) of plants produced in an Italian nursery. ¥ Sequences from these type strains were obtained from a previous work [18], and used as reference for the MLSA in this paper. Analysis of the concatenated sequence (2,858 nucleotide positions) revealed a percent of similarity from 96.86 to 100% within the group encompassed by six X. arboricola reference strains and the 15 strains isolated from Prunus spp. Despite this, strain CITA 44, according to the maximum likelihood analysis, could not be consistently associated with any of the reference strains (Fig 1). The remaining 14 strains isolated from symptomatic hosts were clustered into a group with all the reference strains of X. arboricola pv. pruni. Comparative analysis between these two intra-specific groups revealed a total of 43 nucleotide variations in the concatenated sequence of CITA 44 (11 variable nucleotide sites for dnaK, nine for fyuA, 16 for gyrB and seven for rpoD genes), or a sequence similarity of 98.49% between this strain and all the other X. arboricola isolated from Prunus. These nucleotide changes could be associated with silent mutations, due to the fact that the translated amino acid sequence of each locus showed a 100% of sequence coverage and 99–100% of sequence identity to several strains of X. arboricola according to the Blastp results. 10.1371/journal.pone.0161977.g001Fig 1 Maximum likelihood tree of concatenated nucleotide sequences for partial sequences of the genes dnaK, fyuA, gyrB and rpoD of selected strains of X. arboricola isolated from Prunus. Target strains are in bold. For comparative purposes other X. arboricola strains are included. X. citri subsp. citri strain ICMP 24 was considered as outgroup. Bootstrap values (1,000 replicates) are indicated below and above the branches. Phylogenetic analysis of the 18 X. arboricola strains, based on the core genome sequence (S1 and S2 Tables), showed that the three strains that cause disease on stone fruits and almond (CITA 33, IVIA 2626.1 and MAFF 301420) clustered according to the pathovar classification, as observed previously using the multilocus sequence analysis (MLSA) approach. In the same manner, strain CITA 44 did not cluster with the other xanthomonads isolated from Prunus nor to any of the well-established pathovars described for X. arboricola. Instead, CITA 44 was included in a clade along with the strain 3004 of X. arboricola, recently described as the causal agent of disease on barley (Hordeum vulgare) [19] (Fig 2). 10.1371/journal.pone.0161977.g002Fig 2 Phylogenetic analysis of 18 strains of X. arboricola based on the core genome sequence of the species (2,525 CDS). Sequences were aligned using MAFFT and Maximum likelihood analysis was carried out using RaxML. Bootstrap values (1,000 replicates) are presented above or below the branches. The strain of X. campestris pv. campestris ATCC 33913 was used as an outgroup. After automatic annotation of 18 genome sequences publicly available of X. arboricola [7,16,17,19–23], 3,927 protein coding sequences (CDS) were predicted for CITA 44, whereas in the draft genome sequence of the other X. arboricola strains that did not represent a well-established pathovar (strains 3004, CFBP 7634 and CFBP 7651), a total of 5,221 different CDS were determined. For the strains of the pathovars celebensis, juglandis and pruni, 4,485; 5,700 and 5,048 CDS were predicted, respectively (S1 Table). The core genome sequence of the analyzed strains was comprised by at least 2,525 CDS (S2 Table). CITA 44 shared 3,387 CDS with the three strains of the pathovar pruni (Fig 3A, S1 Fig) and 3,720 CDS with all the strains of X. arboricola (Fig 3B). This strain shared the highest number of CDS with strain 3004 of X. arboricola (3,485 CDS), not classified in any pathovar, whereas the lowest number of CDS was shared with the strain MAFF 301420 of the pathovar pruni (3,181 CDS). Unique CDS for strains CITA 44, CITA 33 and IVIA 2626.1 were predicted. CITA 44 showed 206 exclusive CDS, 55 of which were classified into 18 clusters of orthologous groups (COG) functional categories, being predominantly those related to cell motility and carbohydrate transport and metabolism. Signal peptide cleavage sites and transmembrane helices were predicted only in 19 and 33 CDS, respectively. In the case of the X. arboricola pv. pruni strains, 149 unique CDS were found, 54 of them were classified into 20 COG functional categories, with a predominant presence of CDS related to cell wall/membrane biogenesis and amino acids transport and metabolism. Additionally, 17 CDS showed signal peptide cleavage sites and transmembrane helices were predicted in 33 CDS (Fig 3C, S3 Table). 10.1371/journal.pone.0161977.g003Fig 3 Venn diagram representing the CDS shared among X. arboricola. CDS shared by Prunus-pathogenic strains and CITA 44 (A), CDS shared by CITA 44 and X. arboricola (B). The distribution of the unique CDS found in in the draft genome sequences of CITA 44, CITA 33 and IVIA 2626.1 that have been classified into the COGs functional categories are also represented (C). Different letters represent the standardized code of the COGs functional categories. Carbon sources utilization and chemotaxis profile The profile of carbon sources metabolized by CITA 44 was determined using the BIOLOG GN2 microplate system and compared to that obtained from the strains of the pathovars corylina, juglandis, populi and pruni (Table 1). Seventeen carbon sources compounds over 95 were utilized by CITA 44, meanwhile 45 carbon sources were not utilized and 33 showed variable reactions and, consequently, were considered as not informative. On the other hand, the profile observed in 20 strains isolated from Prunus and identified as X. arboricola pv. pruni, as well as the patterns observed in the remaining strains of X. arboricola, were different compared to CITA 44, as represented in the dendrogram obtained from the similarity analysis (Fig 4). All the strains of the pathovar pruni, unlike CITA 44, were able to metabolize dextrin and proline and unable to metabolize D-saccharic acid. 10.1371/journal.pone.0161977.g004Fig 4 Carbon compounds utilization profile of X. arboricola. Biolog GN2 profile for CITA 44 (A), comparative cluster analysis including carbon compounds profile from X. arboricola isolates (B). Cluster analysis was constructed based on 63 informative substrates. Data were computed using the UPGMA model. Reliability of the tree was determined by the cophenetic correlation index (r > 0.98). In order to determine the chemotactic effect of 18 carbon compounds on CITA 44 and on two reference strains of X. arboricola pv. pruni, a microtiter plate chemotaxis assay was conducted [24]. Strain CITA 44 was attracted only by serine (200 mM), and repelled by citric acid (10 mM), galacturonic acid (10 mM), glucuronic acid (10 mM), arginine (10 mM), cysteine (10 mM) and galactose (10 mM). No chemotactic activity was recorded toward the remaining compounds assayed. Strains CFBP 5530 and CITA 33 showed a different chemotactic pattern compared to CITA 44. The first two strains were attracted to glycerol (2%), maltose (10 mM) and serine (10 mM and 200 mM), and repelled by citric acid (10 mM), galacturonic acid (10 mM), glucuronic acid (10 mM) and cysteine (10 mM). Detailed chemotactic patterns for each strain are shown in Table 2. 10.1371/journal.pone.0161977.t002Table 2 Chemotactic profile of 18 carbon compounds of three representative strains of Xanthomonas arboricola isolated from Prunus spp. Carbon compounds CFBP 5530 CITA 33 CITA 44 Alanine 10 mM + 0 0 Alanine 250 mM + 0 0 Arginine 10 mM 0 - - Arginine 100 mM + 0 0 Citric acid 10 mM - - - Cysteine 10 mM - - - Fructuose 10 mM + 0 0 Galactose 10 mM + 0 - Galacturonic acid 10 mM - - - Glucose 10 mM 0 0 0 Glucuronic acid 10 mM - - - Glycerol 0.2% 0 0 0 Glycerol 2% + + 0 Leucine 10 mM 0 0 0 Leucine 150 mM 0 + 0 Maltose 10 mM + + 0 Mannitol 0.2% 0 0 0 Serine 10 mM + + 0 Serine 200 mM + + + Sodium citrate 10 mM + 0 0 Succinic acid 10 mM 0 0 0 Sucrose 10 mM 0 + 0 Xylose 10 mM 0 0 0 +, Chemotactic activity -, Chemorepellent activity 0, No chemotactic activity In addition to the phenotypic variants observed in the features mentioned above, genome comparison also revealed variants in the profile of the environmental receptors studied within this species. X. arboricola strains harbored 14 of the 28 TonB-dependent transporters (TBDTs) analyzed [25] (S4 Table), and from these, only two homologs to the TBDTs encoded by the loci XAC3620 (outer membrane receptor FepA) and XCC3595 (ferric pseudobactin receptor), described in X. citri and X. campestris, respectively, were shared by all the X. arboricola strains. All strains isolated from Prunus spp., both pathogenic and non-pathogenic, harbored eight TBDTs. The presence of homologous sequences to the TonB-dependent receptor loci XCC1719 and XAC3077, and the absence of XCC0304 and XCC2867 in the strains of X. arboricola pv. pruni, differentiated them from the non-pathogenic strain CITA 44. The distribution of the TBDTs revealed that none of these variations was unique for those strains that inhabit the Prunus hosts. Nevertheless, the TBDT cirA (XAC3077) was found only in those strains that caused disease on stone fruits and Turkish hazel (Corylus colurna). Regarding the distribution of sensors of two-component regulatory system (STCRS) [25], 58 orthologous CDS of 86 genes previously described in Xanthomonas were found in X. arboricola. All the analyzed strains shared 44 STCRS (S4 Table), and CITA 33, CITA 44 and IVIA 2626.1 have 53 STCRS in common. Homologous CDS to a diguanylate cyclase of X. citri (XAC2804) was only present in the pathogenic strains, meanwhile homologues of four STCRS (XAC1819, XAC2804, XAC0136 and XAC1345) were only found in CITA 44. Additionally, as compared to other X. arboricola strains, CITA 33 and IVIA 2626.1 presented two unique STCRS homologous to the loci XAC0136 and XAC1345 of X. citri, respectively (Fig 5). 10.1371/journal.pone.0161977.g005Fig 5 Graphical circular representation of the CDS associated with pathogenesis in X. arboricola isolated from Prunus. The contigs of the draft genome sequence of strains CITA 33, CITA 44 and IVIA 2626.1 were arranged by Mauve [26], using the genome sequence of X. arboricola pv. juglandis Xaj 417 as the reference. The circular map was constructed using CGview. From outside to center: CDS on forward strand, CDS on reverse strand, GC content and GC skew. Out of 26 methyl-accepting chemotaxis proteins (MCPs) previously described [25], 22 were found in the genome sequences of X. arboricola (S4 Table). The number of MCPs varied from 16 in strain CFBP 7634 to 22 in strains 3004, CFBP 7651 and NCPPB 1630; all the X. arboricola strains shared a MCPs pattern composed by 13 genes (S4 Table). Prunus-associated strains shared 20 MCPs, pathogenic strains differed from CITA 44 in the existence of a CDS orthologous to the MCP XCV1933 and the absence of the MCP XCV1938, both described in X. campestris pv. vesicatoria. No MCPs could be detected as specific for a particular host plant, however, orthologous CDS for three MCPs described in X. citri and X. campestris pv. vesicatoria (XAC3768, XCV1952 and XCV1938) were absent only in the strain that causes disease on Turkish hazel (S4 Table). X. arboricola genome sequences harbored 15 to 17 genes homologous to the chemotactic related che genes [25] (S4 Table). Pathogenic strains of Prunus had 17 homologous CDS for these genes, and CITA 44 had 16 CDS. CITA 44 did not show homologous CDS to the locus XAC2447 (cheW) described in X. citri, which was present in all the remaining genome sequences of X. arboricola (Fig 5). Flagella, fimbrial and non-fimbrial adhesins associated with motility and attachment in X. arboricola Swarming motility was evaluated in all the 21 X. arboricola strains isolated from Prunus (Table 3). Two kinds of swarmer colony phenotype were observed. Dendritic pattern, encompassed by several tendrils that extended away from a central colony, was observed in CITA 44 as well as in other 13 strains of the pathovar pruni after 12–15 hours post inoculation (hpi). Despite this, CITA 44 showed some unique characteristic such as the presence of shorter and wider tendrils in the swarming colony (Fig 6A). The second colony morphology was characterized by a circular shape shown by seven strains of X. arboricola pv. pruni (Fig 6A). Light and electron microscopy observations discarded the existence of hyper-flagellated bacterial cells in the swarming colony for all strains, including CITA 44 (S2 Fig). 10.1371/journal.pone.0161977.g006Fig 6 Flagella and type IV pilus mediated motility on semisolid and solid surfaces in some representative strains of Xanthomonas arboricola. Swarming dendritic (A1 and A2) and circular (A3) phenotypes were determined on PYM 0.5% agar plates after 24 hpi. Variable swimming phenotypes (B1, B2 and B3) observed after 72 hpi on MMA 0.3% agar plates. Twitching type motility (C1, C2 and C3) was observed after 72 hpi on polyestyrene plate surface after inoculation through PYM 1.5% agar (40X). Each motility assay was conducted in three independent experiments with three replicates. 10.1371/journal.pone.0161977.t003Table 3 Motility profiles of 21 strains of Xanthomonas arboricola on semisolid surfaces. Strain Pathovar Swarmingα Surfactant activityβ Swimmingγ 100.343 pruni 17.15±6.22* 3.06±0.42 0.58±0.10 100.400 pruni 4.91±2.82* 3.64±0.74 0.55±0.05 100.439 pruni 30.25±14.84* 1.63±0.42 0.55±0.05 CFBP 3894 pruni 4.30±1,14° 1.59±0.48 0.31±0.03 CFBP 5530 pruni 3.32±0,66° 2.87±0.32 0.65±0.13 CFBP 5724 pruni 4.78±1.70° 2.18±0.33 1.25±0.05 CITA 4 pruni 13.47±7.04* 6.62±0.93 0.50±0.05 CITA 9 pruni 17.78±7.66* 1.94±0.56 0.65±0.05 CITA 11 pruni 4.24±1.54* 4.22±1.22 0.55±0.05 CITA 33 pruni 18.72±9.62* 2.80±0.66 0.60±0.05 CITA 44 - 13.42±6.02* 3.69±0.54 2.42±0.10 CITA 46 pruni 22.16±11.91* 8.54±0.16 0.50±0.05 CITA 70 pruni 15.30±4.33* 5.44±0.79 0.52±0.08 IVIA 2626.1 pruni 2.75±0.83° 4.44±0.76 0.30±0.05 IVIA 2647-1-2 pruni 2.32±0.42° 3.94±0.78 0.55±0.05 IVIA 2647-3-1 pruni 23.55±14.67* 3.91±0.64 0.40±0.05 IVIA 2826–1 pruni 17.98±8.23* 3.55±0.45 0.45±0.05 IVIA 2832–10 pruni 21.45±7.41* 2.94±0.52 0.55±0.05 IVIA 3161–2 pruni 3.15±1.04° 2.97±0.73 0.45±0.05 IVIA 3162 pruni 2.71±0.55° 0.52±0.10 0.78±0.08 IVIA 3487–1 pruni 5.89±2.39* 11.98±0.89 0.50±0.00 α Mean ± SD of the area (cm2) covered by the swarmer colony on PYM 0.5% agar plates after 24 hpi * Dendritic swarming phenotypes shown by the analyzed strains. ° Circular swarming phenotypes shown by the analyzed strains. β Mean ± SD of the ratio between the radius of the bright halo and the area of the colony observed 24 hpi after atomizing mineral oil over the bacterial colony. γ Mean ± SD of the radius (cm) of the swimming colony after 72 hpi on MMA 0.3% agar plates. Surfactant activity was evaluated for the 21 strains mentioned above according to the atomized oil assay [27]. A bright halo, associated with a change in the surface tension, was observed around the colonies of all the strains tested. Surfactant production, estimated as the ratio between the radius of the bright halo and the area of the colony, was variable among strains (Table 3). CITA 44 showed a mean ratio of 3.69 ± 0.54 (mean ± standard deviation), meanwhile strains of the pathovar pruni showed a mean ratio of 3.94 ± 2.62. Strains IVIA 3162 and IVIA 3847–1 showed mean ratios of 0.52 ± 0.10 and 11.98 ± 0.89, being the lowest and the highest activity identified, respectively (Table 3). The swimming ability of these strains was also assayed on 0.3% agar MMA plates. After inoculation, all the strains showed a circular and turbid growth around the inoculating point (Fig 6B). This activity, recorded as the radius of the bacterial colony, was variable among the strains (Table 3). CITA 44 presented the highest activity with an average of the radius of the halo of 2.42 ± 0.10, meanwhile the mean for pathogenic strains was 0.65 ± 0.46, with a maximum value of 1.25 ± 0.05 for CFBP 5724 and a minimum value of 0.30 ± 0.05 for IVIA 2626.1. Twitching motility, which is carried out by type IV pilus [28], was observed in all the strains on the plastic surface of the culture plate, after crystal violet staining (Fig 6C). Microscopic observation at the edge of the colony revealed a twitching zone composed by bacterial rafts moving away from the colony, creating a concentric pattern preceded by a lattice like network and microcolonies. Genome analysis performed to characterize the presence of the major structural components of the two motility and adherence structures described above [25,29], revealed that 33 (X. arboricola 3004 and X. arboricola pv. corylina) to 35 (X. arboricola pv. pruni and X. arboricola CFBP 7634 and CFBP 7651) flagellar components were found in X. arboricola. The genomes in these species shared 29 CDS (S4 Table). The strains isolated from Prunus presented 34 (CITA 44) and 35 (CITA 33 and IVIA 2626.1) structural components; CITA 44 did not show homologous sequence to fliD (XAC1974), which is present in all the analyzed strains (Fig 5, S4 Table). Despite the similar pattern of flagellar components observed among the X. arboricola strains, variants in the amino acid sequence of the flagellin protein were observed between pathogenic strains (CFBP 7179, CITA 33, IVIA 2626.1, MAFF 301420 and NCCB 100457), which showed the identical protein WP_039814449.1, and the non-pathogenic or low-pathogenic strains (3004, CFBP 7634, CFBP 7651, NCPPB 1630 and CITA 44), which showed the identical protein WP_024939608.1. Both proteins showed 354 identical amino acids of a total of 399 (pairwise identity of 88.7%). A remarkable difference, associated with the potential to infect in other xanthomonads, is the substitution of aspartic acid by valine in the amino acid position 43 of the N-terminal region of the flagellin gene in the pathovars corylina, juglandis and pruni (S3 Fig). Regarding the structural and regulatory components of the type IV pilus, 22 out of 31 CDS homologous to the components described in X. citri subsp. citri strain 306 [30] were found in the nine genome sequences of X. arboricola; 16 of these CDS were shared by all the strains (S4 Table). The other Prunus-associated strains presented the same components, including CITA 44. The remaining strains of X. arboricola showed a similar pattern, with the exception of NCPPB 1630, CFBP 7634 and CFBP 7651, which did not have homologous sequence to the locus XAC0259 of X. citri. No sequence associated with pilF was observed in strain CFBP 7634 and none homologues of pilQ was found in strain 3004 (S4 Table). Regarding the presence of the genes associated with type IV pilus, that are present in most of the Xanthomonas species, the cluster pilB, C, D, R, S was found in all the strains, as well as the cluster that encodes the minor pilins, pilE, V, W, X, Y1 and fimT, that showed a sequence identity lower than 80.0%. Presence of pilE, pilV, pilY1 and fimT was variable among the strains, meanwhile pilW and pilX were found in all the genome sequences. Prepilin pilA was found in all the strains of X. arboricola, despite its identity to the locus XAC3505 of X. citri was lower than 80.0% (S4 Table). In addition to the fimbrial adhesins described above, the repertoire of non-fimbrial adhesins [25] of X. arboricola comprised five genes (S4 Table). All the strains shared homologous CDS to panB (XAC1816), yapH (XAC2151) and xadA (XCV3670) of X. citri and X. campestris pv. vesicatoria. Prunus-pathogenic strains harbored an adhesin homologous to the locus XAC3672 of X. citri which was absent in CITA44 (Fig 5; S4 Table). Finally, the hemagglutinin encoded by the locus XAC444 of X. citri was found only in those strains that were able to colonize hosts of the genus Corylus, Junglans and Prunus. Pathogenicity tests and genomic components associated with late stages of infection in Prunus-associated strains Detached leaf assay was carried out by inoculating 21 X. arboricola strains, isolated from Prunus spp., on almond (cv. Ferraduel), apricot (cv. Canino), peach (cv. Calanda) and European plum (cv. Golden Japan). CITA 44 did not cause bacterial spot disease symptoms on almond, apricot, peach or plum 28 days post inoculation (dpi). Significant differences (p < 0.05) in the virulence among strains of the pathovar pruni were shown (Table 4). These strains were able to induce disease symptoms on almond, peach and European plum but none of them caused clear disease symptoms on apricot. Representatives of this assay are shown in Fig 7. 10.1371/journal.pone.0161977.g007Fig 7 Schematic representation of bacterial spot symptoms observed on almond, apricot, peach and European plum 28 dpi. Detached leaves were inoculated with a sterile cotton swap damped with 1x108 CFU/mL of bacteria or with 10 mM MgCl2 as negative control and maintained on 0.5% water agar plates. 10.1371/journal.pone.0161977.t004Table 4 Differences in the virulence of 24 Xanthomonas arboricola strains isolated from Prunus on detached leaves of almond, apricot, peach and European plum. Strain Pathovar Almondα Apricot Peach European plum 10.0343 pruni 69.21 ± 12.97b DEF 0a A 79.27 ± 14.13b DE 65.00 ± 20.94b C 10.0439 pruni 55.04± 13.21b BCDEF 0a A 69.09 ± 35.54b CDE 65.23 ± 13.10b C 10.0400 pruni 60.59 ± 7.17b CDEF 0a A 62.82 ± 26.55b CDE 57.35 ± 26.97b BC CFBP 3894 pruni 54.11 ± 13.57c BCDEF 0a A 32.94 ± 29.19bc BC 24.12 ± 33.64ab ABC CFBP 5530 pruni 24.74 ± 20.37b ABC 0a A 76.70 ± 14.45c CDE 57.23 ± 22.960c BC CFBP 5724 pruni 61.65 ± 15.72b CDEF 0a A 60.23 ± 16.91b CDE 46.38 ± 29.31b BC CITA 4 pruni 36.36 ± 27.88b BCDE 0a A 52.80 ± 34.04b BCDE 40.92 ± 31.23b BC CITA 9 pruni 63.09 ± 13.47b DEF 0a A 50.28 ± 32.30b BCDE 42.36 ± 19.17b BC CITA 11 pruni 60.52 ± 18.77b CDEF 0a A 48.18 ± 29.78b BCDE 44.95 ± 20.58b BC CITA 33 pruni 56.30 ± 15.37b BCDEF 0a A 72.96 ± 22.19b CDE 56.87 ± 25.41b BC CITA 44 - 0a A 0a A 0a A 0a A CITA 46 pruni 65.93 ± 11.94c DEF 0a A 77.90 ± 20.09c DE 39.55 ± 22.71b BC CITA 70 pruni 61.62 ± 14.56b CDEF 0a A 42.99 ± 28.09b BCDE 50.83 ± 20.71b BC IVIA 2647-1-2 pruni 51.74 ± 15.80b BCDEF 0a A 73.05 ± 19.87b CDE 55.21 ± 24.10b BC IVIA 2647-3-1 pruni 74.41 ± 22.75c EF 0.03 ± 0.08a A 69.38 ± 20.59c CDE 47.17 ± 18.42b BC IVIA 2826–1 pruni 57.87 ± 16.05c BCDEF 0.60 ± 1.67a A 60.52 ± 26.40c CDE 33.11 ± 18.42b ABC IVIA 2832–10 pruni 65.07 ± 16.01b DEF 0a A 81.49 ± 13.81c E 53.32 ± 14.05b BC IVIA 3161–2 pruni 71.72 ± 16.47b DEF 0a A 75.29 ± 18.70b CDE 40.92 ± 33.80b BC IVIA 3162 pruni 54.62 ± 14.67bc BCDEF 0a A 70.60 ± 28.90c CDE 41.97 ± 28.26b BC IVIA 3847–1 pruni 52.98 ± 33.83b BCDEF 0a A 69.61 ± 12.97b CDE 51.80 ±23.05b BC IVIA2626.1 pruni 75.38 ± 15.99c F 0a A 59.80 ± 25.27c CDE 32.26 ± 26.27b ABC α Mean ± SD of the percentage of symptomatic leaf area 28 dpi using a cotton swab damped with bacterial inoculum. Different capital letters indicate statistically significant differences among the assayed strains on each host. Non capital letters indicate significant differences of each strain on four assayed hosts (p = 0.05). The secretory pathway components, and plant cell wall-degrading enzymes, were also analyzed in X. arboricola genomes (S4 Table). Regarding the encoding elements for the two type II secretory systems (T2SS) described in Xanthomonas [31,32], all strains harbored the 11 genes of the xps cluster but, in the case of the xcs cluster, only CDS homologous to xcsD, xcsE, xcsF, xcsG and xcsJ were found. The remaining seven components of this secretory system were automatically annotated, but amino acid sequence analysis showed for all of them an identity lower than 80% when compared with the xcs cluster described in X. campestris pv. campestris [32]. Variability in the profile of type II secreted virulence factors was found in X. arboricola (S4 Table). A total of 11 pectolytic enzymes [13,33] were found in this species and only the polygalacturonase, encoded by the locus XCC3459, and the rhamnogalacturonan acetylesterase, encoded by XCC0154 in X. campestris, were shared by all the strains. With respect to the strains isolated from Prunus, they shared, in addition, one pectate lyase (XCC2815), one polygalacturonase (XCC2266) and one rhamnogalacturonase (XAC3505). CITA 44 had homologous sequences to a pectate lyase (XCC0112), a pectin methylesterase (XCC0121) and a pectinmethylesterase (XCC2265) that were absent in strains CITA 33 and IVIA 2626.1, meanwhile these two pathogenic strains had a homologous sequence to the degenerated pectate lyase of X. citri (XAC2373) which was absent in CITA 44 (Fig 5). Variation in the profile of cellulolytic enzymes [13] was also found in X. arboricola, 14 of these degrading molecules were present in this species but only nine of them were shared by all the strains. None of the enzymes were unique in those strains isolated from Prunus spp. Pathogenic strains differed from CITA 44 in the presence of a homologous CDS to a beta-glucosidase (XCC1775), and in the absence of the cellulases encoded by the loci XAC3516 and XCC2387, which were present only in CITA 44 (Fig 5, S4 Table). Moreover, 11 CDS homologous to hemicellulolytic enzymes, previously reported for Xanthomonas [13], were found in X. arboricola and eight of these were shared by all the strains (S4 Table). CITA 33 and IVIA 2626.1 presented a xylosidase which was absent in CITA 44 (Fig 5, S4 Table). Those strains of X. arboricola which belongs to the pathovars corylina, juglandis and pruni harbored a xylosidase/arabinosidase enzyme which was absent in all the non-pathogenic strains of X. arboricola and in those strains with a lower pathogenic ability, such as the one described from the pathovar celebensis. Finally, the lipase virulence factor, LipA [34], was found in all the genomes analyzed. The gum gene cluster, associated with the biosynthesis of xanthan in Xanthomonas [35], was analyzed. None of the strains presented homologous sequences to gumG, and CITA 44 and 3004 did not show homologous sequences to the gumF. (S4 Table). Other crucial pathogenic elements analyzed were the T3SS components and the type III effectors [13,33,36–42] in the nine genome sequences of X. arboricola (S4 Table). All the strains harbored CDS that encoded for HpaR2, HpaS, HrpG and HrpX, which are involved in the regulation of the T3SS, and for many of the T3Es and some cell-wall degrading enzymes [43]. However, the strains 3004, CFBP 7634 and CITA 44, as described before [7,17,19], did not present the components of the T3SS that have been found in other Xanthomonas. Besides the master regulons mentioned above, all the remaining X. arboricola strains encompassed and shared 14 hypersensitive response and pathogenicity genes (hpa2, hpaB, hrcC, hrcJ, hrcN, hrcR, hrcS, hrcT, hrcU, hrcV, hrpB1, hrpB2, hrpB5, hrpD6) (S4 Table). The T3E profile of X. arboricola composed of 22 virulence effectors (avrBs2, avrXccA2, hpaA, hrpW, xopA, xopAF, xopAH, xopAI, xopAQ, xopB, xopE2, xopE3, xopF1, xopG, xopK, xopL, xopN, xopQ, xopR, xopV, xopX and xopZ1). The non-pathogenic strain CITA 44 from Prunus, and the pathogen of barley strain 3004, did not have any T3Es. The remaining non-pathogenic strains, CFBP 7634 and CFBP 7651, harbored two (avrsB2 and xopR) and six (avrsB2, hpaA, hrpW, xopA, xopF1 and xopR) T3Es, respectively [7] (S4 Table). In addition, the pathogenic strain NCPPB 1630 from Musa sp. showed six T3Es, the pathogenic strain CFBP 7179 from J. regia showed 16 T3Es and the strain NCCB 100457, that causes disease on C. colurna, showed 19 T3Es. X. arboricola pv. pruni strains CITA 33 from P. amygdalus and IVIA 2626.1 from P. salicina harbored 21 T3Es. All the strains, regardless of their hosts, shared six T3Es (avrBs2, hpaA, hrpW, xopA, xopF1 and xopR). The effector xopB was unique in the strain that causes disease on Juglans, meanwhile xopAQ, which has not been previously reported in X. arboricola, and xopE3, were only found in those strains that cause disease on Prunus. CDS homologous to these two T3Es were specifically present in the plasmid pXap41, which is unique in the strains of the pathovar pruni [44] (Fig 8). 10.1371/journal.pone.0161977.g008Fig 8 Nucleotide sequence similarity among the sequence of the exclusive plasmid of X. arboricola pv. pruni pXap41 and the draft genome sequences of nine strains of X. arboricola. Blastn was used for the comparative sequence analysis with an expected value threshold of 0.001. Circular graphic representation was constructed using the BLAST Ring Image Generator (BRIG) tool. See legend for information related to each ring of the map. Discussion MLSA has been shown to be useful to characterize strains of X. arboricola [2,5,18,45,46]. In this work, we utilized a MLSA scheme proposed by Young and collaborators [18], which is based in the analysis of partial sequences of the genes dnaK, fyuA, gyrB and rpoD. Here, we have concluded that for 14 X. arboricola strains isolated from Prunus, the selected MLSA scheme was a good approach to characterize and discriminate the members of the pathovar pruni from atypical or commensal strains of X. arboricola. Classification of CITA 44 as an atypical strain of X. arboricola was corroborated by the phylogenetic analysis conducted with 2,525 genes identified as the components of the core genome of this species. This atypical strain isolated from Prunus did not present the same phylogenetic origin as the pathogenic strains classified as pathovar pruni, but it was more similar to strain 3004 of X. arboricola, which did not cluster within any of the well-established pathovars of this taxon. Even though CITA 44 was closely related to 3004, this strain did not produce symptoms on inoculated barley in assays performed in our group. This result is similar to a previous work on X. arboricola strains isolated from walnut that were more similar to strains isolated from Musa sp. than to those of the pathovar juglandis isolated from walnut [7]. Moreover, the study of these atypical strains, found on barley, walnut and here in P. mahaleb, open the discussion about the origin and evolution of pathogenicity in this species as it has been proposed previously for the genus Xanthomonas [43]. Based in the current data, it is not possible to determine if these non-pathogenic strains were predecessors of the pathogenic groups or the result of the loss of their pathogenicity. Further exploratory studies regarding some other Prunus-associated strains with other phenotypic and genotypic variants are needed to elucidate the evolutionary process of pathogenicity in this bacterial species. In addition, this study contributes to our understanding of the diversity of X. arboricola associated with stone fruit trees and almond. Moreover, it also provides information to develop tools to identify and discriminate pathogenic and non-pathogenic Xanthomonas strains found in Prunus spp. This precise characterization of atypical strains of X. arboricola will avoid bacterial identification mistakes, as occurred sometimes in other bacterial models which implied unnecessary control measures and resulted in high economic losses [47,48]. The phylogenetic variation among strains of X. arboricola isolated from Prunus concurred with the existence of differences in several phenotypic features including pathogenicity. Additionally, genomic information generated from xanthomonads isolated from Prunus [16,17] and from other X. arboricola [7,19–23], has been used to reveal those features which could be involved in the disease process. Carbon source utilization profiles, as expected [1,2,5], showed high homogeneity among strains of the pathovar pruni. Nevertheless, the profile of carbon sources utilization shown by CITA 44 was different to the one presented by the other strains as well as the one provided in the description of X. arboricola species [1]. This disparity from the original metabolic description of the species has been also described for some strains of other Xanthomonas species, such as X. vesicatoria [49] and X. campestris pv. campestris [50]. The discordant profile of CITA 44 could be due to the fact that the original description of the species X. arboricola was based on strains of seven pathogenic pathovars which presented a high intra-pathovar homogeneity [1,2] and did not consider a wider ecological diversity including the atypical non-pathogenic strains currently of great research interest. Initial stages of bacterial adaptation and host colonization encompass a series of sensors and receptors that detect stimuli, providing to the cells the information of their biotic and abiotic environment which triggers a series of processes such as the cell motility, chemotaxis, quorum sensing, biofilm formation and many other cellular events [15,51]. At this stage, a group of protein complexes denominated TBDTs, STCRS and MCPs play a crucial role [25]. The repertoire of TBDTs in X. arboricola, which are bacterial outer membrane proteins associated with the transport of different substrates including carbohydrates [52], was extensive compared to other species of this genus [25,33], resulting similar to the one observed in X. campestris and other epiphytic xanthomonads [25]. Additionally, this repertoire is in accordance with the one previously observed in two not publicly available genome sequences of X. arboricola pv. fragariae (strains LMG 19145 and LMG 19146) and one of X. arboricola pv. pruni (strain LMG 25862) [33]. This high number of TBDTs has been associated in other xanthomonads with the need for carbohydrates scavenging in variable conditions encountered in epiphytic niches [53]. Bacterial sensing and chemotaxis are essential components of the initial infection processes [54]. Regarding STCRS proteins, which are part of the dominant molecular mechanism by which unicellular organisms respond to environmental stimuli, the large repertoire observed in Xanthomonas arboricola was similar to that observed in X. campestris [25], and it was in accordance with the number of STCRS observed in most of the complete genome sequences of Xanthomonas, with the exception of those species which inhabit in restricted niches such as X. albilineans [15,51]. Evaluation of TBDTs and STCRS content revealed differences between CITA 44 and X. arboricola pv. pruni strains, despite of the general low variability observed among the different X. arboricola strains evaluated. Beside this, CITA 44, compared to strains of the pathovar pruni, also showed a distinct chemotactic pattern, being influenced by just a few compounds. This kind of intraspecific chemotactic variability has been also observed in other xanthomonads models like X. citri subsp. citri with dissimilar pathogenic ability [24]. In order to explain these chemotactic behaviors, MCP content was evaluated. MCPs sense beneficial or toxic environmental compounds and transduce the signal to the cytoplasm by CheW protein, causing changes in the flagella direction and rotation speed, which finally guides bacterial cells to favorable environments [54,55]. Here, we have found that the core repertoire of the MCPs in nine genome sequences of X. arboricola included at least 13 chemoreceptors. Differences in the MCPs content were observed among the genome-sequenced strains and could be associated with their host range. Differences in MCPs content were found between CITA 44 and other xanthomonads from Prunus spp. Moreover, a variation in the cheW locus was also identified. Although functional analyses are necessary to confirm the role of these sensors at initial stages of the infection process, our results point out that these processes and the genes involved may mark the diverse behavior of the different strains. Moreover, the very initial stages of the bacterial-host interaction described above could trigger other molecular routes which involve components such as the flagella and the type IV pilus, that are related not only to motility on liquid or solid environments, but also with attachment to the host or the development of biofilm structures [30,56,57]. Motility on solid and semisolid surfaces, which are controlled by flagella or type IV pilus, was also variable among the assayed strains of X. arboricola isolated from Prunus. This variability among different strains in swimming, swarming and twitching motility have been shown in other xanthomonads such as X. citri [58], as well as in X. arboricola strains isolated from walnut [7]. Strain CITA 44 showed the higher ability to swim that maybe connected to a more restricted niche to survive and higher requirements to locate it. Similarly, this enhanced ability to swim has been described in other bacterial models like X. citri subsp. citri, which less virulent strains were described to have a higher swimming ability [58]. In relation to the surface motility in X. arboricola, two different colony phenotypes were observed; the circular one matched with the one observed for other species such as X. oryzae and X. citri [59,60], described previously as independent of flagella and defined as sliding type motility instead of swarming [57]. Contrary to this, some other strains showed dendritic swarmer colonies, which had not been previously described in Xanthomonas but confirmed in other bacterial species such as Pseudomonas aeruginosa as a real swarming type motility [61]. In addition, these strains showed other swarming-related features such as the production of surfactants and a rapid outward migration [62]. Again, CITA 44 showed a different pattern to that of X. arboricola pv. pruni, presenting a colony with an intermediate phenotype between dendritic and circular and also an intermediate surfactant production. Genome analysis of the flagellum components in CITA 44 revealed an interesting amino acid substitution in the N-terminal region of the flagellin that was also found in other non-pathogenic xanthomonads. Previously, Cesbron and collaborators [7] reported an amino acid polymorphism (Asp-43/Val-43) in the flagellin domain flg22 among pathogenic and non-pathogenic strains of X. arboricola isolated from walnut. Here we report that this variation is not only present in X. arboricola pv. juglandis strains but it is also found in the other two major pathogenic pathovars of this species, corylina and pruni. In X. campestris, strains with Val-43 in flg22 were not detected by the flagellin sensing 2 kinase (FLS2) of Arabidopsis, therefore they did not elicit the pathogen-associated molecular (PAMP)-triggered immunity and, additionally, these strains were more virulent than those with an Asp-43 residue in flg22 [63]. The genomic analysis also revealed the absence of fliD in CITA 44. In several bacterial species the absence of fliD was associated with non-flagellated and non-motile cells [64,65]. This is not the case of CITA 44, which showed a single polar flagellum like X. arboricola pv. pruni strains [17]. The type IV pilus is an important structure related to the movement across the surface, adhesion, microcolony formation, secretion of proteases and colonization factors, being a key pathogenesis factor [66]. In general, strains isolated from Prunus, despite their pathogenicity, harbor all the four subcomplexes that permit the biogenesis and function of this structure [30]; additionally these components were demonstrated as functional in the twitching motility assay. As the type IV pilus, non-fimbrial adhesins are involved in bacterial attachment to host surfaces; in X. oryzae, they play an important role in pathogenesis and are particularly involved at the initial stage of leaf attachment and penetration into the host [67,68]. X. arboricola, shows various combinations of adhesins among the different strains studied. This variability could be associated to the different bacterial-plant compatible interactions of the different strains and hosts [69]. Once more, CITA 44 lacked one of the genes involved in adhesion synthesis which was present in X. arboricola pv. pruni. Further functional studies in this way could be valuable to define the definitive role of these adhesins in the pathogenicity and host specificity observed in the pathovars of X. arboricola. After the very initial stages of pathogenesis described above, some virulence genes related to host colonization, multiplication and development of symptoms are expressed. T2SS permits the export of proteins from the bacterial cell and it is involved in the translocation of degradative enzymes which causes damage to the host cells and tissues [70]. Most species of Xanthomonas encode the components of two T2SS. One of them is Xps, which contributes to bacterial virulence by the secretion of xylanases and proteases in X. campestris and X. oryzae [32]. In X. arboricola, all the components of the xps gene cluster were conserved. As in all the Xanthomonas species, X. arboricola showed a core repertoire of cell wall degrading enzymes which is formed by at least 30 CDS. Beside this, a high variation in the repertoire of these enzymes was also found among the analyzed strains. These variations could be related to the different cell wall composition of the hosts or tissues and the requirements to produce symptoms during the infection [69]. Differences in the profile of degrading enzymes were identified among the Prunus-pathogenic strains and other strains of X. arboricola, including CITA 44. Another gene cluster associated with pathogenesis in Xanthomonas is the xanthan producing gum gene cluster which is composed by 12 genes. Deletion of this group of genes inhibits the disease development in X. campestris on Arabidopsis and Nicotiana [71]. In X. arboricola, this cluster was conserved, with the exception of gumG that was absent in all the sequenced strains and the O-acetyltransferases encoded by gumF lacking in strains CITA 44 and 3004. Nevertheless, the absence of these two genes has been demonstrated to not affect the xanthan polymerization [72]. T3SS, as well as the T3Es, plays an essential role in pathogenicity once bacteria have penetrated the host tissue and could be related to host specificity in Xanthomonas [37]. As occurred in other Xanthomonas species, all the X. arboricola strains harbored orthologs CDS to the genes that encode HpaR2, HpaS, HrpG and HrpX, which are involved in the regulation of T3SS, T3Es as well as other pathogenic factors such as those encoded by pehA and pehD genes [43]. This was also previously reported in the strains isolated from walnut as well as for strains 3004, CITA 44 and IVIA 2626.1 [7,17,43]. It is important to note the localization in X. arboricola of an ortholog to the T3E, called xopAQ, in the plasmid pXap41, next to other two virulence associated proteins, xopE3 and mltB [44]. This effector has been previously found in the chromosome of X. citri and X. gardneri and a similar sequence has been found in Ralstonia solanacearum [40,73]. Nucleotide sequence analysis revealed a high pairwise nucleotide identity (99.7%) to the xopAQ sequence of X. citri subsp. citri Aw12879, as well as a putative plant-inducible promoter box (PIP-box) sequence 67 bps upstream the start codon, but this conserved sequence associated with the regulon HrpX showed a variant in one nucleotide (S4 Fig). The presence of at least three T3Es in pXap41 reinforces the hypothesis that this plasmid may contribute to the virulence of this pathovar as well as to the specialization of the pathovar pruni towards hosts of the Prunus genus [44]. CITA 44 did not show xopAQ or any of the T3Es found, which is consistent with the absence of pXap41 and its non-pathogenic character. Currently, functional studies on the effect of pXap41 and its related T3Es are being conducted in our group to confirm their role in the pathogenicity and the host specificity of the pathovar pruni strains. In conclusion, this study of strains with diverse virulence range, and particularly strain CITA 44, has provided an opportunity to elucidate essential mechanisms in the host-bacteria interaction in Xanthomonas arboricola pv. pruni. In addition, it is useful to understand the evolution of the pathogenicity in this species. Furthermore, the comparative analysis reported here highlights several differences regarding to key genotypic and phenotypic features of initial and late stages of the infection process. The different TBDT, STCR, MCP profiles, the genomic variation in the components of flagellin, the type IV pilus and in cell-wall degrading enzymes repertoire, and the alteration in main virulence factors provide information that contributes to explain the phenotypic differences observed between pathogenic and non-pathogenic strains. Although the work presented shows a global overview of mechanisms involved in virulence, further functional studies are needed to test and improve the understanding of the role of all the virulence-related features mentioned here. Materials and Methods Bacterial strains and culture conditions X. arboricola strain CITA 44, isolated from asymptomatic leaves of P. mahaleb in the Spanish region of Aragón, and fourteen bacterial strains of X. arboricola pv. pruni isolated from trees showing bacterial spot symptoms in Spain and Italy, were used in this study (Table 1). From these fourteen strains, three were isolated from almond [Prunus amygdalus Batsch, syn. P. dulcis (Miller) D. A. Webb], four from peach [Prunus persica (L.) Batsch], six from Japanese plum (Prunus salicina Lindley) and one from the hybrid rootstock Garnem (GxN15) (P. persica x P. amygdalus). Reference strains of X. arboricola pv. pruni (10.0343, 10.0400, 10.439, CFBP 3894, CFBP 5530, CFBP 5724), X. arboricola pv. corylina (CFBP 1846), X. arboricola pv. populi (CFBP 3123) and X. arboricola pv. juglandis (IVIA 2113) were also included in this study (Table 1). Strains were routinely cultured on 1.5% agar plates of Luria Bertani (LB) or in LB broth at 27°C for 48 h. All Xanthomonas strains are conserved in the collections of Centro de Investigación y Tecnología Agroalimentaria de Aragón (CITA, Aragón, Spain) and Instituto Valenciano de Investigaciones Agrarias (IVIA, Valencia, Spain) and no specific permission was required for their use. Multi Locus Sequence Analysis (MLSA) Bacterial DNA was extracted from cultures in LB broth obtained after 24 h incubation at 27°C, using a QIAamp DNA miniKit according to the manufacturer's instructions (Qiagen, Barcelona, Spain). DNA was used for PCR or stored at -20°C until further use. Degenerate primers, previously determined as useful in MLSA analysis conducted in Xanthomonas [18,46], were used for PCR amplification of partial sequences of the housekeeping genes dnaK, fyuA, gyrB and rpoD. PCR amplifications were carried out in a 50 μL volume containing 1X PCR buffer (10 mM Tris-HCl, 50 mM KCl, 0.1% Triton X-100 [pH 9.0]); 0.2 μM of each primer; 1.25 U Taq DNA polymerase (Biotools, Madrid, Spain); 0.2 mM each dNTP (Biotools Madrid, Spain); 1.5 mM MgCl2 and 1.0 μg/μL of DNA template. All PCR reactions were performed in an ABI 2720 thermal cycler (Applied Biosystems, Foster Urban district, CA, USA) with an initial denaturation at 94°C for 5 min, 40 cycles of denaturation at 94°C for 1 min, annealing at 55°C for 1 min and extension at 72°C for 2 min, and a final extension step at 72°C for 10 min. PCR products were visualized under UV light in 2% agarose gels stained with ethidium bromide and purified with the Wizard SV Gel and PCR Clean-up System Kit (Promega Corporation, Madison, USA). PCR products were sequenced at STAB VIDA (Lisbon, Portugal), and edited using BioEdit Sequence Alignment Editor [74]. Additionally, sequences of the housekeeping genes used for MLSA analysis from X. arboricola pathovars celebensis (ICMP 1488), corylina (ICMP 5726) and juglandis (ICMP 35) as well as X. axonopodis pv. citri strain ICMP 24, included as outgroup, were obtained from the National Center for Biotechnology Information database (NCBI) (http://www.ncbi.nlm.nih.gov). Nucleotide sequences were aligned with ClustalW version 1.83 [75] using default parameters. Both ends of each alignment were trimmed to the following sizes: dnaK, 864 positions; fyuA, 607 positions; gyrB, 631 positions and rpoD, 756 positions. Nucleotide sequences from all strains analyzed here and from those Xanthomonas spp. available in databases [18] were aligned and concatenated to give a total length of 2,858 nucleotide positions. The programs jModelTest 0.1.1 [76] and MEGA 5.05 [77] were used to determine the best model of evolution for Maximum Likelihood analysis (ML) based in the akaike information criterion (AIC) [78]. For the concatenated gene dataset the model selected was TN93+G. Maximum likelihood trees, using 1,000 bootstrap re-samplings, were generated in MEGA 5.05. Nucleotide sequences were deposited in GenBank. Accession numbers for the partial sequences of the genes used in this study are: KR054426 to KR054449 for dnaK; KR054450 to KR054473 for fyuA; KR054474 to KR054497 for gyrB and KR054498 to KR054521 for rpoD. Genomic Analysis Based in the MLSA analysis, three X. arboricola strains, CITA 33 (Synonymy Xap 33), CITA 44 and IVIA 2626.1, were selected for complete genome sequencing and genome features for these strains have been previously announced [16,17]. Nucleotide sequences of the draft genome sequences of these three strains (Deposited at DDBJ, EMBL, GenBank under the accession numbers JHUQ00000000 for CITA 33, LJGM00000000 for CITA 44, and LJGN00000000 for IVIA 2626.1), as well as nucleotide sequences of three strains of X. arboricola (3004, CFBP 7634 and CFBP 7651) [7,19], two strains of X. arboricola pv. celebensis (NCPPB 1630 and NCPPB 1832) [23], one strain of X. arboricola pv. corylina (NCCB 100457) [20], eight strains of X. arboricola pv. juglandis (CFBP 2528, CFBP7179, NCPPB 1447, XajA3, Xaj2, Xaj4-1, Xaj43a and Xaj417) [7,21,22] and one strain of X. arboricola pv. pruni (MAFF 301420) were obtained from the NCBI´s genome database (S1 Table). Genome scaffolds presented in each one of the bacterial genomes mentioned above were arranged and oriented by Mauve [26] using the complete genome sequence of X. arboricola pv. juglandis strain Xaj417 [22] as reference. For the purpose of homogeneity for further comparison, all the genome sequences were annotated using Prokka [79]. GFF3 archives generated by Prokka were used as the input to determine the core and the dispensable accessory genes in the analyzed genomes using Roary [80]. The online tool available at bioinformatics.psb.ugent.be, was used to generate the Venn diagram to compare the genome composition of the Prunus-isolated strains with the remaining subspecific groups of X. arboricola. Signal peptides and transmembrane domains for the unique protein coding sequences (CDS) of X. arboricola CITA 44 and X. arboricola pv. pruni CITA 33 or IVIA 2626.1, were determined using the signalP 4.1 server and the TMHMM server version 2.0 [81,82]. Beside this, the assignment of these genes to the clusters of orthologous groups (COG) database [83] was performed with the NCBI`s conserved domain database using an expected value of 0.001 [84]. The core genome sequence obtained for each strain was aligned using MAFFT [85] for further phylogenetic analysis. Subsequently, a maximum likelihood tree, using 1,000 bootstrap resamplings, was constructed to accurately determine the phylogenetic position of the atypical strain isolated from Prunus, CITA 44, within X. arboricola. Maximum likelihood tree was carried out using the RaxML tool [86] and the dendrogram obtained was visualized using Dendroscope [87]. X. campestris pv. campestris strain ATCC 33913 was used as an outgroup in the analysis. Carbon sources utilization analysis and environmental sensors profile To prepare the inocula, bacterial strains used in the MLSA analysis (Table 1) were cultured in LB plates and resuspended in sterile phosphate-buffered saline (PBS) (OD600 = 0.3). Suspensions (150 μL) were inoculated into each well of the Biolog GN2 microplates (Biolog Inc., USA). Microplates were read at 570 nm using a Labsystems Multiskan RC spectrophotometer (Fisher Scientific, Walthman, USA) after 24 h incubation at 27°C [1,88]. Three independent assays were performed, each including two microplates per strain and three reads per well. Means from all the reads were performed and used in further analysis. Three substrate utilization levels were defined based in the percentage of utilization referred to water control [89]: positive (+, > 160%); borderline, considered as non-informative (±, 130–160%) and negative (-, < 130%). The carbon sources utilization profile of strain CITA 44 was determined and compared with those obtained for strains of the pathovars corylina, juglandis, populi and pruni; results were converted to a binary form by scoring the metabolism observed for each compound as 0 (no catabolism of the carbon source) and 1 (catabolism of the carbon source). Similarity for pairs was calculated according to the Jaccard´s coefficient and was subjected to Unweight Pair Group Method with Arithmetic Mean (UPGMA) cluster analysis. Finally, the reliability of the similarity trees was determined using the cophenetic correlation index. All the analysis were computed on NTSYS 2.11T (Exeter Software, Setauket, NY). Profiles of sensors of two-component regulatory system (STCRS) and TonB-dependent transporters (TBDTs) for X. arboricola strains isolated from Prunus (CITA 33, CITA 44 and IVIA 2626.1) and six other Xanthomonas (3004, CFBP 7634, CFBP 7651, NCPPB 1630, NCCB 100457 and CFBP 7179) (S1 Table) were determined based in the analysis of the complete genomes and the search of homologous sequences for 86 STCRS and 28 TBDTs previously described in other Xanthomonas species [25]. Sequence homology searches were conducted using the Blast tool according to the protocol proposed by the NCBI [90], using the nucleotide sequence that encodes each one as the input. Those CDS which presented an identity and a query coverage percentage over 80% were considered as orthologous genes in the target genomes of X. arboricola. Chemotaxis assay and repertoire of methyl accepting chemotaxis proteins Eighteen carbon compounds that induce variation in the chemotactic response in other Xanthomonas [24] were evaluated to determine differences in the chemotactic profile between the atypical strain CITA 44 and X. arboricola pv. pruni strains CITA 33 and CFBP 5530, for which complete genome sequence had been already obtained [16,17,91]. The chemotactic compounds tested were: alanine (10 and 250 mM), arginine (10 and 100 mM), citric acid (10 mM), cysteine (10 mM), fructose (10 mM), galactose (10 mM), galacturonic acid (10 mM), glucose (10 mM), glucuronic acid (10 mM), glycerol (0.2 and 2%), leucine (10 and 150 mM), maltose (10 mM), mannitol (0.2%), serine (10 and 200 mM), sodium citrate (10 mM), succinic acid (10 mM), sucrose (10 mM) and xylose (10 mM). The chemotactic effect of the carbon compounds was evaluated according to a microtiter plate chemotaxis assay previously developed [24]. Briefly, 10 μL tips containing 5 μL of the carbon source tested were inserted into 48 of a 96 wells plate previously inoculated with 200 μL of bacterial suspension in 10 mM MgCl2 (108 CFU/mL). The number of bacteria that moved into the tip during one hour was estimated by means of serial dilutions of the tip content plated on 1.5% agar LB plates. Significant differences (p < 0.05) between the means from each carbon source tested and the negative control (10 mM MgCl2) were determined by an analysis of variance (ANOVA) according to the Student-Newman-Keuls method. Statistical analyses were performed using STATGRAPHICS Plus v.5.1 (Manugistics Inc. Rockville Maryland, USA). A carbon source was considered as a chemoattractant when the average number of bacteria contained in the tip (6 replicates in 2 independent assays) was significantly higher to the blank control (10 mM MgCl2), and considered as a chemorepellent when the average was significantly lower (p < 0.05). Orthologous genes for the 26 methyl accepting chemotaxis proteins (MCPs) as well as for 18 specific-chemotaxis che genes [25], described in X. campestris pv. campestris, X. campestris pv. vesicatoria, X. citri subsp. citri and X. oryzae, were searched in the genome sequences of nine X. arboricola strains as mentioned above. Those CDS with more than 80% of identity and more than 80% of sequence coverage containing MCPs periplasmic domains were considered as orthologous genes. Motility in solid and semisolid surfaces and surfactant activity Swarming, swimming and twitching motility assays were conducted for strain CITA 44 as well as for 20 strains classified as X. arboricola pv. pruni by MLSA. Bacterial cultures in exponential growth phase were centrifuged at 6,350 g for 15 min, then washed and resuspended in 10 mM MgCl2 (OD600 = 1.0). To analyze swarming motility, 0.5% agar PYM swarming plates (peptone 0.5%, yeast extract 0.3%, malt extract 0.3%, glucose 1.0%) were inoculated with 10 μL of bacterial suspension. For swimming motility assay, cultures resuspended in 10 mM MgCl2 were centrifuged at 6,350 g during 15 min and the pellets were inoculated using a sterile toothpick in the center of semisolid 0.3% agar minimal medium A (MMA) plates (K2HPO4 0.7%, KH2PO4 0.3%, MgSO4·7H2O 0.01%, (NH4)2SO4 0.1%, sodium citrate 0.005%, glycerol 0.2%). For twitching assay, bacterial cultures were prepared as mentioned above, and then inoculated with a sterile toothpick through a 5 mm PYM 1.5% agar layer to the bottom of the plate. After 72 h, culture medium was removed and the bottom of the plate was stained with 0.3% crystal violet for 15 min and washed using sterile distilled water. Surfactant production, which is closely related to bacterial motility on solid surfaces, was also assessed according to the atomized oil assay [27]. Briefly, bacteria were inoculated using a sterile toothpick on 1.5% agar LB plates, and after 24 hpi, a fine mist of mineral oil droplets were sprayed on the colony. Bacterial strains that instantaneously showed a bright halo around the colony were recorded as surfactant producers. Plates from swarming motility and surfactant activity assays were incubated at 27°C during 24 h, whilst swimming and twitching plates were incubated for 72 h. Images from all the plates were recorded. All assays were performed in three independent experiments with three replicates each time. Homologous CDS to the major structural components associated with flagellum [25,29,92] as well as the fimbrial adhesin type IV pilus [30] and a group of non-fimbrial adhesins [25] were identified in the three genomes of X. arboricola isolated from Prunus and in other six X. arboricola strains as described above. Pathogenicity tests on Prunus spp Bacteria were grown on 1.5% agar LB plates and incubated at 27°C for 48 h, then a single bacterial colony was resuspended in 30 mL of LB broth and incubated at 27°C on a rotary shaker for 24 h. After incubation, cultures were centrifuged at 6,350 g for 15 min and washed three times with 10 mM MgCl2. Finally, cultures were adjusted to 108 CFU/mL (OD600 = 0.1) and used to inoculate detached leaves of different Prunus species. Young, fully expanded leaves from greenhouse-grown plants of almond (cv. Ferraduel), apricot (cv. Canino), peach (cv. Calanda) and European plum (cv. Golden Japan) were collected, brought to the laboratory and washed three times in sterile distilled water. After surface sterilization with 70% ethanol and 0.05% sodium hypochlorite, the leaves were rinsed three times with sterile distilled water and dried on a hood. Three leaves per host were selected randomly for being inoculated with each strain in three independent assays. Surface sterilized leaves were placed on 0.5% water agar plates and the abaxial surfaces were gently inoculated using a sterile cotton swab damped with the bacterial inoculum. Sealed plates were incubated in a grow chamber adjusted at 27°C, 80–90% relative humidity and 12 h photoperiod. Inoculated leaves were digitally recorded at 28 dpi and the percentage of symptomatic area per host and strain was quantified using the ImageJ 1.45s software. Samples, inoculated with sterile 10 mM MgCl2, were used as blank control for comparison. Additionally to the genomic analysis related to the early events in the bacteria-host interaction described below, homologous CDS to several genes associated with later stages of pathogenesis, like the type III secretory system (T3SS) and the type III effectors (T3Es), were searched in the genome of the nine strains of X. arboricola according to the percentage of the sequence length and identity with the amino acid sequence of the genes previously associated with these features [36–41,93,94]. In the same manner, orthologous CDS to other remarkable processes and features associated with virulence such as the quorum sensing [95], the xanthan biosynthesis [35], the type II secretion system [31,32] as well as the cellulolytic, hemicellulolytic, pectolytic and lipases enzymes, previously described in other species of Xanthomonas [13,33,34,96], were searched in the genome sequences of the nine strains of X. arboricola. Only those CDS with a percentage of coverage and identity over 80.0% were considered as orthologous sequences. Finally, the presence of the plasmid pXap41 [44], putatively involved in virulence in X. arboricola pv. pruni, was searched in the analyzed genomes based in the nucleotide sequence similarity and graphically represented using the BLAST Ring Image Generator (BRIG) tool [97]; blastn was used for the sequence comparative analysis with an expected value threshold of 0.001. All the CDS associated with pathogenesis, mentioned above, were represented in a circular genome map using CGView [98]. For this purpose, contigs of the draft genome sequence of CITA 33, CITA 44 and IVIA 2626.1 were arranged by Mauve [26] using the circularized genome sequence of X. arboricola pv. juglandis strain Xaj 417 as the reference [22]. Supporting Information S1 Fig Nucleotide genome sequence comparison of X. arboricola pv. pruni strains CITA 33 and IVIA 2626.1 and the non-pathogenic strain CITA 44. Genome sequences were compared against each other based on Blastn results and represented as a circular map using the CGview tool. From outside to center: strain CITA 33, strain IVIA 2626.1, strain CITA 44, GC content, GC skew + and GC skew -. (TIF) Click here for additional data file. S2 Fig Transmission electron microscopy of X. arboriola strain CITA 44. Negative staining showed monoflagellated cell from the edge of the dendritic swarming colony 24 hpi in 0.5% PYM agar plates. (TIF) Click here for additional data file. S3 Fig Amino acid alignment of two variants of the FliC found in X. arboricola. Sequence alignment, performed using ClustalW, showed the amino acid change in the position number 43 of the amino terminal region of FliC. (TIF) Click here for additional data file. S4 Fig Nucleotide alignment of type III effector xopAQ and the PIP-box found in X. arboricola pv. pruni. Sequence alignment, performed using ClustalW, showed a slight variation among X. arboricola and other Xanthomonas (A) as well as a variant in PIP-box sequence (B). (TIF) Click here for additional data file. S1 Table Genome statistics of 18 X. arboricola strains according to the automatic annotation performed using Prokka. (XLSX) Click here for additional data file. S2 Table Annotated protein sequences shared by 18 pathogenic and non-pathogenic strains of X. arboricola. (XLSX) Click here for additional data file. S3 Table Unique protein coding sequences (CDS) and related features found in the draft genome sequence of X. arboricola strain CITA 44 and X. arboricola pv. pruni strains CITA 33 and IVIA 26262.1. (XLSX) Click here for additional data file. S4 Table Orthologous protein coding sequences (CDS) of nine strains of X. arboricola associated with pathogenesis. (XLSX) Click here for additional data file. We would like to thank to Elisa Ferragud and Isabel M. Berruete for technical assistance on the phenotypic features evaluated as well as to Pilar Sabuquillo for assistance on searching orthologous genes. ==== Refs References 1 Vauterin L , Hoste B , Kersters K , Swings J . Reclassification of Xanthomonas . Int J Syst Bacteriol . 1995 ;45 (3 ): 472 –489 . 2 Fischer-Le Saux M , Bonneau S , Essakhi S , Manceau C , Jacques M-A . Aggressive emerging pathovars of Xanthomonas arboricola represent widespread epidemic clones distinct from poorly pathogenic strains, as revealed by multilocus sequence typing . Appl Environ Microbiol . 2015 ;81 (14 ): 4651 –4568 . 10.1128/AEM.00050-15 25934623 3 EFSA PLH Panel (EFSA Panel on Plant Health) . Scientific opinion on pest categorisation of Xanthomonas arboricola pv. pruni (Smith, 1903) . EFSA J . 2014 ;12 : 3857 –3882 . 4 Stefani E . Economic significance and control of bacterial spot/canker of stone fruits caused by Xanthomonas arboricola pv. pruni . J Plant Pathol . 2010 ;92 (1 , Supplement): S1.99 –S1.103 . 5 Boudon S , Manceau C , Nottéghem JL . Structure and origin of Xanthomonas arboricola pv. pruni populations causing bacterial spot of stone fruit trees in Western Europe . Phytopathology . 2005 ;95 (9 ): 1081 –1088 . 10.1094/PHYTO-95-1081 18943306 6 Denancé N , Lahaye T , Noël LD . Editorial: Genomics and effectomics of the crop killer Xanthomonas . Front Plant Sci . 2016 ;7 : 71 10.3389/fpls.2016.00071 26870077 7 Cesbron S , Briand M , Essakhi S , Gironde S , Boureau T , Manceau C , et al Comparative genomics of pathogenic and non-pathogenic strains of Xanthomonas arboricola unveil molecular and evolutionary events linked to pathoadaptation . Front Plant Sci . 2015 ;6 : 1126 10.3389/fpls.2015.01126 26734033 8 Essakhi S , Cesbron S , Fischer-Le Saux M , Bonneau S , Jacques M-A , Manceau C . Phylogenetic and VNTR analysis identified non-pathogenic lineages within Xanthomonas arboricola lacking the canonical type three secretion system . Appl Environ Microbiol . 2015 ;81 (16 ): 5395 –5410 . 10.1128/AEM.00835-15 26048944 9 Vernière CJ , Gottwald TR , Pruvost O . Disease development and symptom expression of Xanthomonas axonopodis pv. citri in various citrus plant tissues . Phytopathology . 2003 ;93 (7 ): 832 –43 . 10.1094/PHYTO.2003.93.7.832 18943164 10 Gordon JL , Lefeuvre P , Escalon A , Barbe V , Cruveiller S , Gagnevin L , et al Comparative genomics of 43 strains of Xanthomonas citri pv. citri reveals the evolutionary events giving rise to pathotypes with different host ranges . BMC Genomics . 2015 ;16 : 1098 10.1186/s12864-015-2310-x 26699528 11 Kamoun S , Kado CI . Phenotypic switching affecting chemotaxis, xanthan production, and virulence in Xanthomonas campestris . Appl Environ Microbiol . 1990 ;56 (12 ): 3855 –3860 16348384 12 Crossman L , Dow JM . Biofilm formation and dispersal in Xanthomonas campestris . Microbes Infect . 2004 ;6 (6 ): 623 –629 . 15158198 13 da Silva CR , Ferro J , Reinach FC , Farah CS , Furlan LR , Quaggio RB , et al Comparison of the genomes of two Xanthomonas pathogens with differing host specificities . Nature . 2002 ;417 (6887 ): 459 –463 . 12024217 14 Triplett LR , Hamilton JP , Buell CR , Tisserat NA , Verdier V , Zink F , et al Genomic analysis of Xanthomonas oryzae isolates from rice grown in the united states reveals substantial divergence from known X. oryzae pathovars . Appl Environ Microbiol . 2011 ;77 (12 ): 3930 –3937 . 10.1128/AEM.00028-11 21515727 15 Qian W , Han Z-J , He C . Two-component signal transduction systems of Xanthomonas spp.: a lesson from genomics . Mol Plant-Microbe Interact . 2008 ;21 (2 ): 151 –161 . 10.1094/MPMI-21-2-0151 18184059 16 Garita-Cambronero J , Sena-Vélez M , Palacio-Bielsa A , Cubero J . Draft genome sequence of Xanthomonas arboricola pv. pruni strain Xap33, causal agent of bacterial spot disease on almond . Genome Announc . 2014 ;2 (3 ): e00440 –14 . 10.1128/genomeA.00440-14 24903863 17 Garita-Cambronero J , Palacio-Bielsa A , López MM , Cubero J . Draft genome sequence for virulent and avirulent strains of Xanthomonas arboricola isolated from Prunus spp. in Spain . Stand Genomic Sci . 2016 ;11 : 12 10.1186/s40793-016-0132-3 26823958 18 Young JM , Park DC , Shearman HM , Fargier E . A multilocus sequence analysis of the genus Xanthomonas . Syst Appl Microbiol . 2008 ;31 (5 ): 366 –377 . 10.1016/j.syapm.2008.06.004 18783906 19 Ignatov AN , Kyrova EI , Vinogradova S V , Kamionskaya AM , Schaad NW , Luster DG . Draft genome sequence of Xanthomonas arboricola strain 3004, a causal agent of bacterial disease on barley . Genome Announc . 2015 ;3 (1 ): e01572 –14 . 10.1128/genomeA.01572-14 25700410 20 Ibarra Caballero J , Zerillo MM , Snelling J , Boucher C , Tisserat N . Genome sequence of Xanthomonas arboricola pv. corylina, isolated from Turkish filbert in Colorado . Genome Announc . 2013 ;1 (3 ): e00246 –13 . 10.1128/genomeA.00246-13 23704178 21 Higuera G , González-Escalona N , Véliz C , Vera F , Romero J . Draft genome sequences of four Xanthomonas arboricola pv. juglandis strains associated with walnut blight in Chile . Genome Announc . 2015 ;3 (5 ): e01160 –15 . 10.1128/genomeA.01160-15 26450732 22 Pereira UP , Gouran H , Nascimento R , Adaskaveg JE , Goulart LR , Dandekar AM . Complete genome sequence of Xanthomonas arboricola pv. juglandis 417, a copper-resistant strain isolated from Juglans regia L . Genome Announc . 2015 ;3 (5 ): e01126 –15 . 10.1128/genomeA.01126-15 26430043 23 Harrison J , Grant MR , Studholme DJ . Draft genome sequences of two strains of Xanthomonas arboricola pv. celebensis isolated from banana plants . Genome Announc . 2016 ;4 (1 ): e01705 –15 . 10.1128/genomeA.01705-15 26868395 24 Sena-Vélez M. Mecanismos implicados en las etapas iniciales de infección en la cancrosis de los cítricos provocada por Xanthomonas citri subsp. citri. Doctoral dissertation. Politechnic University of Madrid. 2015. Available: http://oa.upm.es/38610/ 25 Mhedbi-Hajri N , Darrasse A , Pigné S , Durand K , Fouteau S , Barbe V , et al Sensing and adhesion are adaptive functions in the plant pathogenic xanthomonads . BMC Evol Biol . 2011 ;11 : 67 10.1186/1471-2148-11-67 21396107 26 Darling ACE , Mau B , Blattner FR , Perna NT . Mauve: multiple alignment of conserved genomic sequence with rearrangements . Genome Res . 2004 ;14 (7 ): 1394 –1403 . 15231754 27 Burch AY , Shimada BK , Browne PJ , Lindow SE . Novel high-throughput detection method to assess bacterial surfactant production . Appl Environ Microbiol . 2010 ;76 (16 ): 5363 –5372 . 10.1128/AEM.00592-10 20562275 28 Mattick JS . Type IV pili and twitching motility . Annu Rev Microbiol . 2002 ;56 : 289 –314 . 12142488 29 Chevance FFV , Hughes KT . Coordinating assembly of a bacterial macromolecular machine . Nat Rev Microbiol . 2008 ;6 (6 ): 455 –465 . 10.1038/nrmicro1887 18483484 30 Dunger G , Llontop E , Guzzo CR , Farah CS . The Xanthomonas type IV pilus . Curr Opin Microbiol . 2016 ;30 : 88 –97 . 10.1016/j.mib.2016.01.007 26874963 31 Filloux A . The underlying mechanisms of type II protein secretion . Biochim Biophys Acta (BBA)—Mol Cell Res . 2004 ;1694 (1–3 ): 163 –179 . 32 Szczesny R , Jordan M , Schramm C , Schulz S , Cogez V , Bonas U , et al Functional characterization of the Xcs and Xps type II secretion systems from the plant pathogenic bacterium Xanthomonas campestris pv vesicatoria . New Phytol . 2010 ;187 (4 ): 983 –1002 . 10.1111/j.1469-8137.2010.03312.x 20524995 33 Vandroemme J , Cottyn B , Baeyen S , De Vos P , Maes M . Draft genome sequence of Xanthomonas fragariae reveals reductive evolution and distinct virulence-related gene content . BMC Genomics . 2013 ;14 : 829 10.1186/1471-2164-14-829 24274055 34 Subramoni S , Suárez-Moreno ZR , Venturi V . Lipases as pathogenicity factors of plant pathogens In: Timmis KN , editor. Handbook of hydrocarbon and lipid microbiology . Springer ; 2010 pp. 3269 –3277 . 35 Vorhölter FJ , Schneiker S , Goesmann A , Krause L , Bekel T , Kaiser O , et al The genome of Xanthomonas campestris pv. campestris B100 and its use for the reconstruction of metabolic pathways involved in xanthan biosynthesis . J Biotechnol . 2008 ;134 (1–2 ): 33 –45 . 10.1016/j.jbiotec.2007.12.013 18304669 36 Bonas U , Stall RE , Staskawicz B . Genetic and structural characterization of the avirulence gene avrBs3 from Xanthomonas campestris pv. vesicatoria . Mol Gen Genet . 1989 ;218 (1 ): 127 –136 . 2550761 37 White FF , Potnis N , Jones JB , Koebnik R . The type III effectors of Xanthomonas . Mol Plant Pathol . 2009 ;10 (6 ): 749 –766 . 10.1111/j.1364-3703.2009.00590.x 19849782 38 Bogdanove AJ , Koebnik R , Lu H , Furutani A , Angiuoli SV ., Patil PB , et al Two new complete genome sequences offer insight into host and tissue specificity of plant pathogenic Xanthomonas spp . J Bacteriol . 2011 ;193 (19 ): 5450 –5464 . 10.1128/JB.05262-11 21784931 39 Guo Y , Figueiredo F , Jones J , Wang N . HrpG and HrpX play global roles in coordinating different virulence traits of Xanthomonas axonopodis pv. citri . Mol Plant-Microbe Interact . 2011 ;24 (6 ): 649 –661 . 10.1094/MPMI-09-10-0209 21261465 40 Potnis N , Krasileva K , Chow V , Almeida NF , Patil PB , Ryan RP , et al Comparative genomics reveals diversity among xanthomonads infecting tomato and pepper . BMC Genomics . 2011 ;12 : 146 10.1186/1471-2164-12-146 21396108 41 Hajri A , Pothier JF , Le-Saux MF , Bonneau S , Poussier S , Boureau T , et al Type three effector gene distribution and sequence analysis provide new insights into the pathogenicity of plant-pathogenic Xanthomonas arboricola . Appl Environ Microbiol . 2012 ;78 (2 ): 371 –384 . 10.1128/AEM.06119-11 22101042 42 Li RF , Lu GT , Li L , Su HZ , Feng G , Chen Y , et al Identification of a putative cognate sensor kinase for the two-component response regulator HrpG, a key regulator controlling the expression of the hrp genes in Xanthomonas campestris pv. campestris . Environ Microbiol . 2014 ;16 (7 ): 2053 –2071 . 10.1111/1462-2920.12207 23906314 43 Jacobs JM , Pesce C , Lefeuvre P , Koebnik R . Comparative genomics of a cannabis pathogen reveals insight into the evolution of pathogenicity in Xanthomonas . Front Plant Sci . 2015 ;6 : 431 10.3389/fpls.2015.00431 26136759 44 Pothier JF , Vorhölter FJ , Blom J , Goesmann A , Pühler A , Smits THM , et al The ubiquitous plasmid pXap41 in the invasive phytopathogen Xanthomonas arboricola pv. pruni: Complete sequence and comparative genomic analysis . FEMS Microbiol Lett . 2011 ;323 (1 ): 52 –60 . 10.1111/j.1574-6968.2011.02352.x 21732961 45 Burokiene D , Pulawska J . Characterization of Xanthomonas arboricola pv. juglandis isolated from walnuts in Lithuania . J Plant Pathol . 2012 ; 94 (1 , Supplement): S1.23 –S1.27 . 46 Kaluzna M , Pulawska J , Waleron M , Sobiczewski P . The genetic characterization of Xanthomonas arboricola pv. juglandis, the causal agent of walnut blight in Poland . Plant Pathol . 2014 ;63 (6 ): 1404 –1416 . 47 Graham JH , Gottwald TR . Research perspectives on eradication of citrus bacterial diseases in Florida . Plant Dis . 1991 ;75 : 1193 . 48 Jalan N , Aritua V , Kumar D , Yu F , Jones JB , Graham JH , et al Comparative genomic analysis of Xanthomonas axonopodis pv. citrumelo F1, which causes citrus bacterial spot disease, and related strains provides insights into virulence and host specificity . J Bacteriol . 2011 ;193 (22 ): 6342 –6357 . 10.1128/JB.05777-11 21908674 49 Stoyanova M , Vancheva T , Moncheva P , Bogatzevska N . Differentiation of Xanthomonas spp. causing bacterial spot in Bulgaria based on biolog system . Int J Microbiol . 2014 ; 2014 ; 2014 : 495476 25197281 50 Massomo SMS , Nielsen H , Mabagala RB , Mansfeld-Giese K , Hockenhull J , Mortensen CN . Identification and characterisation of Xanthomonas campestris pv. campestris strains from Tanzania by pathogenicity tests, biolog, rep-PCR and fatty acid methyl ester analysis . Eur J Plant Pathol . 2003 ;109 (8 ): 775 –789 . 51 Pieretti I , Royer M , Barbe V , Carrere S , Koebnik R , Couloux A , et al Genomic insights into strategies used by Xanthomonas albilineans with its reduced artillery to spread within sugarcane xylem vessels . BMC Genomics . 2012 ;13 : 658 10.1186/1471-2164-13-658 23171051 52 Noinaj N , Guillier M , Barnard TJ , Buchanan SK . TonB-dependent transporters: regulation, structure, and function . Annu Rev Microbiol . 2010 ;64 : 43 –60 . 10.1146/annurev.micro.112408.134247 20420522 53 Blanvillain S , Meyer D , Boulanger A , Lautier M , Guynet C , Denancé N , et al Plant carbohydrate scavenging through TonB-dependent receptors: A feature shared by phytopathogenic and aquatic bacteria . PLoS One . 2007 ;2 (2 ): e224 17311090 54 Wadhams GH , Armitage JP . Making sense of it all: bacterial chemotaxis . Nat Rev Mol Cell Biol . 2004 ;5 (12 ): 1024 –1037 . 15573139 55 Moreira LM , Facincani AP , Ferreira CB , Ferreira RM , Ferro MIT , Gozzo FC , et al Chemotactic signal transduction and phosphate metabolism as adaptive strategies during citrus canker induction by Xanthomonas citri . Funct Integr Genomics . 2015 ;15 (2 ): 197 –210 . 10.1007/s10142-014-0414-z 25403594 56 Kearns DB . A field guide to bacterial swarming motility . Nat Rev Microbiol . 2010 ;8 (9 ): 634 –44 . 10.1038/nrmicro2405 20694026 57 Malamud F , Torres PS , Roeschlin R , Rigano LA , Enrique R , Bonomi HR , et al The Xanthomonas axonopodis pv. citri flagellum is required for mature biofilm and canker development . Microbiology . 2011 ;157 (Pt 3 ): 819 –829 . 10.1099/mic.0.044255-0 21109564 58 Sena-Vélez M , Redondo C , Gell I , Ferragud E , Johnson E , Graham JH , et al Biofilm formation and motility of Xanthomonas strains with different citrus host range . Plant Pathol . 2015 ;64 (4 ): 767 –775 . 59 Shen Y , Chern M , Silva FG , Ronald P . Isolation of a Xanthomonas oryzae pv. oryzae flagellar operon region and molecular characterization of flhF . Mol Plant-Microbe Interact . 2001 ;14 (2 ): 204 –213 11204784 60 Gottig N , Garavaglia BS , Garofalo CG , Orellano EG , Ottado J . A filamentous hemagglutinin-like protein of Xanthomonas axonopodis pv. citri, the phytopathogen responsible for citrus canker, is involved in bacterial virulence . PLoS One . 2009 ;4 (2 ): e4358 10.1371/journal.pone.0004358 19194503 61 Köhler T , Curty LK , Barja F , van Delden C , Pechere JC . Swarming of Pseudomonas aeruginosa is dependent on cell-to-cell signaling and requires flagella and pili . J Bacteriol . 2000 ;182 (21 ): 5990 –5996 . 11029417 62 Verstraeten N , Braeken K , Debkumari B , Fauvart M , Fransaer J , Vermant J , et al Living on a surface: swarming and biofilm formation . Trends Microbiol . 2008 ;16 (10 ): 496 –506 . 10.1016/j.tim.2008.07.004 18775660 63 Sun W , Dunning FM , Pfund C , Weingarten R , Bent AF . Within-species flagellin polymorphism in Xanthomonas campestris pv campestris and its impact on elicitation of Arabidopsis FLAGELLIN SENSING2-dependent defenses . Plant Cell . 2006 ;18 (3 ): 764 –779 . 16461584 64 Ikeda T , Oosawa K , Hotani H . Self-assembly of the filament capping protein, FliD, of bacterial flagella into an annular structure . J Mol Biol . 1996 ;259 (4 ): 679 –686 . 8683574 65 Kim JS , Chang JH , Chung S Il, Yum JS . Molecular cloning and characterization of the Helicobacter pylori fliD gene, an essential factor in flagellar structure and motility . J Bacteriol . 1999 ;181 (22 ): 6969 –6976 . 10559162 66 Craig L , Li J . Type IV pili: paradoxes in form and function . Curr Opin Struct Biol . 2008 ;18 (2 ): 267 –277 . 10.1016/j.sbi.2007.12.009 18249533 67 Ray SK , Rajeshwari R , Sharma Y , Sonti R V . A high-molecular-weight outer membrane protein of Xanthomonas oryzae pv.oryzae exhibits similarity to non-fimbrial adhesins of animal pathogenic bacteria and is required for optimum virulence . Mol Microbiol . 2002 ;46 (3 ): 637 –647 . 12410822 68 Das A , Rangaraj N , Sonti R V . Multiple adhesin-like functions of Xanthomonas oryzae pv. oryzae are involved in promoting leaf attachment, entry, and virulence on rice . Mol Plant-Microbe Interact . 2009 ;22 (1 ): 73 –85 . 10.1094/MPMI-22-1-0073 19061404 69 Ryan RP , Vorhölter FJ , Potnis N , Jones JB , Van Sluys MA , Bogdanove AJ , et al Pathogenomics of Xanthomonas: understanding bacterium-plant interactions . Nat Rev Microbiol . 2011 ;9 (5 ): 344 –355 . 10.1038/nrmicro2558 21478901 70 Cianciotto NP . Type II secretion: A protein secretion system for all seasons . Trends Microbiol . 2005 ;13 (12 ): 581 –588 . 16216510 71 Yun MH , Torres PS , El Oirdi M , Rigano LA , Gonzalez-Lamothe R , Marano MR , et al Xanthan induces plant susceptibility by suppressing callose deposition . Plant Physiol . 2006 ;141 (1 ): 178 –187 . 16531487 72 Katzen F , Ferreiro DU , Oddo CG , Ielmini MV , Becker A , Pühler A , et al Xanthomonas campestris pv. campestris gum mutants: Effects on xanthan biosynthesis and plant virulence . J Bacteriol . 1998 ;180 (7 ): 1607 –1617 9537354 73 Jalan N , Kumar D , Yu F , Jones JB , Graham JH , Wang N . Complete genome sequence of Xanthomonas citri subsp. citri strain Aw12879, a restricted-rost-range citrus canker-causing bacterium . Genome Announc . 2013 ;1 (3 ): e00235 –13 . 10.1128/genomeA.00235-13 23682143 74 Hall T . BioEdit: An important software for molecular biology . GERF Bull Biosci . 2011 ;2 (1 ): 60 –61 . 75 Goujon M , McWilliam H , Li W , Valentin F , Squizzato S , Paern J , et al A new bioinformatics analysis tools framework at EMBL-EBI . Nucleic Acids Res . 2010 ;38 (Web Server issue): W695 –W699 . 10.1093/nar/gkq313 20439314 76 Posada D . jModelTest: Phylogenetic model averaging . Mol Biol Evol . 2008 ;25 (7 ): 1253 –6 . 10.1093/molbev/msn083 18397919 77 Tamura K , Peterson D , Peterson N , Stecher G , Nei M , Kumar S . MEGA5: Molecular evolutionary genetics analysis using maximum likelihood, evolutionary distance, and maximum parsimony methods . Mol Biol Evol . 2011 ;28 (10 ): 2731 –2739 . 10.1093/molbev/msr121 21546353 78 Posada D , Buckley TR . Model selection and model averaging in phylogenetics: advantages of akaike information criterion and bayesian approaches over likelihood ratio tests . Syst Biol . 2004 ;53 (5 ): 793 –808 . 15545256 79 Seemann T . Prokka: rapid prokaryotic genome annotation . Bioinformatics . 2014 ;30 (14 ): 2068 –2069 . 10.1093/bioinformatics/btu153 24642063 80 Page AJ , Cummins CA , Hunt M , Wong VK , Reuter S , Holden MTG , et al Roary: rapid large-scale prokaryote pan genome analysis . Bioinformatics . 2015 ;31 (22 ): 3691 –3693 . 10.1093/bioinformatics/btv421 26198102 81 Krogh A , Larsson B , von Heijne G , Sonnhammer EL . Predicting transmembrane protein topology with a hidden Markov model: application to complete genomes . J Mol Biol . 2001 ;305 (3 ): 567 –580 . 11152613 82 Petersen TN , Brunak S , von Heijne G , Nielsen H . SignalP 4.0: discriminating signal peptides from transmembrane regions . Nat Methods . 2011 ;8 (10 ): 785 –786 . 10.1038/nmeth.1701 21959131 83 Tatusov RL . The COG database: a tool for genome-scale analysis of protein functions and evolution . Nucleic Acids Res . 2000 ;28 (1 ): 33 –36 . 10592175 84 Marchler-Bauer A , Derbyshire MK , Gonzales NR , Lu S , Chitsaz F , Geer LY , et al CDD: NCBI’s conserved domain database . Nucleic Acids Res . 2014 ;43 (Database issue): D222 –6 . 10.1093/nar/gku1221 25414356 85 Katoh K , Misawa K , Kuma K , Miyata T . MAFFT: a novel method for rapid multiple sequence alignment based on fast Fourier transform . Nucleic Acids Res . 2002 ;30 (14 ): 3059 –3066 . 12136088 86 Stamatakis A . RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies . Bioinformatics . 2014 ;30 (9 ): 1312 –1313 . 10.1093/bioinformatics/btu033 24451623 87 Huson DH , Richter DC , Rausch C , Dezulian T , Franz M , Rupp R . Dendroscope: An interactive viewer for large phylogenetic trees . BMC Bioinformatics . 2007 ;8 : 460 18034891 88 Vernière C , Pruvost O , Civerolo EL , Gambin O , Jacquemoud-Collet JP , Luisetti J . Evaluation of the biolog substrate utilization system to identify and assess metabolic variation among strains of Xanthomonas campestris pv. citri . Appl Environ Microbiol . 1993 ;59 (1 ): 243 –249 16348849 89 Maharjan RP , Seeto S , Ferenci T . Divergence and redundancy of transport and metabolic rate-yield strategies in a single Escherichia coli population . J Bacteriol . 2007 ;189 (6 ): 2350 –2358 . 17158684 90 National Center for Biotechnology Information (NCBI). How to: Find a homolog for a gene in another organism. 2016. Available: http://www.ncbi.nlm.nih.gov/guide/howto/find-homolog-gene/ 91 Pothier JF , Smits THM , Blom J , Vorhölter FJ , Goesmann A , Pühler A , et al Complete genome sequence of the stone fruit pathogen Xanthomonas arboricola pv. pruni . Phytopathology . 2011 ;101 : S144 –S145 . 92 Blocker A , Komoriya K , Aizawa S-I . Type III secretion systems and bacterial flagella: insights into their function from structural similarities . Proc Natl Acad Sci USA . 2003 ;100 (6 ): 3027 –3030 . 12631703 93 Büttner D , Bonas U . Regulation and secretion of Xanthomonas virulence factors . FEMS Microbiol Rev . 2010 ;34 (2 ): 107 –133 . 10.1111/j.1574-6976.2009.00192.x 19925633 94 Abby SS , Rocha EPC . The non-flagellar type III secretion system evolved from the bacterial flagellum and diversified into host-cell adapted systems . PLoS Genet . 2012 ;8 (9 ): e1002983 10.1371/journal.pgen.1002983 23028376 95 He Y-W , Zhang L-H . Quorum sensing and virulence regulation in Xanthomonas campestris . FEMS Microbiol Rev . 2008 ;32 (5 ): 842 –857 . 10.1111/j.1574-6976.2008.00120.x 18557946 96 Nascimento R , Gouran H , Chakraborty S , Gillespie HW , Almeida-Souza HO , Tu A , et al The type II secreted lipase/esterase LesA is a key virulence factor required for Xylella fastidiosa pathogenesis in grapevines . Sci Rep . 2016 ;6 : 18598 10.1038/srep18598 26753904 97 Alikhan NF , Petty NK , Ben Zakour NL , Beatson SA . BLAST Ring Image Generator (BRIG): simple prokaryote genome comparisons . BMC Genomics . 2011 ;12 : 402 10.1186/1471-2164-12-402 21824423 98 Grant JR , Stothard P . The CGView server: a comparative genomics tool for circular genomes . Nucleic Acids Res . 2008 ; 36 (Web Server issue): W181 –W184 : 181–184. 10.1093/nar/gkn179 18411202

PMC005xxxxxx/PMC5003340.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 27571209PONE-D-16-0795110.1371/journal.pone.0161024Research ArticleEngineering and TechnologyFluidicsMicrofluidicsPhysical SciencesMaterials ScienceMaterials by StructureAmorphous SolidsGelsPhysical SciencesMaterials ScienceMaterials by StructureMixturesGelsPhysical SciencesPhysicsClassical MechanicsContinuum MechanicsFluid MechanicsFluid DynamicsFlow RatePhysical SciencesChemistryChemical CompoundsOrganic CompoundsAlcoholsEthanolPhysical SciencesChemistryOrganic ChemistryOrganic CompoundsAlcoholsEthanolEngineering and TechnologySignal ProcessingEngineering and TechnologyElectronicsPhysical SciencesPhysicsClassical MechanicsContinuum MechanicsFluid MechanicsFluid DynamicsFluid FlowEngineering and TechnologyElectrical EngineeringElectrical CircuitsElectronic CircuitsIntegrated Microfluidic Membrane Transistor Utilizing Chemical Information for On-Chip Flow Control Integrated Microfluidic Transistor for On-Chip Flow ControlFrank Philipp 1Schreiter Joerg 23Haefner Sebastian 1Paschew Georgi 1Voigt Andreas 13Richter Andreas 13* 1 Polymeric Microsystems, Institute for Semiconductors and Microsystems, Technische Universität Dresden, Dresden, Germany 2 Highly-Parallel VLSI-Systems and Neuro-Microelectronics Chair, Institute for the Fundamentals of Electroncis, Technische Universität Dresden, Dresden, Germany 3 Center for Advancing Electronics Dresden (cfaed), Dresden, Germany Han Arum Editor Texas A&M University College Station, UNITED STATES Competing Interests: The authors have declared that no competing interests exist. Conceived and designed the experiments: PF GP AV AR. Performed the experiments: PF. Analyzed the data: PF JS AV. Contributed reagents/materials/analysis tools: PF JS SH. Wrote the paper: PF JS AV AR. Designed the software used in experiment and analysis: PF. * E-mail: andreas.richter7@tu-dresden.de2016 29 8 2016 11 8 e016102424 2 2016 28 7 2016 © 2016 Frank et al2016Frank et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Microfluidics is a great enabling technology for biology, biotechnology, chemistry and general life sciences. Despite many promising predictions of its progress, microfluidics has not reached its full potential yet. To unleash this potential, we propose the use of intrinsically active hydrogels, which work as sensors and actuators at the same time, in microfluidic channel networks. These materials transfer a chemical input signal such as a substance concentration into a mechanical output. This way chemical information is processed and analyzed on the spot without the need for an external control unit. Inspired by the development electronics, our approach focuses on the development of single transistor-like components, which have the potential to be used in an integrated circuit technology. Here, we present membrane isolated chemical volume phase transition transistor (MIS-CVPT). The device is characterized in terms of the flow rate from source to drain, depending on the chemical concentration in the control channel, the source-drain pressure drop and the operating temperature. http://dx.doi.org/10.13039/501100001659Deutsche ForschungsgemeinschaftRTG 1865 "Hydrogel-based Microsystems"Frank Philipp http://dx.doi.org/10.13039/501100001659Deutsche ForschungsgemeinschaftRTG 1865 "Hydrogel-based Microsystems"Haefner Sebastian http://dx.doi.org/10.13039/501100001659Deutsche ForschungsgemeinschaftEXC 1056 "Center for Advancing Electronics"Voigt Andreas http://dx.doi.org/10.13039/501100001659Deutsche ForschungsgemeinschaftEXC 1056 "Center for Advancing Electronics"Schreiter Joerg Free State of SaxonyResearch Group "Diagnostic Integrated Chemical Circuits"Paschew Georgi This work was supported by Deutsche Forschungsgemeinschaft, Research Training Group 1865 "Hydrogel-based Microsystems", Recipients: PF, SH (URL: http://www.dfg.de/en/research_funding/programmes/list/projectdetails/index.jsp?id=211944370); Deutsche Forschungsgemeinschaft, EXC 1056 "Center for Advancing Electronics Dresden", Recipients: JS, AV (URL: http://www.dfg.de/en/research_funding/programmes/list/projectdetails/index.jsp?id=194636624); and Free State Saxony "Diagnostic Integrated Chemicl Circuits", Recipient: GP. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Lab-on-a-chip-technology as an integrated circuit (IC) concept is intended to provide economical and functional advantages in microfluidics in a manner comparable to microelectronics in electronic information technology. However, Albert Folch noted that “…this vision is not as fully realized as the exponential progression of integrated circuit improvement known as Moore’s law, simply because there is no such thing as a microfluidic transistor on which to sustain the same economy of scale as there was for the electronic transistor.” [1]. The dominating microelectromechanical system (MEMS)-based lab on a chip (LoC) technology employs computer controlled microfluidic components to carry out microfluidic IC programs [2]. Whereas electronic ICs are the heart of computers, current MEMS-based LoCs are simply complex computer-controlled devices. One common approach is the use of elastomeric valves that can be activated by a pneumatic or hydraulic input signal, and output a pre-determined microfluidic signal. Using this concept, basic circuits like logic gates and oscillators have been shown [3–6] constituting fluid-mechanical counterparts to the electronic circuits. Against this background, there is a high motivation to investigate the possibilities of a microfluidic transistor concept as the component base for integrated circuits, eventually providing a comparable economy of scale in microfluidics as there is for microelectronics. While in microelectronics electric charges transport information, in microfluidics the fluid containing reactive molecules and ions is the carrier of information. This extends the scope of information channels and information types of the technology resulting in not only the manipulation of information, but of matter. At the core of our concept, a microfluidic, (or better: chemo-fluidic) transistor connects the signal domain of the fluid as multi-dimensional carrier of information with the mechanical or microfluidic domain. The coupling of domains is facilitated by intrinsically active materials such as stimuli-responsive hydrogels, which react to small chemical changes in a fluid with a volume phase transition, where the gels undergo a significant change in volume. Typical devices facilitating valving in micro channels such as pneumatic membrane valves [7], vacuum valves [8], thermally [9] or mechanically activated valves [10] rely on externally applied energy in order to function. The energy is triggered via a control signal, which also originates from an external control setup e.g. a computer with according circuitry. The chemo-fluidic membrane transistor in contrast uses the input from the process liquid in order to make a switching decision, even more so it is able to conduct a fluid with according fluids to subsequent chemo-fluidic transistors, and therefore carry out a signal propagation. Most microfluidic valves conceptionally lack the ability to effect the switching decision of the subsequently following device, contradicting an autonomous information processing and further a conclusive IC concept. Inspired by microelectronics, the chemo-fluidic transistor paves the way for a transistor-based circuitry as an on-chip flow control extracting inputs directly from the process fluid, eventually rendering all external control devices obsolete. In our concept, integrated circuits, just like in electronics, conduct the manipulation (of fluid) purely on the chip level. In order to lay the foundation for this comprehensive and complex circuit concept for LoCs we describe in this article the key device, the chemo-fluidic transistor. The transistor offers a flow control in one channel while the signal input is given by a separate channel. The channels are coupled hydraulically via a flexible membrane instituting the devices name: membrane-isolated chemo-fluidic volume phase transition transistor (MIS-CVPT), or chemo-fluidic membrane transistor. Hydrogels as active material Similar to the electronic model, the MIS-CVPT is a monolithic device, whose functioning relies on intrinsically active materials, here stimuli-responsive hydrogels. These polymers show unique material properties. Hydrogels are three-dimensional networks of polymer chains synthesized via a cross-linking polymerization or a cross-linking reaction [11]. The polymers building hydrogels show solubility in water. While a polymer will dissolve in its solvent, a cross-linked polymer, due to its interconnections, will not dissolve but retain the solvent within its network, leading to a swelling of the structure. The equilibrium swelling degree depends on the solubility in the solution utilized. Further, stimuli-responsive hydrogels exhibit a reversible first order volume phase transition. By small changes of the thermodynamic properties of the solution, the gel can be made to undergo a volume phase transition between a shrunken state and a swollen state. In the shrunken state the polymer-polymer interactions are prevalent and the gel is highly hydrophobic. In the swollen state (mixing state) solvent-polymer interactions are dominant. Hydrogels have been designed to show sensitivity to a large variety of thermodynamic parameters including concentrations of solvents, specific ions, biomolecules and pH value [12–17], but also to electrical field [18], light [19] or temperature [20]. For the MIS-CVPT a co-polymer of poly(N-isopropylacrylamide) (PNIPAAm) and the superabsorbent sodium acrylate (NaA) is employed. The composition of both monomers combines the sensitivities of NIPAAm (alcohol and temperature sensitivity [21, 22]) with the high swelling degree of NaA. The swelling behavior of PNIPAAm and the PNIPAAm-NaA composition is shown in S1 Fig over the ethanol concentration. The co-polymer exhibits a higher swelling degree, while showing a change of equilibrium volume at roughly the same values of ethanol concentration as pure PNIPAAm. The gel particle is produced photolithographically [23] and is placed in the device acting mechanically on chemical inputs by changing its volume. Device principle The chemo-fluidic membrane transistor consists of two operational layers, the flow layer and the control layer including a hydrogel particle, which are separated by a thin flexible membrane (Fig 1A–1C and 1E) derived from a series of works [24–27] but specifically extending the concept of the chemo-mechanical valve [21, 28]. The thermal responsiveness of hydrogels as microfluidic valve has been exploited before by [9], but without utilizing the coupling of chemical and fluidic domain. The flow layer is interrupted by a channel break, forming, in combination with the membrane, a doormat-like valve [29–31]. The doormat setup was chosen as it closes a channel tightly, in contrast to a membrane valve setup without the channel break. In order to ensure good tightness of the valve the flow channel was designed to be smaller than the underlying hydrogel particle. A microscope picture of the device is given in Fig 1A for the closed and open state. Typically in the closed state, where the hydrogel is fully swollen, the gel particle is not visible, because its refractive index is close to the one of water, making it indistinguishable from the aqueous phase. 10.1371/journal.pone.0161024.g001Fig 1 (A), Microscope image of the MIS-CVPT (scale bar 500 μm). (A-top), Fully swollen hydrogel closing the transistor for a low substance concentration. (A-bottom), Shrunken hydrogel opening the transistor for a high substance concentration. (B), Schematic 3D CAD model of the MIS-CVPT closed and open. (C-D), Comparison of fluidic and electronic domain, devices with according terminals. Voltage corresponds to pressure and current corresponds to flow. (C), Schematic top view of the MIS-CVPT with analogue electronic ports of the MOSFET. (D), Graphical symbol of the n-type enhancement MOSFET with its terminals. (E), Cross section view with open and closed state of the MIS-CVPT including the used materials. In reference to the electric-fluidic analogy, the MIS-CVPT is presumed to be the counterpart of the n-type enhancement metal-oxide-semiconductor field-effect transistor (MOSFET) in electronics: Drain-source voltage UDS corresponds to the excitation pressure pDS from (D) to (S) in the flow channel (Fig 1C and 1D), the drain-source current IDS corresponds to the flow rate QDS in the flow channel. Gate-source voltage UGS, which is the control voltage, corresponds to the substance concentration csub (port (C)), which determines the pressure applied by the hydrogel against the membrane (Fig 1E). At csub = 0wt% the pressure applied by the gel is high and decreases with increasing substance concentration. Analogue to the n-type enhancement MOSFET the MIS-CVPT is also normally non-conductive. Similar to the threshold voltage of a MOSFET, there is a minimum concentration csub,0 for the MIS-CVPT to open. While in electronics temperature is often regarded a perturbation, for the chemo-fluidic transistor it is more of an additional control input, and can be used e.g. to adjust the threshold. In the MOSFET, the gate is electrically isolated from the source-drain channel. Likewise, there is no fluid flow from the control port to the source-drain channel in the MIS-CVPT. The sensitivity and specificity to a certain input is bound directly to the composition of the gel. By changing the composition of the gel a different information can be processed and be utilized as trigger. PNIPAAm for instance is known for a distinct sensitivity towards a series of alcohols [21] and towards temperature change [32]. For this study we use a copolymer of PNIPAAm and sodium acrylate. This gel composition has shown to have a high swelling degree from the sodium acrylate while still keeping the sensitivity of the PNIPAAm [33]. The concentration cEth of ethanol is used as the chemical input of the chemo-fluidic membrane transistor. Experimental The chemo-fluidic membrane transistor was fabricated using multi-layer soft lithography (Fig 2). Prior to the PDMS moulding process photolithographically structured masters for each functional layer were produced. While the control layer was produced by spin coating to achieve a thin layer the flow layer was casted resulting in a fairly thick layer. The assembling of the layers and the incorporation of the hydrogel particle concluded the fabrication process. 10.1371/journal.pone.0161024.g002Fig 2 Overview of the fabrication procedure. Master fabrication: (A) Lamination of dry film resist onto substrate. (B) Exposure with UV light through photo mask. (C) Post-exposure bake. (D) Development and rinsing with following hard bake. Chip fabrication: (E) Spin coating of PDMS on control layer master. (F) Moulding of PDMS on flow layer master. Chip assembling: (G) Inhibition of the channel break in the flow layer. (H) Plasma bonding with aligning of the PDMS layers. (I) Incorporation of the hydrogel particle into the control channel. (J) Plasma bonding of the multi-layer chip onto a cover glass. Master fabrication An overview about the master fabrication is given in Fig 2A–2D. A glass substrate (Borofloat33, Schott Jena, Germany) with a thickness of 700μm was cleaned with acetone, isopropanol as well as distilled water and consecutively dried using nitrogen. A dehydration bake at 150°C for 20 min was performed on a hot plate as all following baking steps. One layer resist (Ordyl SY355, Elga Europe, Italy) of thickness 50μm was laminated and soft baked at 85°C for 3 min. This step was repeated for a second layer of 50μm (Fig 2A). The substrate was exposed through a polymer film mask for 90 s for the flow layer and 65 s for the control layer (Fig 2B). The intensity was previously measured and determined to be 1.66W cm−2. The post-exposure bake was performed at 85°C for 30 min (Fig 2C). For development the Ordyl developer and rinser (Ordyl Developer and Rinser, Elga Europe, Italy) was used. The development was finished off with rinsing with isopropanol and distilled water (Fig 2D). The process was completed with a hard bake at 120°C for 1.5 h. Chip fabrication Control layer The control layer contains the membrane which has a thickness of about 30 μm. Therefore spin-coating was chosen as the method of fabrication. 22.0 g of PDMS (Sylgard 184, Dow Corning, USA) in a ratio 10:1 is mixed with an electric stirrer for 5 min. It is then degased for 45 min. The bubble-free PDMS was used to spin coat the control master (Fig 2E) creating a layer thickness of 130 μm. The PDMS was then cured in a convection oven at 60°C for 2 h. Flow layer 27.5 g of polydimethylsiloxane (PDMS) (Sylgard 184, Dow Corning, USA) in a ratio 10:1 was mixed with an electric stirrer for 5 min. It was then degased for 45 min. The bubble-free PDMS was poured on top of the flow master (Fig 2F) creating a layer thickness of 4 mm. The PDMS was then cured in a convection oven at 60°C for 2 h. Hydrogel particle The hydrogel solution was prepared under protective argon atmosphere. For a batch of 25 ml 2.069 g PNIPAAm (poly(N-isopropylacrylamide), Sigma-Aldrich, USA), 0.044 g Sodium acrylate (Sigma-Aldrich, USA), 0.043 g N,N’-Methylenbisacrylamide (Sigma-Aldrich, USA) and 0.042 g photo initiator (2-methylpropiophenone, Sigma-Aldrich, USA) were weighed out. The solids were dissolved in 25 ml de-ionized water. The hydrogel solution was pipetted into a PDMS mould under argon protective gas and closed up with a glass slide. The mould was then exposed to UV light for 20 s. The gel structures were dried in the mould at room temperature overnight. Chip assembling The assembling is additionally documented in Fig 2G–2J. Prior to the bonding of the two functional layers, the doormat setup in the flow layer is inhibited from bonding by dispensing a small drop of oil on top (Fig 2G). The prepared flow layer was placed into a plasma oven (DREVA Clean 450, Vakuumtechnik Dresden GmbH, Germany) together with the control layer (still on the master). The two layers were exposed to an oxygen plasma for 2 min at 50W with an oxygen flow of 20 sccm. After treatment, the two layers were aligned (Fig 2H). The fluidic connections were punched with a biopsy tool, diameter of 1.2 mm. The hydrogel structure was then placed in the chamber in the control layer (Fig 2I) and the chip was bonded on top of a glass substrate (Fig 2J) via oxygen plasma activation (as above). Measuring setup The flow channel inlet was supplied with pressure by a pressure flow pump (AF-1, Elveflow, France), while the outlet was set to ambient pressure. Distilled water was used as the fluid. The flow rate was measured with a flow sensor providing a flow range from 0 μL min−1 to 50μL min−1 (Flow sensor 50 μL min−1, Elveflow, France). The control channel was supplied with a syringe pump (LA-100, HLL Landgraf Laborsysteme, Germany) at a constant flow of 15 μL min−1. For perfusing the control channel incorporating the hydrogel a solution of ethanol at various concentrations diluted in distilled water was used. The microfluidic chip system was placed on a tempering element connected to a circulation thermostat (Haake A10, Thermo Fischer Scientific Inc., USA) at a constant temperature for each particular experiment. The temperature compliance of the thermostat was regularly checked with a digital thermometer (Qtemp 600, VWR International GmbH). The chip was characterized in a stationary and dynamic manner, the dependent variable was always the flow rate. For the stationary characterization ethanol concentration, temperature and pressure were varied. The ethanol concentration the gel was exposed to was varied in a range from 0 wt% to 15.0 wt% in 2.5 wt% steps. The temperature was varied in a range from 10.0°C to 30.0°C in 2.5K steps. A home-made Python software was employed for controlling the pressure pump. The inlet pressure of the flow channel was varied in a range from 0mbar to 700mbar in 25mbar steps. During the measurement the pressure was applied for 5 s and then released for 15 s, this was repeated three times for each pressure step. For the dynamic characterization one operating point of ethanol concentration (cEth = 2.5wt%) and temperature (ϑ = 20.0°C) was chosen, and a sine pressure function was applied. The period of the sine function was varied, the amplitude of the sine was 50mbar, and the offset wa 400mbar. The measured data was analyzed and processed with home-made Python software (Python 2.7). In the dynamic investigation as pressure driven flow pump we used the system described in [34] in connection with a flow sensor (SLI-1000, Sensirion AG, Switzerland). Results and discussion As the chemo-fluidic membrane transistor is a novel microfluidic device, it is important to analyze its behavior to draw conclusions for its future applicability. Here we present and discuss a quantitative characterization of the device when controlled by a static source-drain-pressure (“output characteristic”) and a static chemical control concentration (“transfer characteristic”). In addition we show the dynamic reaction both to a change in pressure and in chemical concentration. We present a quasi-static model of the MIS-CVPT to be used for circuit design, and finally discuss the potential of the device within the field of microfluidics. Output characteristic Figs 3 and 4 show the output characteristic of the chemo-fluidic membrane transistor in terms of the flow rate QDS versus the excitation pressure pDS. In order to give an overview, Fig 4 shows the data over the entirety of the parameter field in four 3D-plots. 10.1371/journal.pone.0161024.g003Fig 3 3D-plot of measured output characteristics (flow rate QDS vs. pressure pDS vs. ethanol concentration cEth) of the MIS-CVPT at different temperatures ϑ ranging from 22.5°C to 30.0°C in 2.5K steps. 10.1371/journal.pone.0161024.g004Fig 4 Measured output characteristics (flow rate QDS vs. pressure pDS) of the MIS-CVPT with the different ethanol concentrations cEth 0wt%, 7.5wt%, and 15.0wt% at different temperatures ϑ ranging from 15.0°C to 30.0°C in 5.0K steps. The drain-source channel of the MIS-CVPT becomes conductive for pressures higher than the opening pressure which depends on both temperature and ethanol concentration. This opening pressure decreases with increasing temperature and increasing ethanol concentration. Fig 4 shows a selection of fewer data for better clarity. Here the graphs of the MIS-CVPT for ethanol concentration 0wt%, 7.5wt% and 15.0wt% over the temperatures ϑ from 15°C to 30°C are plotted in 5.0°C steps. The increase of temperature as well as the increase in concentration causes the gel to shrink, releasing the pressure off the membrane. In case of the curve for 0wt% ethanol concentration the opening pressure decreases T from 425mbar down to 100 mbar by a change of temperature of 15K. The curve for 15.0wt% ethanol concentration at 25°C remains the same for 30.0°C.This behavior can be explained by the fact that the gel is collapsed to a degree where it does not influence the switching of the membrane anymore. So for this part the mechanical properties of the doormat setup dominate the behavior of the chemical transistor completely, i.e. no chemical input is handled anymore, only the passive flow resistance remains. The device is in a state of overdrive, where an additional increase of concentration will not change the behavior. The curve for 7.5wt% at 30.0°C shows the same behavior. For the hydrogel a saturation of the ethanol concentration sets in. For every temperature step of 2.5K the next higher ethanol concentration in 2.5wt% steps reaches a saturation for the gel particle to shrink. When the device is in the overdriven state, the graphs of flow rate vs. pDS are the same, independent of temperature and concentration. In order to assure that the pressure drop in the control channel dos not influence the opening behaviour, the flow was limited to 15 μl min−1. We simulated the according pressure in an CFD simulation. The results presented in S2 Fig and S1 File show a neglectable magnitude of the pressure. At higher pressures, the output characteristic of the MIS-CVPT is expected to reach saturation with only linear increase of the flow rate. According to the design of the device, the opening between membrane and channel break has a limited size. This would lead to a constant conductance of the channel and hence a linear dependence of flow rate vs. pressure. In the current set-up the device characteristic could not yet be obtained for high values of pDS, since the flow sensor saturates at 50μL min−1. The expected saturation however would correspond to the saturation region in the output characteristic of the electronic MOSFET. Transfer characteristic The transfer characteristic of a MOSFET describes the relation of Drain-Source current IDS versus the Gate-Source voltage UGS. In the case of the MIS-CVPT this relation is described by the flow rate QDS versus the ethanol concentration cEth. The corresponding graphs for the transfer characteristic of the MIS-CVPT are shown in Fig 5 for several temperatures. The graphs are at constant pressure 100mbar pDS. The concentration determines the flow rate through the device. With rising ethanol concentration the MIS-CVPT becomes conductive. The temperature in this model can be used for biasing of the MIS-CVPT and hence corresponds to the bulk voltage Ub in a MOSFET. Fig 5 demonstrates, that the concentration at which the MIS-CVPT becomes conductive (threshold concentration) can be adjusted via the change of the temperature ϑ. Here an increase of 2.5K in temperature decreases the concentration at which the device opens by 2.5wt% ethanol. 10.1371/journal.pone.0161024.g005Fig 5 Transfer characteristic of the MIS-CVPT. Flow rate QDS over ethanol concentration cEth at a constant pressure pDS = 100mbar for the temperatures ϑ 22.5°C; 25.0°C; 27.5°C; 30.0°C. Dynamic behavior Against the background that hydrogels are perceived to work rather slowly and keeping in mind, that the switching speed determines the future application and acceptance of such a system, the dynamic behavior of the MIS-CVPT is of particular interest. Here, we analyze the dynamic response of the transistor to a change in chemical concentration. An analysis of the response to varying pressures (mechanical response) is presented in S3 Fig and S2 File. The approach of the dynamic investigation regarding a change of chemical concentration is shown in Fig 6. In the beginning of the experiment the transistor is closed. At time t0 (= 0 s) the ethanol concentration in the control channel is switched from c0 (= 0wt%) to c1 (= 30wt%). c1 was set to be 30wt% ethanol as the swelling curve of the hydrogel (S1 Fig) exhibits minimum volume at this concentration and therefore sets the transistor to overdrive. The flow channel of the transistor is fed with a constant pressure pDS of 400mbar and a flow rate sensor was installed following the chip. Over time the volume phase change transition takes place and the gel particle shrinks leading to the opening of the transistor.The flow rate in the flow channel reaches saturation plateau after some time. The dynamic investigation was conducted with three different transistor sizes. An exemplary measurement of the flow rate over time is shown in Fig 6B. In order to quantify the results three characterization parameters were derived and are marked in Fig 6C. Time t10 corresponds to 10% of the maximum flow rate QDS,max, time t90 corresponds to 90% of the maximum flow rate QDS,max, and the time interval Δt is the difference between t90 and t10. These three parameters indicate the response time of the system when it is exposed to a jump of ethanol from a low to a high concentration. 10.1371/journal.pone.0161024.g006Fig 6 Dynamic investigation of the gate switching. (A), Experimental approach of the investigation. Switch event at t0 carried out by a change of the input flow in the control channel from c0 (= 0wt%) to c1 (= 30wt%), while applying a constant pressure over the chemo-fluidic transistor. The flow rate generated in the flow channel is measured via a flow sensor. (B), Typical measurement of pressure and flow rate data over time, starting from time t0 (= 0s). (C), Schematic graph with characteristic parameters to quantify the experimental data. (D), Bar graph of the characteristic parameter t10, t90 and Δt for three different dimensions of gel particles. The results for the three tested transistor sizes (defined by the gel particle size L [μm] x B [μm]) over the characterization parameter t10, t90 and Δt are shown in Fig 6D. We systematically scaled the device from dimensions (1200 x 800) μm2 down to dimensions of (700 x 500) μm2 while keeping the height of the system constant at 100 μm. For a spherical hydrogel particle of radius r0 the characteristic reaction time can be estimated via τ=s2π2Dcoop,(1) where Dcoop is the cooperative diffusion coefficient, which is typically in the order of 10−11m2s−1 for neutral hydrogels [11]. As liquid supply to the gel is expected to occur mostly from the sides in the current valve set-up, the half-width B/2 is considered to be the characteristic length determining gel deswelling. Values of the reaction time estimated with Eq 1 and a value of Dcoop = 5 × 10−12m2s−1agree with the measured valve reaction times t90 within about 30%, if the radius of the gel r0 is substituted by B/2. Especially, the square dependence of reaction time on length scale is roughly shown in the values of t90. If the gel is in the shrunken state, a switch from high concentration c1 back to low concentration c0 will trigger gel swelling and hence the closing of the source-drain channel. Gel swelling is typically slower than gel shrinking (e.g. by a factor of three reported in [35]). Therefore a qualitatively similar, but accordingly slower response of the MIS-CVPT is expected for this inverse process. The diffusion-based transport of liquid in and out the gel allows a setting of the time scale by the spatial dimensions of the gel particle. For dimensions in the range of several 10 μm the time regime changes to a scale of seconds. Hence a faster response of the MIS-CVPT can be achieved simply by reducing its size. The dynamic switching of the MIS-CVPT and a MOSFET show similarities to an extent. In both cases an element is loaded (or unloaded) at a limited rate in the switching process. For the MOSFET this element is the gate electrode. Its capacitance for electric charge depends on size parameters (area of electrode, thickness of dielectric layer) and material parameters (dielectric constant). The charging rate is determined by the capacitance and the total resistance of the charging circuit, while the internal gate resistance and the gate capacitance of the MOSFET determine its maximum operating speed. For the MIS-CVPT, the element being loaded is the hydrogel particle, which is filled with or emptied of liquid. The loading rate is determined by the relaxation of the gel network which depends on geometric parameters as well as material parameters such as elastic moduli and polymer-solvent friction. Both MIS-CVPT and MOSFET exhibit a threshold in their transfer characteristic: In the MOSFET, the source-drain channel only becomes conductive, when the gate electrode has been charged above the threshold voltage of the transistor, while in the MIS-CVPT the hydrogel has to shrink below a certain threshold size to facilitate the opening of the channel at t10. Acting together, limited charging rate and threshold behavior cause a characteristic delay after the trigger signal until the source-drain channel becomes conductive in both devices. Compact Model Quasi-static model In order to simulate larger circuits with design automation tools, compact models of the components involved are needed. Here we present a physically motivated quasi-static compact model of the MIS-CVPT based on a fit to experimental data. Given the temperature, the chemical concentration in the control channel, the pressure in the control channel and the source-drain pressure, the model can be used to calculate the source-drain flowrate. The model is based on assigning a fluidic conductance G to the source-drain channel. Analogous to Ohm’s law in electronics the flowrate from source to drain can then be calculated via Q = GpDS, where pDS is the source-drain pressure. We distinguish two cases: Either the hydrogel is swollen and always touches the membrane exerting some pressure, or the hydrogel is collapsed. We first consider the former case. If the hydrogel were allowed to swell unobstructedly, it would reach a volume Vfree. Due to mechanical constraints the gel is restricted to a size Vgel. The pressure resulting from this compression can be roughly estimated via KVfree-VgelVfree, where K is the bulk modulus of the gel. If the membrane is not deflected yet (Vgel = Vcage), a threshold pressure p0 needs to be overcome to open the source-drain channel for conduction. With the effective pressure peff = (pdrain − psource)/2 − pc > p0 this conditition can be expressed as peff > p0, where pdrain is the pressure in the drain inlet, psource is the pressure in the source outlet and pc is the pressure in the control channel. For higher pressures the gel will be further compressed (with a linear dependence of volume on pressure in this approximation) and hence the membrane will be deflected further. The deflection of the membrane can lead to a deepening of the channel or a broadening of the channel. According to Hagen-Poiseuille’s law the conductivity of a wide rectangular channels shows a linear dependence on the width of the channel and a cubic dependence on the height of the channel. In the experiment, the dependence of G on peff showed to be quadratic to a good approximation, indicating a mixture of the two effects, see S4 Fig. The opening threshold pressure p0 as a function of temperature and chemical concentration was determined from the experimental data and fitted by a 2d polynomial (Fig 7). 10.1371/journal.pone.0161024.g007Fig 7 Opening pressure extracted from the measurement data and modeled as a 2nd order 2D polynomial. If the gel is in the shrunken state, the membrane can be deflected by very low pressure values until it touches the gel where higher pressures will be needed for further deflection. As this exact behavior is difficult to model we decided to fit the measured data by an adequate routine instead to obtain the parameters for the model. Here, a linear dependence of the conductance on peff showed good agreement with experimental data, see S4 Fig. In the current state, this model can be used to calculate steady-state phenomena by microfluidic circuit analysis tools. In future work the description of the transient behavior will be implemented as well, to allow the simulation of switching processes and oscillations. More details about the mathematical implementation of the model and a comparison of the fitted model with experimental data can be found in S3 File. Discussion Both microfluidics and electronics deal with signal propagation in circuits. If we consider pure laminar flows the mathematical isomorphism is given by the hydraulic analogy, equating pressure drop with voltage and flow rate with current. Fluidic resistances can be calculated by the Hagen-Poiseuille law, capacities stem from deflectable membranes and “inductances” come from the inertia of the fluid (which is usually negligible). However, the key difference between electronics and microfluidics is the existence of (one or more) chemical concentrations within the flow. In current approaches these concentrations are usually passively transported with the flow which is controlled by purely mechanical means. While the pressure and flow rate signals travel at the speed of sound, concentration propagation is governed by transport equations. This entails flow rate-dependent signal delays. These can prove limiting in some cases, but they can also be employed for tailored stimulus/reaction control and feedback mechanisms. The similarity of electronics and microfluidics allows the employment of tools from electronic design automation, albeit in a modified form to account for chemical signals [35]. In contrast to other actuation schemes, transistors based on smart hydrogels couple from the chemical to the fluidic domain. As fluidic signals can be converted back to chemical signals by well-designed mixing-junctions, this facilitates signal transduction purely based on chemical signals. In this domain of chemical signals, circuits know from electronics, such as oscillators, comparators, logic gates and flip-flops can be built. Since no external sensing and actuation units are used, there is no limitation due to the constraints implied by external connections of the microfluidic system. Therefore hydrogel-based transistors show a high potential for large-scale integration. In addition, chemical-fluidic coupling can be utilized for the smart control of chemical quantities that are difficult or impossible to sense otherwise by external sensing units. A specific potential of the chemo-fluidic membrane transistor lies in the separation of control and flow layer, and in the use of a channel break. The resulting output characteristic has a well-defined region of full closure of the source-drain channel, while the transfer characteristic shows a sharp response to a change in concentration. This is particularly important for digital applications and sensitive species detection. In our next plans, we will build elementary circuits, wherein the device is operated at drain pressure values below the threshold pressure of the swollen state, and switching between the blocked and conducting states is controlled purely by concentration in the control channel. The separation of control and flow layer is the most important feature of the MIS-CVPT. In contrast to the chemo-fluidic transistor used in [35] it facilitates a much more flexible setup of chemo-fluidic control operation. The controlling flow and the controlled flow can be kept separate, can be connected, and can even be swapped in a rather arbitrary fashion by the design of the microfluidic system. In a digital setup, the separation allows for chemical signal recovery without pressure build-up, elevating the limits for cascading of logic stages. These features pave the way for complex control schemes determined by programs coded by complex chemical signals. Conclusion Within this study we presented a microfluidic monolithically integrated transistor-like device that couples the chemical with the fluidic domain. The transfer and output characteristics of the device show that the chemo-fluidic membrane transistor is able to process both discrete and analogue chemical signals. Any thermodynamic parameter able to provoke a volume phase transition of stimuli-responsive hydrogels within an aqueous solutions, for example concentration of organic solvents such as alcohols, ions (e.g. of salts), hydrogen (pH-value) and biomolecules like glucose can be coupled with the fluidic domain via chemo-fluidic transistors. For a clearer understanding the article solely covers the characterization of the device with ethanol concentration as model system. Additionally the threshold concentration of the MIS-CVPT can be remotely controlled by the threshold temperature ϑth. As we aim for a rethinking of the LoC concept, we are quite aware of the importance of the dynamic behavior of such a technology in order to gain acceptance. Even though we realize the large time scale of the response, our results clearly state, that the dynamics of the device can be improved by merely scaling down the technology. By focussing our efforts on conquering the next technology node for this platform, we are confident to sustainably speed up the dynamic behavior. As we have moved forward in our concept we have developed a series of devices contributing to the chemo-fluidic circuit concept. E.g. a chemo-fluidic delay line oscillator [35], which was recently published by Paschew et. al., allows the bidirectional coupling of the chemical and the fluidic domain introducing event-based control on the component level. This offers possibilities that exceed the capabilities of present-day microfluidics. The report of the chemo-fluidic oscillator also addressed the crucial importance of the utilization of computer aided design (CAD) in microfluidics. The compact model in this report follows this pave and aims for a conclusive development process incorporating CAD, modeling and simulation. Another approach demonstrated the autonomous and purely fluidic transistor-based control in contrast to the current dominant concept of computer controlled labs-on-chips. A proof of principle of complex circuit programs based on chemo-fluidic components has been shown by Greiner [36]. The next challenge that arises is the development of chemo-fluidic circuits for a more sophisticated on-chip flow control. Here discrete circuits such as logic gates but also analogue approaches seem feasible. By combining the efforts of technological downscaling with model-based computer aided design we hope to enhance the potential of the LoC platform by the effects of integration and automation known from electronics. Supporting Information S1 Fig Swelling Graph. Swelling behaviour of pure PNIPAAm and a co-polymerization of PNIPAAm with a 2.5% fraction of sodium acrylate over a series of ethanol concentrations (0wt% to 100.0wt%). (PDF) Click here for additional data file. S1 File Control channel CFD simulation. Analysis of the control channel CFD simulation. (PDF) Click here for additional data file. S2 Fig Control channel CFD simulation. CFD simulation in the control channel. Target value is the pressure within the hydrogel seat. (left)—Simulation without hydrogel particle. (right)—Simulation with swollen hydrogel particle. (PDF) Click here for additional data file. S2 File Dynamic mechanical investigation. Analysis of the dynamic mechanical investigation. (PDF) Click here for additional data file. S3 Fig Dynamic mechanical investigation. Dynamic mechanical behavior of the MIS-CVPT. Normalized amplitude of the flow rate and the exciting pressure over the frequency. (PDF) Click here for additional data file. S3 File Data analysis and model fit. Explanation of the data analysis and model fit. (PDF) Click here for additional data file. S4 Fig Data analysis and model fit. Exemplary plot of data points of the flow rate over pressure and ethanol concentration at 25°C with the according model fit. (PDF) Click here for additional data file. The authors acknowledge the support of Cesare Pini for overseeing the design process and contributing his expertise. Marcus Tietze is thanked for his support in polymer synthesis. Further we thank Dr. Hielscher from the packaging laboratories at Technische Universität Dresden (IAVT, TU Dresden) for his support in the mask technology. We also thank the German Research Foundation DFG for financing the research within the Research Training Group (RTG) 1865 “Hydrogel-Based Microsystems”, the Center for Advancing Electronics Dresden (cfaed), and the Free State Saxony for funding the research group “Diagnostic Integrated Chemical Circuits” for financial support. ==== Refs References 1 Folch A . Introduction to BioMEMS . CRC Press ; 2012 Available from: https://books.google.com/books?hl=de&lr=&id=ejrSBQAAQBAJ&pgis=1. 2 Thorsen T , Maerkl SJ , Quake SR . Microfluidic Large-Scale Integration . Science . 2002 ;298 (5593 ):580 –584 . Available from: http://www.sciencemag.org/content/298/5593/580.abstract. 10.1126/science.1076996 12351675 3 Jensen EC , Grover WH , Mathies RA . Micropneumatic digital logic structures for integrated microdevice computation and control . J Microelectromech S . 2007 ;16 (6 ):1378 –1385 . 10.1109/JMEMS.2007.906080 4 Mosadegh B , Kuo CH , Tung YC , Torisawa YS , Bersano-Begey T , Tavana H , et al Integrated Elastomeric Components for Autonomous Regulation of Sequential and Oscillatory Flow Switching in Microfluidic Devices . Nat Phys . 2010 6 ;6 (6 ):433 –437 . Available from: http://www.nature.com/nphys/journal/v6/n6/pdf/nphys1637.pdf. 10.1038/nphys1637 20526435 5 Mosadegh B , Bersano-Begey T , Park JY , Burns Ma , Takayama S . Next-generation integrated microfluidic circuits . Lab Chip . 2011 ;11 (17 ):2813 –2818 . 10.1039/c1lc20387h 21799977 6 Kim SJ , Yokokawa R , Lesher-Perez SC , Takayama S . Constant flow-driven microfluidic oscillator for different duty cycles . Anal Chem . 2012 ;84 (2 ):1152 –1156 . 10.1021/ac202866b 22206453 7 Unger MA , Chou HP , Thorsen T , Scherer A , Quake SR . Monolithic Microfabricated Valves and Pumps by Multilayer Soft Lithography . Science . 2000 ;288 (5463 ):113 –116 . Available from: http://www.sciencemag.org/content/288/5463/113.abstract. 10.1126/science.288.5463.113 10753110 8 Baek JY , Park JY , Ju JI , Lee TS , Lee SH . A pneumatically controllable flexible and polymeric microfluidic valve fabricated via in situ development . Journal of Micromechanics and Microengineering . 2005 ;15 (5 ):1015 Available from: http://stacks.iop.org/0960-1317/15/i=5/a=017. 10.1088/0960-1317/15/5/017 9 Wang J , Chen Z , Mauk M , Hong KS , Li M , Yang S , et al Self-actuated, thermo-responsive hydrogel valves for lab on a chip . Biomedical microdevices . 2005 ;7 (4 ):313 –322 . 10.1007/s10544-005-6073-z 16404509 10 Gu W , Chen H , Tung YCC , Meiners JC , Takayama S . Multiplexed hydraulic valve actuation using ionic liquid filled soft channels and Braille displays . Applied Physics Letters . 2007 ;90 (3 ):157 –173 . Available from: http://scitation.aip.org/content/aip/journal/apl/90/3/10.1063/1.2431771. 10.1063/1.2431771 11 Richter A , Paschew G , Klatt S , Lienig J , Arndt KF , Adler HJP . Review on Hydrogel-based pH Sensors and Microsensors . Sensors . 2008 ;8 (1 ):561 –581 . 10.3390/s8010561 27879722 12 Siegel RA , Firestone BA . pH-dependent equilibrium swelling properties of hydrophobic polyelectrolyte copolymer gels . Macromolecules . 1988 11 ;21 (11 ):3254 –3259 . Available from: 10.1021/ma00189a021 . 10.1021/ma00189a021 13 Liu K , Ji HF . Detection of Pb2+ using hydrogel swelling microcantilever sensor . Anal Sci . 2004 ;20 (January ):9 –11 . 10.2116/analsci.20.9 14753251 14 Zhang Y , Ji HF , Brown GM , Thundat T . Detection of CrO4 2- using a hydrogel swelling microcantilever sensor . Anal Chem . 2003 ;75 (18 ):4773 –4777 . 10.1021/ac0343026 14674453 15 Irie M , Misumi Y , Tanaka T . Stimuli-responsive polymers: chemical induced reversible phase separation of an aqueous solution of poly(N-isopropylacrylamide) with pendent crown ether groups . Polymer . 1993 11 ;34 (21 ):4531 –4535 . Available from: http://www.sciencedirect.com/science/article/pii/003238619390160C. 10.1016/0032-3861(93)90160-C 16 Tanaka T , Wang C , Pande V , Grosberg AY , English A , Masamune S , et al Polymer gels that can recognize and recover molecules . Faraday Discuss . 1995 1 ;101 :201 Available from: http://pubs.rsc.org/en/Content/ArticleLanding/1995/FD/FD9950100201. 10.1039/fd9950100201 17 Li W , Zhao H , Teasdale PR , John R , Zhang S . Synthesis and characterisation of a polyacrylamide—polyacrylic acid copolymer hydrogel for environmental analysis of Cu and Cd . React Funct Polym . 2002 ;52 :31 –41 . 10.1016/S1381-5148(02)00055-X 18 Tanaka T , Nishio I , Sun ST , Ueno-Nishio S . Collapse of gels in an electric field . Science (New York, NY) . 1982 10 ;218 (4571 ):467 –9 . Available from: http://www.sciencemag.org/content/218/4571/467.short. 10.1126/science.218.4571.467 19 Suzuki A , Tanaka T . Phase transition in polymer gels induced by visible light . Nature . 1990 7 ;346 (6282 ):345 –347 . Available from: 10.1038/346345a0 . 10.1038/346345a0 20 Tanaka T . Collapse of gels and the critical endpoint . Phys Rev Lett . 1978 ;40 (12 ):820 –823 . 10.1103/PhysRevLett.40.820 21 Richter A , Kretschmer K , Wenzel J . Mechanically adjustable chemostats based on stimuli-responsive polymers . Sensor Actuat B-Chem . 2007 ;125 (2 ):569 –573 . Available from: http://linkinghub.elsevier.com/retrieve/pii/S0925400507001669. 10.1016/j.snb.2007.03.002 22 Kuckling D , Adler HjP , Ling L , Habicher WD . Temperature sensitive polymers based on 2- (dimethyl maleinimido)—N -ethyl-acrylamide:Copolymers with N -isopropylacrylamide . Polym Bull . 2000 ;276 :269 –276 . 10.1007/s002890050602 23 Haefner S , Rohn M , Frank P , Paschew G , Elstner M , Richter A . Improved PNIPAAm-Hydrogel Photopatterning by Process Optimisation with Respect to UV Light Sources and Oxygen Content . MDPI Gels . 2016 ;2 (1 ):10 Available from: http://www.mdpi.com/2310-2861/2/1/10. 10.3390/gels2010010 24 Beebe DJ , Moore JS , Bauer JM , Yu Q , Liu RH , Devadoss C , et al Functional hydrogel structures for autonomous flow control inside microfluidic channels:Abstract:Nature . Nature . 2000 ;404 (6778 ):588 –590 . Available from: http://www.nature.com/doifinder/10.1038/35007047∖npapers3://publication/doi/10.1038/35007047. 10.1038/35007047 10766238 25 Eddington D . Flow control with hydrogels . Adv Drug Deliver Rev . 2004 ;56 (2 ):199 –210 . Available from: http://www.sciencedirect.com/science/article/pii/S0169409X03001947. 10.1016/j.addr.2003.08.013 26 Gu Y, Baldi A, Ziaie B, Siegel RA. Modulation of drug delivery rate by hydrogel-incorporating MEMS devices. 2nd Annual International IEEE/EMBS Conference on Microtechnologies in Medicine and Biology. 2002;p. 406–409. Available from: http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=1002356. 27 Liu RH , Beebe DJ . Fabrication and characterization of hydrogel-based microvalves . J Microelectromech S . 2002 ;11 (1 ):45 –53 . Available from: http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=982862. 10.1109/84.982862 28 Arndt KF , Kuckling D , Richter A . Application of Sensitive Hydrogels in Flow Control . Polym Adv Technol . 2000 ;11 :496 –505 . 10.1002/1099-1581(200008/12)11:8/12<496::AID-PAT996>3.3.CO;2-Z 29 Grover WH , Skelley AM , Liu CN , Lagally ET , Mathies Ra . Monolithic membrane valves and diaphragm pumps for practical large-scale integration into glass microfluidic devices . Sensor Actuat B-Chem . 2003 ;89 (3 ):315 –323 . 10.1016/S0925-4005(02)00468-9 30 Hosokawa K , Maeda R . A pneumatically-actuated three-way microvalve fabricated with polydimethylsiloxane using the membrane transfer technique . J Micromech Microeng . 2000 ;10 (3 ):415 –420 . 10.1088/0960-1317/10/3/317 31 Li N , Hsu CH , Folch A . Parallel mixing of photolithographically defined nanoliter volumes using elastomeric microvalve arrays . Electrophoresis . 2005 ;26 (19 ):3758 –3764 . 10.1002/elps.200500171 16196107 32 Richter A , Türke A , Pich A . Controlled Double-Sensitivity of Microgels Applied to Electronically Adjustable Chemostats . Adv Mater . 2007 ;19 (8 ):1109 –1112 . Available from: http://doi.wiley.com/10.1002/adma.200601989. 10.1002/adma.200601989 33 Huang H , Zhang D , Mao Z , Yu W , Yan H . Fast responsive and strong swelling hydrogels based on N -isopropylacrylamide with sodium acrylate . J Appl Polym Sci . 2009 4 ;112 (1 ):123 –128 . Available from: http://doi.wiley.com/10.1002/app.28457. 10.1002/app.28457 34 Frank P , Haefner S , Elstner M , Richter A . Fully-Programmable, Low-Cost, “Do-It-Yourself” Pressure Source for General Purpose Use in the Microfluidic Laboratory . MDPI Inventions 2016, Vol 1, Page 13. 2016 ;1 (2 ):13 . 35 Paschew G , Schreiter J , Voigt A , Pini C , Chávez JP , Allerdißen M , et al Autonomous Chemical Oscillator Circuit Based on Bidirectional Chemical-Microfl uidic Coupling . Advanced Materials Technology . 2016 ;p. 1 –6 . 36 Greiner R , Allerdissen M , Voigt A , Richter A . Fluidic microchemomechanical integrated circuits processing chemical information . Lab Chip . 2012 ;12 (23 ):5034 –5044 . 10.1039/c2lc40617a 23038405

PMC005xxxxxx/PMC5003341.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757126710.1371/journal.pone.0161827PONE-D-16-10125Research ArticleBiology and Life SciencesOrganismsPlantsLegumesPeasBiology and Life SciencesAnatomyCardiovascular AnatomyBlood VesselsArteriesPulmonary ArteriesMedicine and Health SciencesAnatomyCardiovascular AnatomyBlood VesselsArteriesPulmonary ArteriesMedicine and Health SciencesSurgical and Invasive Medical ProceduresBiopsyResearch and Analysis MethodsImaging TechniquesNeuroimagingComputed Axial TomographyBiology and Life SciencesNeuroscienceNeuroimagingComputed Axial TomographyMedicine and Health SciencesDiagnostic MedicineDiagnostic RadiologyTomographyComputed Axial TomographyResearch and Analysis MethodsImaging TechniquesDiagnostic RadiologyTomographyComputed Axial TomographyMedicine and Health SciencesRadiology and ImagingDiagnostic RadiologyTomographyComputed Axial TomographyMedicine and Health SciencesPharmacologyDrugsVasodilatorsPhysical SciencesChemistryChemical ElementsOxygenMedicine and Health SciencesVascular MedicineBlood PressureResearch and Analysis MethodsMathematical and Statistical TechniquesStatistical MethodsAnalysis of VariancePhysical SciencesMathematicsStatistics (Mathematics)Statistical MethodsAnalysis of VarianceSevere Pulmonary Arteriopathy Is Associated with Persistent Hypoxemia after Pulmonary Endarterectomy in Chronic Thromboembolic Pulmonary Hypertension Pulmonary Arteriopathy and Hypoxemia after PEA in CTEPHhttp://orcid.org/0000-0001-5725-1810Jujo Takayuki 12*Tanabe Nobuhiro 12Sakao Seiichiro 1Ishibashi-Ueda Hatsue 3Ishida Keiichi 4Naito Akira 1Kato Fumiaki 1Takeuchi Takao 1Sekine Ayumi 12Nishimura Rintaro 12Sugiura Toshihiko 1Shigeta Ayako 1Masuda Masahisa 5Tatsumi Koichiro 11 Department of Respirology (B2), Graduate School of Medicine, Chiba University, 1-8-1, Inohana, Chuo-Ku, Chiba City, 260–8670, Japan2 Department of Advanced Medicine in Pulmonary Hypertension, Graduate School of Medicine, Chiba University, 1-8-1 Inohana, Chuo-Ku, Chiba City, 260–8670, Japan3 Department of Pathology, National Cerebral and Cardiovascular Center, 5-7-1, Fujishiro-Dai, Suita City, Osaka, 565–8565, Japan4 Department of Cardiovascular Surgery, Graduate School of Medicine, Chiba University, 1-8-1 Inohana, Chuo-Ku, Chiba City, 260–8670, Japan5 Department of Cardiovascular Surgery, Chiba Medical Center, National Hospital Organization, 4-1-2, Tsubakimori, Chuo-ku, Chiba City, 260–8606, JapanKuwana Masataka EditorProfessor, JAPANCompeting Interests: I have read the journal's policy and the authors of this manuscript have the following competing interests: T.J, N.T, A.S. and R.N. were members of an endowed department sponsored by Actelion Pharmaceuticals; N.T received lecture honoraria from Bayer Daiichi-Sankyo Actelion; K.T received lecture honoraria from GlaxoSmithKline and Pfizer. The other authors had no conflict of interest. This does not alter the authors' adherence to PLOS ONE policies on sharing data and materials. Conceptualization: TJ NT SS. Formal analysis: TJ NT AN FK TT AS RN TS AS MM KT. Investigation: TJ AN. Methodology: TJ NT SS HI-U. Resources: HI-U KI TS MM. Writing – original draft: TJ NT. Writing – review & editing: TJ NT SS HI-U KI AN FK TT AS RN TS AS MM KT. * E-mail: naikamo_resp19184@chiba-u.jp29 8 2016 2016 11 8 e01618279 3 2016 14 8 2016 © 2016 Jujo et al2016Jujo et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Background Chronic thromboembolic pulmonary hypertension (CTEPH) is characterized by occlusion of pulmonary arteries by organized chronic thrombi. Persistent hypoxemia and residual pulmonary hypertension (PH) following successful pulmonary endarterectomy (PEA) are clinically important problems; however, the underlying mechanisms remain unclear. We have previously reported that residual PH is closely related to severe pulmonary vascular remodeling and hypothesize that this arteriopathy might also be involved in impaired gas exchange. The purpose of this study was to evaluate the association between hypoxemia and pulmonary arteriopathy after PEA. Methods and Results Between December 2011 and November 2014, 23 CTEPH patients underwent PEA and lung biopsy. The extent of pulmonary arteriopathy was quantified pathologically in lung biopsy specimens. We then analyzed the relationship between the severity of pulmonary arteriopathy and gas exchange after PEA. We observed that the severity of pulmonary arteriopathy was negatively correlated with postoperative and follow-up PaO2 (postoperative PaO2: r = -0.73, p = 0.0004; follow-up PaO2: r = -0.66, p = 0.001), but not with preoperative PaO2 (r = -0.373, p = 0.08). Multivariate analysis revealed that the obstruction ratio and patient age were determinants of PaO2 one month after PEA (R2 = 0.651, p = 0.00009). Furthermore, the obstruction ratio and improvement of pulmonary vascular resistance were determinants of PaO2 at follow-up (R2 = 0.545, p = 0.0002). Severe pulmonary arteriopathy might increase the alveolar-arterial oxygen difference and impair diffusion capacity, resulting in hypoxemia following PEA. Conclusion The severity of pulmonary arteriopathy was closely associated with postoperative and follow-up hypoxemia. the Respiratory Failure Research Group from the Ministry of Health, Labor and Welfare of JapanTatsumi Koichiro the Pulmonary Hypertension Research Group from Japan Agency for Medical Research and Development, AMEDTatsumi Koichiro T.J was supported by The Cardiovascular Research Fund. N.T was supported by a research grant from the Ministry of Education, Culture, Sports, Science and Technology of Japan (25461148), and a grant from the Ministry of Health, Labour and Welfare specifically designated for the Respiratory Failure Research Group and Cardiovascular Diseases. S.S was supported by research grants from the Respiratory Failure Research Group from the Ministry of Health, Labour and Welfare, Japan, a grant to the Pulmonary Hypertension Research Group (No. 27280401) from the Japan Agency for Medical Research and Development, AMED, and a Grant-in-Aid for Scientific Research (JSPS KAKENHI Grant Number 15K09210) from the Japanese Ministry of Education and Science; K.I was supported by a Grant-in-Aid for Scientific Research from the Japan Society for the Promotion of Science; K.T was supported by Japanese Ministry of Health, Labour and Welfare research grants specifically designated for the Respiratory Failure Research Group and Cardiovascular Diseases. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityRelevant summary data are within the paper and its Supporting Information files. However, all underlying data cannot be made publicly available due to participant privacy. Additional data are available upon request from the corresponding author, Takayuki Jujo (naikamo_resp19184@chiba-u.jp).Data Availability Relevant summary data are within the paper and its Supporting Information files. However, all underlying data cannot be made publicly available due to participant privacy. Additional data are available upon request from the corresponding author, Takayuki Jujo (naikamo_resp19184@chiba-u.jp). ==== Body Introduction Chronic thromboembolic pulmonary hypertension (CTEPH) is characterized by occlusion of pulmonary arteries by organized chronic thrombi [1,2]. It is thought that inappropriate resolution of acute pulmonary embolism results in the formation of persistent chronic thrombi leading to pulmonary hypertension [3]. Pulmonary endarterectomy (PEA) is a surgical procedure in which the organized pulmonary thrombi are removed. PEA has been shown to improve hemodynamics, cardiac function, and the six-minute walk distance in patients with CTEPH [4–6]. However, following PEA, many patients experience persistent hypoxemia and residual pulmonary hypertension (PH). Hypoxemia and residual PH have been observed in 50–60% [7] and 5–35% [8] of CTEPH patients, respectively, following successful PEA and are closely related to poor functional capacity (NYHA functional class III-IV) [9]. Ventilation-perfusion mismatch has been postulated as a major contributor to hypoxemia in CTEPH patients [10,11]. However, the anatomical location and the pathophysiology underlying this finding remain unclear. We recently reported that residual PH was associated with severe pulmonary arteriopathy [12] and hypothesized that severe pulmonary vascular remodeling might also be involved in impaired gas exchange (Fig 1). Thus, the purpose of this study was to evaluate the relationship between pulmonary arteriopathy and gas exchange following PEA. 10.1371/journal.pone.0161827.g001Fig 1 Pulmonary arteriopathy in biopsied lung tissues. Severe pulmonary arteriopathy in the high-obstruction group (A). Pulmonary arteriopathy was composed of severe fibrous intimal thickening, moderate medial hypertrophy, and lumen stenosis. The low-obstruction group (B) demonstrated mild pulmonary intimal thickening and medial hypertrophy. Methods Ethics statement This study was approved by the Ethics Committee of Chiba University (approval number: 1221). Written informed consent for inclusion in this study was obtained from all patients. Subjects Thirty CTEPH patients underwent PEA between December 2011 and November 2014 in our institutions. Lung biopsy was performed in 25 patients. Two of these patients were excluded from the study because the specimens were too small to yield adequate slices for examination. Thus, lung tissues derived from 23 patients were evaluated in this study. All subjects were examined and diagnosed with CTEPH at Chiba University Hospital, which is located about 20 meters above sea level. The diagnosis was confirmed by right heart catheterization (RHC), pulmonary angiography (PAG), lung perfusion scans, and enhanced computed tomography (CT). The criteria for CTEPH diagnosis were as follows: 1) elevation of mean pulmonary arterial pressure (mean Ppa ≥ 25 mmHg) with normal pulmonary arterial wedge pressure (PAWP); 2) symptoms for more than six months; 3) persistent pulmonary embolism confirmed by lung perfusion scans, CT scans and PAGs. General examinations, respiratory function tests, CT scans, RHC, and PAG were performed during the preoperative, postoperative (about one month after surgery), and follow-up periods. Follow-up evaluation was performed 13.1±2.4 months after PEA. For one female patient, postoperative and follow-up CT scans were not obtained due to an anaphylactic reaction to contrast medium. This patient also refused postoperative RHC. RHC RHC was performed in a similar method as described in our previous report [12]. Briefly, Ppa and PAWP were measured using a 7.5-Fr Swan-Ganz catheter (Edwards Lifesciences, USA). Cardiac index (CI) was calculated from cardiac output (CO) measured by thermodilution. Pulmonary vascular resistance (PVR) was calculated according to the following formula: PVRdyne·sec·cm−5=meanPpa−PAWPCO×80 Blood gas analysis Blood samples from the pulmonary artery (mixed venous blood) and a peripheral systemic artery were obtained at the time of pulmonary pressure measurement, while each patient was breathing room air. The partial pressures of oxygen and carbon dioxide in systemic blood (PaO2 and PaCO2, respectively) and the partial pressure of oxygen in mixed venous blood (PvO2) were directly measured by a blood gas analyzer (ABL-555, Radiometer Medical Aps, Copenhagen, Denmark). The alveolar-arterial oxygen difference (A-aDO2) was calculated according to the following formula [13]: A−aDO2mmHg=FIO2Patm−Pvapor−PaCO2R+FIO2×PaCO2×1−RR−PaO2 in which FIO2 is the fraction of inspired oxygen (0.209), Patm is the atmospheric pressure (760 mmHg), Pvapor is the saturated water vapor pressure (47 mmHg), and R is the respiratory quotient (0.83). Two blood samples obtained from two female patients during the postoperative period were excluded because they showed a negative A-aDO2, which was considered a technical error resulting from a lack of oxygen washout before the measurement. PEA PEA was performed by Drs. K.I and M.M at Chiba University Hospital and the Chiba Medical Center. The indications for PEA were similar to those described in our previous report [12]: 1) mean Ppa > 30 mmHg, 2) PVR > 300 dyne·sec·cm-5, 3) WHO functional class ≥ 2, 4) technically operable, and 5) no significant comorbidity. In patients with mild PH (mean Ppa 25–30 mmHg), PEA was performed only if the patient wished to undergo the procedure after the risks and benefits had been fully explained. Lung biopsy and sample preparation Lung biopsy was performed during PEA. The lung biopsy technique and pathological evaluation were similar to our previous report [12]. Lung tissue was resected from the right middle lobe or left lingular segment. Lung biopsy was limited to a single site for ethical and safety reasons. The selection of the areas for biopsy was left to the surgeon’s discretion. The biopsy sites were selected by considering safety, and the selection was unrelated to removal of thrombi. Samples were fixed in 10% buffered formalin and embedded in paraffin. More than 20 sections obtained at intervals of at least 20 μm were prepared for each patient. A total of 555 slices were prepared. The sections were stained with Elastica van Gieson stain for quantification. Pathological quantification of pulmonary arteriopathy Pathological quantification of pulmonary arteriopathy was performed according to the methods described in our previous report [12]. Details are shown in S1 Fig. The obstruction ratio of a pulmonary artery was defined as the ratio of the luminal and medial area to the vascular area; the obstruction ratio was calculated for each pulmonary artery using Image J software (version 1.45). Pathological findings were reviewed by a trained pathologist and a pulmonologist. The mean obstruction ratio for each patient was defined as the average value of all obstruction ratios in the tissue sample. The patients were divided into two groups, the high-obstruction group and the low-obstruction group, based on the median value of all mean obstruction ratios. Evaluation of segmental pulmonary thrombi Chronic thrombi within segmental pulmonary arteries were quantified by enhanced CT scanning according to the modified methods of Qanadli [14]. CT scans were performed using the Aquilion ONE (Toshiba Medical, Tochigi, Japan). Contrast material consisting of 100 ml of 350 mg/ml of iodine was injected using a mechanical injector. CT scan images were obtained using a tube voltage of 120 kV and a scanning delay of 20–30 seconds. Each image was 0.5mm thick. Segmental pulmonary arteries were defined according to Boyden’s classification [15] and consisted of ten right-sided and eight left-sided segmental pulmonary arteries. Eighteen segmental pulmonary arteries were identified on each CT scan image and each segmental artery was evaluated over the entire series of images. Each segmental pulmonary artery was scored as follows: score 0: no thrombi; score 1: the artery was narrowed by chronic thrombi but contrast medium passed to distal areas; and score 2: the artery was obstructed by chronic thrombi and contrast medium did not pass to distal areas. The segmental obstruction index (SGOI) was calculated by the following equation: SGOI=0×n0+1×n1+2×n236 in which n0, n1, and n2 are the numbers of segmental pulmonary arteries with score 0, score 1, and score 2, respectively. We also evaluated the presence or absence of thrombi in the biopsied area. Two segmental pulmonary arteries (A4, A5) in the biopsied side were scored. The segmental obstruction index at the biopsied site (SGOIbiopsy) was calculated by the following equation: SGOIbiopsy=0×n0+1×n1+2×n22 Two investigators interpreted a total of 64 CT scan images in a blinded manner (T.J and A.N). The SGOI and SGOIbiopsy scores obtained by the two independent investigators were significantly correlated (SGOI: r = 0.917, p = 1.7×10−26; SGOIbiopsy: r = 0.803, p = 1.4×10−15). To minimize bias, SGOI and SGOIbiopsy scores for each patient were defined as the average value between the two investigators. Statistical analysis Continuous variables are reported as the mean ± SD unless otherwise indicated. The correlation between variables was determined by Spearman's rank correlation coefficient. Multiple comparisons between more than three groups were performed by one-way ANOVA with the adjustment of p-value by the Bonferroni method. Pairwise comparisons of time-dependent data were made by Repeated-Measures ANOVA with the adjustment of p-value by the Bonferroni method. The change of vasodilator usage rate was analyzed by McNemar's Chi-squared test. Simple and multiple linear regression analyses were used to determine the factors associated with PaO2. A p-value of < 0.05 was considered significant. All data were analyzed using EZR (ver. 1.29, Saitama Medical Center, Jichi Medical University, Saitama, Japan) [16]. Results Clinical background of subjects The 23 study patients included 4 males and 19 females. The mean age at diagnosis was 64.7 ± 8.1 years old. The mean period from symptom onset to PEA was 53.7 ± 52.8 months. Nineteen of 23 patients (82.6%) had a history of deep vein thrombosis. Lung tissues were biopsied from the right middle lobe of five patients and from the left lingular segment of eighteen patients. One patient developed sudden respiratory failure of unknown cause and died 15 days after PEA as described previously [12]. Including this patient, no complication related to the lung biopsies was observed. Data from the three observational periods are shown in Table 1 and S1 Table. Following PEA, the burden of segmental organized thrombi was significantly reduced, and gas exchange parameters and hemodynamics improved. The follow-up PaO2 was significantly higher than the preoperative PaO2 (p = 0.0002). The postoperative PaO2 was higher than the preoperative PaO2, although the difference was not significant (p = 0.1). 10.1371/journal.pone.0161827.t001Table 1 Patient characteristics. Preoperative Postoperative Follow-up Segmental pulmonary thrombi (n = 23) (n = 21) (n = 20) SGOI 0.35 ± 0.15 0.22 ± 0.15* 0.21 ± 0.16*§ Hemodynamics (n = 23) (n = 21) (n = 22) Mean Ppa (mmHg) 45.1 ± 10.6 26.9 ± 9.3* 28.1 ± 8.7* PVR (dyne·sec·cm-5) 767 ± 299 335 ± 166* 374 ± 183* CI (L/min/m2) 2.85 ± 0.79 3.15 ± 0.47 2.83 ± 0.62§ PAWP (mmHg) 7.9 ± 2.8 8.4 ± 3.3 9.0 ± 2.5 Blood gas analysis (n = 23) (n = 19) (n = 22) PaO2 (mmHg) 57.0 ± 10.1 61.5 ± 14.1 71.3 ± 15.1*§ PaCO2 (mmHg) 37.2 ± 3.4 41.3 ± 3.1* 41.5 ± 3.7* PvO2 (mmHg) 33.8 ± 3.7 34.1 ± 3.3 37.2 ± 3.2*§ A-aDO2 (mmHg) 48.8 ± 9.2 38.8 ± 14.8* 29.7 ± 14.1*§ Respiratory function test (n = 22) (n = 21) (n = 22) %VC (%) 88.7 ± 12.4 78.1 ± 10.1* 90.2 ± 13.9§ FEV1.0/FVC (%) 76.3 ± 7.7 79.4 ± 7.4 76.1 ± 7.0 %DLCO/VA (%) 80.4 ± 14.7 71.7 ± 11.2* 74.9 ± 15.9 Vasodilators 14/23 3/22* 7/22* Shunt ratio by perfusion scan (%) 6.4±2.2 (n = 10) 6.5±0.8 (n = 5) 7.3±3.8 (n = 4) SGOI: Segmental obstruction index, Ppa: pulmonary arterial pressure; PVR: pulmonary vascular resistance; CI: cardiac index; PAWP: pulmonary arterial wedge pressure; PaO2: arterial oxygen pressure; PaCO2: arterial carbon dioxide pressure; PvO2: oxygen pressure of mixed venous blood; A-aDO2: alveolar-arterial oxygen difference; %VC: vital capacity as the percent predicted; FEV1.0: forced expiratory volume (FEV) in 1 second; FVC: forced vital capacity; DLCO: diffusing capacity of carbon monoxide; VA: alveolar ventilation. *: p<0.05 vs preoperative §: p<0.05 vs postoperative. The relationship between pulmonary arteriopathy and PaO2 after PEA Quantitative data were obtained from 354 pulmonary muscular arteries in the 23 patients in this study. Pulmonary arteriopathy with intimal thickening and medial hypertrophy was observed in samples from all patients (Fig 1). The mean obstruction ratio was 0.838±0.139 and the median value of the mean obstruction ratios was 0.863933. No difference in the obstruction ratio was observed between biopsy sites (left: 0.837±0.146; right: 0.843±0.128, p = 0.9). The mean obstruction ratio was negatively correlated with the postoperative and follow-up PaO2 (Fig 2A and 2B), but was less correlated with the preoperative PaO2 (Fig 2C). The high obstruction group (mean obstruction ratio ≥0.863933) had lower PaO2 values than the low-obstruction group (mean obstruction ratio <0.863933) (p = 0.009, analyzed by repeated measures ANOVA, Fig 3). 10.1371/journal.pone.0161827.g002Fig 2 Correlation between mean obstruction ratio and PaO2. The mean obstruction ratio was negatively correlated with postoperative (A) and follow-up (B) PaO2. However, no correlation was observed between the mean obstruction ratio and preoperative PaO2 (C). 10.1371/journal.pone.0161827.g003Fig 3 PaO2 in the high- and low-obstruction groups. The high obstruction group had lower PaO2 values than the low-obstruction group (p = 0.009, analyzed by repeated measures ANOVA). (*: p<0.05, vs preoperative data for each group; §: p<0.05, vs postoperative data for each group). Segmental thrombi and gas exchange The macroscopic classification of resected segmental organized thrombi is known as the “Jamieson’s classification” [4]. Samples were classified as follows, Type I, 9 cases; Type II, 9 cases, and Type III, 5 cases. The preoperative SGOI for each Jamieson’s classification was shown as follows, Type I, 0.44 ± 0.14; Type II, 0.33 ± 0.13; and Type III, 0.24 ± 0.14. There were no significant differences in SGOI among the three classes (p = 0.0501, analyzed by one-way ANOVA). The SGOI significantly decreased after PEA (preoperative vs. postoperative: p = 0.000003, preoperative vs. follow-up: p = 0.0000007, Table 1). Neither the preoperative SGOI nor the SGOIbiopsy correlated with the mean obstruction ratio (preoperative SGOI: r = 0.093, p = 0.7; preoperative SGOIbiopsy: r = 0.143, p = 0.5). The SGOI did not correlate with PaO2 (preoperative: r = -0.024, p = 0.9; postoperative: r = -0.298, p = 0.2; follow-up: r = -0.298, p = 0.2) or A-aDO2 (preoperative: r = -0.134, p = 0.5; postoperative: r = 0.436, p = 0.06; follow-up: r = 0.320, p = 0.2). The SGOI weakly correlated with the postoperative %DLCO/VA (r = -0.445, p = 0.04), but not with the preoperative and follow-up values (preoperative: r = 0.101, p = 0.7; follow-up: r = -0.090, p = 0.7). Pulmonary arteriopathy and hypoxemia after PEA Patient age, preoperative PaO2, cardiac and respiratory function, reduction in PVR (%Δ PVR), SGOI, decrease in SGOI (Δ SGOI), and mean obstruction ratio were associated with PaO2. Multiple regression analysis using these factors revealed that the obstruction ratio correlated with both postoperative and follow-up PaO2 (Tables 2 and 3). Single regression analysis revealed that the mean obstruction ratio weakly correlated with preoperative PaO2 (standardized regression coefficient: -0.426, adjusted R-squared: 0.1421, p = 0.04,). The correlations between the other factors and preoperative PaO2 were poor (S2 Table). 10.1371/journal.pone.0161827.t002Table 2 Univariate and multivariate analysis of variables associated with postoperative PaO2. Simple regression analysis Multiple regression analysis β p-value β p-value β p-value Age -0.699 ± 0.164 0.0005 -0.345 ± 0.184 0.08 -0.391 ± 0.171 0.04* Preoperative PaO2 0.620 ± 0.213 0.01 0.060 ± 0.215 0.8 Postoperative SGOI -0.282 ± 0.236 0.2 Δ SGOI 0.268 ± 0.223 0.2 Postoperative CI 0.016 ± 0.238 0.945 %Δ PVR 0.466 ± 0.215 0.04 0.128 ± 0.175 0.5 Obstruction ratio -0.797 ± 0.162 0.0001 -0.492 ± 0.203 0.03 -0.543 ± 0.182 0.009* Posoperative %VC 0.065 ± 0.253 0.8 Postperative FEV1.0/FVC% 0.337 ± 0.220 0.1 Postoperative %DLCO/VA 0.411 ± 0.217 0.08 β = standardized regression coefficient ± standard error. *Adjusted R2 = 0.651, p = 0.00009, See Table 1 for abbreviations. 10.1371/journal.pone.0161827.t003Table 3 Univariate and multivariate analysis of variables associated with follow-up PaO2. Simple regression analysis Multiple regression analysis β p-value β p-value β p-value Age -0.510 ± 0.192 0.02 -0.002 ± 0.200 1.0 Preoperative PaO2 0.517 ± 0.225 0.03 0.194 ± 0. 193 0.3 Follow-up SGOI -0.300 ± 0.222 0.2 Δ SGOI 0.055 ± 0.23 0.8 CI 0.152 ± 0.221 0.5 %Δ PVR 0.613 ± 0.177 0.002 0.338 ± 0. 179 0.08 0.360 ± 0.168 0.045* Obstruction ratio -0.693 ± 0.159 0.0003 -0.462 ± 0. 197 0.03 -0.521 ± 0.167 0.006* Follow-up %VC 0.318 ± 0.212 0.1 Follow-up FEV1.0/FVC% 0.088 ± 0.22 0.7 Follow-up %DLCO/VA 0.415 ± 0.203 0.055 β = standardized regression coefficient ± standard error. *Adjusted R2 = 0.545, p = 0.0002, See Table 1 for abbreviations. Association between pulmonary arteriopathy, A-aDO2, and diffusion capacity after PEA The mean obstruction ratio positively correlated with A-aDO2 at all time points (Fig 4A, 4B and 4C). The high-obstruction group had a higher A-aDO2 than the low-obstruction group (p = 0.004, analyzed by repeated measures ANOVA, Fig 4D). The mean obstruction ratio negatively correlated with the postoperative and follow-up %DLCO/VA (Fig 5A and 5B), but did not correlate with the preoperative %DLCO/VA (Fig 5C). The %DLCO/VA in the high-obstruction group was significantly lower than the low-obstruction group (p = 0.002, analyzed by repeated measures ANOVA, Fig 5D). 10.1371/journal.pone.0161827.g004Fig 4 Relationship between mean obstruction ratio and A-aDO2. The mean obstruction ratio was negatively correlated with the preoperative (A), postoperative (B), and follow-up A-aDO2 (C). The high obstruction group had higher A-aDO2 values than the low-obstruction group (p = 0.004, analyzed by repeated measures ANOVA) (D). 10.1371/journal.pone.0161827.g005Fig 5 Relationship between the mean obstruction ratio and %DLCO/VA. The mean obstruction ratio was negatively correlated with postoperative (A) and follow-up %DLCO/VA (B). However, the mean obstruction ratio did not correlate with preoperative %DLCO/VA (C). The high obstruction group had lower %DLCO/VA values than the low-obstruction group (p = 0.002, analyzed by repeated measures ANOVA) (D). (*: p<0.05, vs preoperative data for each group; §: p<0.05, vs postoperative data for each group). Potential effect of PvO2 on PaO2 in CTEPH PvO2 was positively correlated with PaO2 at all time points (preoperative: r = 0.425, p = 0.04; postoperative: r = 0.773, p = 0.0001; follow-up: r = 0.836, p = 0.000001). Discussion In this study, we observed that severe pulmonary arteriopathy correlates with persistent hypoxemia after PEA. The extent of pulmonary arteriopathy was negatively correlated with postoperative and follow-up PaO2 and was an independent determinant of PaO2 after PEA. We found that severe pulmonary arteriopathy is associated with persistent hypoxemia following PEA in patients with CTEPH. Hypoxemia can be caused by four mechanisms: a ventilation-perfusion mismatch, impaired diffusion capacity, intrapulmonary shunts and hypoventilation [17]. Kapitan and colleagues reported that the ventilation-perfusion mismatch, measured by gas elimination methods, could be a major contributor to hypoxemia in postoperative CTEPH patients [11]. In our study, the mean obstruction ratio positively correlated with A-aDO2, and patients with high obstruction ratios had little improvement in A-aDO2 after PEA (Fig 4). Hence, severe pulmonary arteriopathy could be related to higher A-aDO2 after PEA. A-aDO2 reflects the degree of ventilation-perfusion mismatch [13]; therefore, severe pulmonary arteriopathy might be associated with unresolved ventilation-perfusion mismatch in the pulmonary microcirculation after PEA. An impaired diffusion capacity was also associated with hypoxemia [13,17]. Low DLCO results from a reduction in the pulmonary capillary blood volume (Vc) and/or pulmonary membrane diffusion capacity (DM) [13]. Bernstein et al. reported that Vc and DM in preoperative CTEPH patients were significantly lower than in healthy control subjects and the values did not improve after PEA [18]. It was suggested that the reduction in Vc and DM might be due to the pulmonary arteriopathy [18,19], and our results support this idea (Fig 5). Thus, the severe pulmonary arteriopathy in some postoperative CTEPH patients might be associated with ventilation-perfusion mismatch and impaired diffusion capacity. The adverse effects of PEA, or other as-yet unidentified postoperative factors, might be involved in the delayed recovery of PaO2 one month after PEA [7,20]. In general, the VC value, which could induce hypoventilation, was depressed at one to three months following thoracotomy [21] and recovery of the respiratory function took three or more months following PEA [7]. The vascular steal phenomenon, defined as a disproportionate increase of blood flow to newly perfused lung regions [22], is a major contributor to ventilation perfusion mismatch and hypoxemia during the early postoperative period [20]. An anatomical shunt between the pulmonary and bronchial circulations develops in CTEPH patients [23,24]. Increased blood flow supplied from the two circulations [25] has been shown to induce endothelial damage, resulting in lung edema and hypoxemia in the postoperative period [26]. Intrapulmonary shunt may cause hypoxemia in CTEPH patients; however, current data are contradictory. It has been reported that anatomical shunts between the pulmonary arteries and veins (PA-PV shunt) are present in CTEPH patients [24]. Tanabe et al. reported that preoperative CTEPH patients had an increased absolute shunt ratio, and this did not decrease after PEA [20]. In the present study, the preoperative shunt ratio by perfusion scans was slightly increased over the normal range (8%), and the values after PEA were evaluated in a few patients (data not shown). However, Kapitan et al. found that intrapulmonary shunts do not exist in this patient population and are not responsible for hypoxemia in CTEPH patients [10,11]. The reason for these discrepancies may be that different methods were used in the studies [10,11,20]. There is a discrepancy in the prevalence of hypoxemia and residual PH following PEA in patients with CTEPH [7,8]. The precise mechanisms underlying this inconsistency remain unknown. Some patients demonstrate an increased shunt ratio on preoperative perfusion scans. The preoperative shunt ratio weakly correlated with PaO2 at follow-up (r = -0.502, p = 0.02), but did not correlate with preoperative PaO2 (r = -0.299, p = 0.08) (data not shown). The shunt ratio data following PEA were too sparse for robust analysis, but our group previously reported that the shunt ratio did not change after PEA [20]. Severe vascular remodeling might contribute to persistent shunting, resulting in inconsistent shunt observations in this patient population. Fig 6 illustrates our hypothesis that severe pulmonary arteriopathy contributes to persistent hypoxemia after PEA. Removal of chronic segmental pulmonary thrombi by PEA would be expected to increase blood flow to distal vascular areas, but persistent pulmonary arteriopathy would cause a chronic disruption of blood flow into the pulmonary capillary bed. This disruption could result in a sustained ventilation-perfusion mismatch, reduced Vc, and impaired diffusion capacity. We hypothesize that obstruction of blood flow causes a persistent intrapulmonary shunt. Hypoxemia can be sustained even one year after PEA. The persistent nature of hypoxemia may be due to an irreversible component of the pulmonary arteriopathy. It has been reported that severe pulmonary vascular remodeling persists after epoprostenol therapy [27] and successful balloon pulmonary angioplasty [28]. Reversal of pulmonary artery remodeling might improve both hypoxemia and residual PH and is a goal for future CTEPH treatment.　 10.1371/journal.pone.0161827.g006Fig 6 The effects of severe pulmonary arteriopathy on oxygenation and hemodynamics following PEA in patients with CTEPH. Severe pulmonary arteriopathy might negatively influence the pulmonary microcirculation, resulting in residual hypoxemia after PEA. Vc: pulmonary capillary blood volume. The mechanisms that contribute to poor oxygenation in preoperative CTEPH patients appear to be more complicated. Ventilation-perfusion mismatch could be responsible for preoperative hypoxemia, which was amplified by the low PvO2 and the low cardiac output [10]. Kim et al. described two components, segmental arteries and distal vessels, that affect the pulmonary circulation in preoperative CTEPH patients [29]. It has also been reported that the extent of chronic segmental thrombi, estimated by pulmonary angiography, has minimal association with the ventilation-perfusion mismatch [10], which is in agreement with our data. In our study, the mean obstruction ratio weakly correlated with the preoperative A-aDO2 (Fig 4A), although the relationship between the mean obstruction ratio and the preoperative PaO2 was inconsistent (Fig 2C and S2 Table). Hence, multiple interrelated factors are likely to affect gas exchange in CTEPH patients prior to PEA, and the pulmonary arteriopathy might be a factor. Further investigation is needed to clarify these factors. Pulmonary vasodilators might affect the gas exchange in CTEPH patients. Several pulmonary vasodilators were administered to some patients (S1 and S3 Tables). Riociguat was used to treat only one patient because it was not yet approved in Japan when this study began. Patients who required the pulmonary vasodilators had higher pulmonary pressure and PVR, which might be related to severe pulmonary arteriopathy [12]. All patients tended to have improved PaO2 between the postoperative and follow-up periods. It was difficult to examine the beneficial or harmful effects of vasodilators on gas exchange as well as vascular remodeling for postoperative patients. Some vasodilators, such as endothelin receptor antagonists and epoprostenol, might potentially worsen the gas exchange due to deteriorating ventilation-perfusion inequality [30,31], although it was reported that riociguat and bosentan did not induce hypoxemia in the clinical trials for CTEPH patients [32,33]. The influence of vasodilators on gas exchange was unclear in this study. Further investigation is needed regarding this point. In this study, the biopsied sample size was small and only a single lung specimen was resected and examined from each patient due to safety and ethical reason. Dorfmüller et al. found evidence of pulmonary arteriopathy in 10 lung specimens randomly selected from whole lung tissue samples from each of 17 CTEPH cases [24]. Pulmonary arteriopathy has also been observed in both open and occluded vascular areas [34,35]. These results may suggest that the pulmonary arteriopathy is distributed throughout the entire lung. Finally, Yamaki’s report and our previous study indicate that the extent of pulmonary arteriopathy in a single biopsy specimen correlates with both the hemodynamics and the prognosis in operative CTEPH patients [12,36]. It was proposed that the entire pulmonary circulation might not be fully reflected in the small lung tissue samples. However, the single lung biopsy was interpreted as a random sampling of pulmonary arteries from each patient, and the specimens could provide important information regarding the pulmonary microcirculation. As an alternative to lung biopsy, the noninvasive methods of directly evaluating the pulmonary microcirculation were needed. This study had several limitations. First, the study was performed at a single institution in Japan. Second, the small lung tissue samples might not fully reflect the entire pulmonary circulation as described above. Third, the vascular steal phenomenon was not investigated because perfusion scans were, in most cases, performed only at the time of diagnosis. Fourth, the perfusion/ventilation mismatch was not evaluated by the gas elimination method. Fifth, the chronic segmental thrombi were only evaluated by CT scan images. Finally, the relationship between vasodilators and oxygenation was not investigated. Conclusion The severity of pulmonary arteriopathy was closely associated with PaO2 after PEA and was an independent determinant of postoperative and follow-up hypoxemia. Supporting Information S1 Fig Illustration detailing the quantification of pulmonary arteriopathy. The obstruction ratio of each pulmonary artery was defined as the ratio of the luminal area to the vascular area of the artery. The vascular and luminal areas were defined as the areas enclosed by the external elastic lamina and the luminal wall, respectively. The area was traced and measured using Image J software (ver. 1.45). The obstruction ratio of each pulmonary artery was calculated according to the following formula: Obstructionratio=Vasculararea−LuminalareaVasculararea (TIF) Click here for additional data file. S1 Table Details of vasodilator medical therapy. (DOCX) Click here for additional data file. S2 Table Univariate analysis of variables associated with preoperative PaO2. (DOCX) Click here for additional data file. S3 Table Details of vasodilator medical therapy in the high- and low-obstruction groups. (DOCX) Click here for additional data file. We thank the laboratory staff in the Department of Pathology, National Cerebral and Cardiovascular Center for technical support. We also thank Ms. Misa Ishimura in the Photo Center, Chiba University Hospital for drawing the beautiful illustration (Fig 6). Ms. Ishimura assigned the illustration copyright to the corresponding author. ==== Refs References 1 Kim NH , Delcroix M , Jenkins DP , Channick R , Dartevelle P , Jansa P , et al Chronic thromboembolic pulmonary hypertension . J Am Coll Cardiol 2013 ;62 :D92 –99 . 10.1016/j.jacc.2013.10.024 24355646 2 Tanabe N , Sugiura T , Tatsumi K . Recent progress in the diagnosis and management of chronic thromboembolic pulmonary hypertension . Respir Investig 2013 ;51 :134 –146 . 10.1016/j.resinv.2013.02.005 23978639 3 Sacks RS , Remillard CV , Agange N , Auger WR , Thistlethwaite PA , Yuan JX . Molecular biology of chronic thromboembolic pulmonary hypertension . Semin Thorac Cardiovasc Surg 2006 ;18 :265 –276 . 17185190 4 Jamieson SW , Kapelanski DP , Sakakibara N , Manecke GR , Thistlethwaite PA , Kerr KM , et al Pulmonary endarterectomy: experience and lessons learned in 1,500 cases . Ann Thorac Surg 2003 ;76 :1457 –1462 ; discussion 1462–1464. 14602267 5 Ishida K , Masuda M , Tanabe N , Matsumiya G , Tatsumi K , Nakajima N . Long-term outcome after pulmonary endarterectomy for chronic thromboembolic pulmonary hypertension . J Thorac Cardiovasc Surg 2012 ;144 :321 –326 . 10.1016/j.jtcvs.2011.09.004 21992851 6 van der Plas MN , Surie S , Reesink HJ , van Steenwijk RP , Kloek JJ , Bresser P . Longitudinal follow-up of six-minute walk distance after pulmonary endarterectomy . Ann Thorac Surg 2011 ;91 :1094 –1099 . 10.1016/j.athoracsur.2010.11.061 21440129 7 Zoia MC , D'Armini AM , Beccaria M , Corsico A , Fulgoni P , Klersy C , et al Mid term effects of pulmonary thromboendarterectomy on clinical and cardiopulmonary function status . Thorax 2002 ;57 :608 –612 . 12096204 8 Jenkins DP , Madani M , Mayer E , Kerr K , Kim N , Klepetko W , et al Surgical treatment of chronic thromboembolic pulmonary hypertension . Eur Respir J 2013 ;41 :735 –742 . 10.1183/09031936.00058112 23143539 9 Corsico AG , D'Armini AM , Cerveri I , Klersy C , Ansaldo E , Niniano R , et al Long-term outcome after pulmonary endarterectomy . Am J Respir Crit Care Med 2008 ;178 :419 –424 . 10.1164/rccm.200801-101OC 18556630 10 Kapitan KS , Buchbinder M , Wagner PD , Moser KM . Mechanisms of hypoxemia in chronic thromboembolic pulmonary hypertension . Am Rev Respir Dis 1989 ;139 :1149 –1154 . 2712441 11 Kapitan KS , Clausen JL , Moser KM . Gas exchange in chronic thromboembolism after pulmonary thromboendarterectomy . Chest 1990 ;98 :14 –19 . 2361380 12 Jujo T , Sakao S , Ishibashi-Ueda H , Ishida K , Naito A , Sugiura T , et al Evaluation of the Microcirculation in Chronic Thromboembolic Pulmonary Hypertension Patients: The Impact of Pulmonary Arterial Remodeling on Postoperative and Follow-Up Pulmonary Arterial Pressure and Vascular Resistance . PLoS One 2015 ;10 :e0133167 10.1371/journal.pone.0133167 26252755 13 West JB . Respiratory physiology: the essentials . 5th ed. Williams & Wilkins : Baltimore ; 1995 . 14 Qanadli SD , El Hajjam M , Vieillard-Baron A , Joseph T , Mesurolle B , Oliva VL , et al New CT index to quantify arterial obstruction in pulmonary embolism: comparison with angiographic index and echocardiography . AJR Am J Roentgenol 2001 ;176 :1415 –1420 . 11373204 15 Boyden EA . Semental anatomy of the lungs New York : McGraw-Hill ; 1955 . 16 Kanda Y . Investigation of the freely available easy-to-use software 'EZR' for medical statistics . Bone Marrow Transplant 2013 ;48 :452 –458 . 10.1038/bmt.2012.244 23208313 17 Wagner P . Gas exchange In: Spiro S , Silvesti G , Agsti A , editors. Clinical Respiratory Medicine . 4th ed. Philadelphia : Elsevier ; 2012 pp. 37 –49 . 18 Bernstein RJ , Ford RL , Clausen JL , Moser KM . Membrane diffusion and capillary blood volume in chronic thromboembolic pulmonary hypertension . Chest 1996 ;110 :1430 –1436 . 8989056 19 Steenhuis LH , Groen HJ , Koeter GH , van der Mark TW . Diffusion capacity and haemodynamics in primary and chronic thromboembolic pulmonary hypertension . Eur Respir J 2000 ;16 :276 –281 . 10968503 20 Tanabe N , Okada O , Nakagawa Y , Masuda M , Kato K , Nakajima N , et al The efficacy of pulmonary thromboendarterectomy on long-term gas exchange . Eur Respir J 1997 ;10 :2066 –2072 . 9311504 21 Ali J , Weisel RD , Layug AB , Kripke BJ , Hechtman HB . Consequences of postoperative alterations in respiratory mechanics . Am J Surg 1974 ;128 :376 –382 . 4606381 22 Moser KM , Metersky ML , Auger WR , Fedullo PF . Resolution of vascular steal after pulmonary thromboendarterectomy . Chest 1993 ;104 :1441 –1444 . 8222803 23 Shimizu H , Tanabe N , Terada J , Masuda M , Sakao S , Kasahara Y , et al Dilatation of bronchial arteries correlates with extent of central disease in patients with chronic thromboembolic pulmonary hypertension . Circ J 2008 ;72 :1136 –1141 . 18577824 24 Dorfmuller P , Gunther S , Ghigna MR , Thomas de Montpreville V , Boulate D , Paul JF , et al Microvascular disease in chronic thromboembolic pulmonary hypertension: a role for pulmonary veins and systemic vasculature . Eur Respir J 2014 ;44 :1275 –1288 . 10.1183/09031936.00169113 25142477 25 Ley S , Kreitner KF , Morgenstern I , Thelen M , Kauczor HU . Bronchopulmonary shunts in patients with chronic thromboembolic pulmonary hypertension: evaluation with helical CT and MR imaging . AJR Am J Roentgenol 2002 ;179 :1209 –1215 . 12388501 26 Gan HL , Zhang JQ , Sun JC , Feng L , Huang XY , Lu JK , et al Preoperative transcatheter occlusion of bronchopulmonary collateral artery reduces reperfusion pulmonary edema and improves early hemodynamic function after pulmonary thromboendarterectomy . J Thorac Cardiovasc Surg 2014 ;148 :3014 –3019 . 10.1016/j.jtcvs.2014.05.024 24929804 27 Rich S , Pogoriler J , Husain AN , Toth PT , Gomberg-Maitland M , Archer SL . Long-term effects of epoprostenol on the pulmonary vasculature in idiopathic pulmonary arterial hypertension . Chest 2010 ;138 :1234 –1239 . 10.1378/chest.09-2815 21051399 28 Ogawa A , Kitani M , Mizoguchi H , Munemasa M , Matsuo K , Yamadori I , et al Pulmonary microvascular remodeling after balloon pulmonary angioplasty in a patient with chronic thromboembolic pulmonary hypertension . Intern Med 2014 ;53 :729 –733 . 24694486 29 Kim NH , Fesler P , Channick RN , Knowlton KU , Ben-Yehuda O , Lee SH , et al Preoperative partitioning of pulmonary vascular resistance correlates with early outcome after thromboendarterectomy for chronic thromboembolic pulmonary hypertension . Circulation 2004 ;109 :18 –22 . 14699009 30 Ghofrani HA , Wiedemann R , Rose F , Schermuly RT , Olschewski H , Weissmann N , et al Sildenafil for treatment of lung fibrosis and pulmonary hypertension: a randomised controlled trial . Lancet 2002 ;360 :895 –900 . 12354470 31 Stolz D , Rasch H , Linka A , Di Valentino M , Meyer A , Brutsche M , et al A randomised, controlled trial of bosentan in severe COPD . Eur Respir J 2008 ;32 :619 –628 . 10.1183/09031936.00011308 18448495 32 Jais X , D'Armini AM , Jansa P , Torbicki A , Delcroix M , Ghofrani HA , et al Bosentan for treatment of inoperable chronic thromboembolic pulmonary hypertension: BENEFiT (Bosentan Effects in iNopErable Forms of chronIc Thromboembolic pulmonary hypertension), a randomized, placebo-controlled trial . J Am Coll Cardiol 2008 ;52 :2127 –2134 . 10.1016/j.jacc.2008.08.059 19095129 33 Ghofrani HA , D'Armini AM , Grimminger F , Hoeper MM , Jansa P , Kim NH , et al Riociguat for the treatment of chronic thromboembolic pulmonary hypertension . N Engl J Med 2013 ;369 :319 –329 . 10.1056/NEJMoa1209657 23883377 34 Moser KM , Bloor CM . Pulmonary vascular lesions occurring in patients with chronic major vessel thromboembolic pulmonary hypertension . Chest 1993 ;103 :685 –692 . 8449052 35 Yi E , Kim H , Ahn H , Strother J , Morris T , Masliah E , et al Distribution of obstructive intimal lesions and their cellular phenotypes in chronic pulmonary hypertension . Am J Respir Crit Care Med 2000 ;162 :1577 –1586 . 11029379 36 Yamaki S , Ando M , Fukumoto Y , Higuchi Y , Kaneko K , Maeda K , et al Histopathological examination by lung biopsy for the evaluation of operability and postoperative prognosis in patients with chronic thromboembolic pulmonary hypertension . Circ J 2014 ;78 :476 –482 . 24284920

PMC005xxxxxx/PMC5003342.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 27571416PONE-D-15-3973610.1371/journal.pone.0157988Research ArticleComputer and Information SciencesData VisualizationInfographicsGraphsPhysical SciencesMathematicsApplied MathematicsAlgorithmsResearch and Analysis MethodsSimulation and ModelingAlgorithmsPhysical SciencesMathematicsOptimizationComputer and Information SciencesInformation TechnologyData MiningComputer and Information SciencesSoftware EngineeringSoftware ToolsEngineering and TechnologySoftware EngineeringSoftware ToolsPhysical SciencesMathematicsProbability TheoryProbability DistributionBiology and Life SciencesOrganismsAnimalsVertebratesAmniotesMammalsMarine MammalsDolphinsBiology and Life SciencesMarine BiologyMarine MammalsDolphinsEarth SciencesMarine and Aquatic SciencesMarine BiologyMarine MammalsDolphinsBiology and Life SciencesBehaviorRecreationSportsBiology and Life SciencesSports ScienceSportsA Novel Clustering Methodology Based on Modularity Optimisation for Detecting Authorship Affinities in Shakespearean Era Plays A Novel Clustering Methodology Based on Modularity Optimisationhttp://orcid.org/0000-0002-3360-7680Naeni Leila M. 123*Craig Hugh 4Berretta Regina 12Moscato Pablo 12 1 The Priority Research Centre of Bioinformatics and Information-Based Medicine, The University of Newcastle, Newcastle, New South Wales, Australia 2 School of Electrical Engineering and Computer Science, Faculty of Engineering and Built Environment, The University of Newcastle, Newcastle, New South Wales, Australia 3 School of Built Environment, Faculty of Design, Architecture and Building, University of Technology Sydney, Sydney, Australia 4 Centre for Literary and Linguistic Computing, School of Humanities and Social Science, The University of Newcastle, Newcastle, New South Wales, Australia Schlick Tamar Editor New York University, UNITED STATES Competing Interests: The authors have declared that no competing interests exist. Conceived and designed the experiments: LN PM. Performed the experiments: LN. Analyzed the data: LN PM HC. Contributed reagents/materials/analysis tools: LN PM HC RB. Wrote the paper: LN PM HC. * E-mail: Leila.MoslemiNaeni@uon.edu.au2016 29 8 2016 11 8 e01579888 9 2015 8 6 2016 © 2016 Naeni et al2016Naeni et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.In this study we propose a novel, unsupervised clustering methodology for analyzing large datasets. This new, efficient methodology converts the general clustering problem into the community detection problem in graph by using the Jensen-Shannon distance, a dissimilarity measure originating in Information Theory. Moreover, we use graph theoretic concepts for the generation and analysis of proximity graphs. Our methodology is based on a newly proposed memetic algorithm (iMA-Net) for discovering clusters of data elements by maximizing the modularity function in proximity graphs of literary works. To test the effectiveness of this general methodology, we apply it to a text corpus dataset, which contains frequencies of approximately 55,114 unique words across all 168 written in the Shakespearean era (16th and 17th centuries), to analyze and detect clusters of similar plays. Experimental results and comparison with state-of-the-art clustering methods demonstrate the remarkable performance of our new method for identifying high quality clusters which reflect the commonalities in the literary style of the plays. http://dx.doi.org/10.13039/501100000923Australian Research CouncilDP140104183Moscato Pablo http://dx.doi.org/10.13039/501100000923Australian Research CouncilDP120101955Craig Hugh The authors acknowledge support from the Australian Research Council (ARC, http://www.arc.gov.au/) to Pablo Moscato (Discovery Project, DP140104183) and Hugh Craig (Discovery Project, DP120101955). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Following great advancements in technology, all scientific fields have been faced with huge empirical datasets. Identifying groups of objects, patterns or elements with “similar characteristics” has always drawn the attractions of researchers. Data clustering is one of the most widely studied problems in data mining and machine learning with a wide variety of applications in many fields ranging from the analysis of genomic data in biology [1, 2] to classifying customers for efficient marketing strategies [3, 4]. Clusters are natural groups in data such that elements within the same cluster are more similar to each other than to elements belonging to other clusters [5]. Thousands of techniques have been proposed to address a wide variety of clustering problems in different disciplines (see the review of data mining and clustering techniques by Han and Kamber [6]). The main clustering methods include probabilistic and generative models [7, 8], distance-based clustering [9], density and grid-based clustering [10, 11], spectral clustering [12, 13] and graph clustering [14]. Graph clustering is the process of grouping vertices of a graph into clusters according to the structure of the graph. Generally speaking, a good graph clustering contains clusters with many edges within the clusters and few between clusters. However this definition is unclear, thus many quality functions have been proposed to evaluate graph clusters such as ratio cut [15], normalized cut [16] and modularity [17]. In the recent literature a cluster of vertices in a graph is mostly called a community [18] and graph clustering discussed as the community detection problem [19]. The graph clustering problem is a fundamental algorithmic problem, however it has not been solved satisfactory and it is a notably challenging member of the NP-complete class [20, 21]. This study proposes a new data clustering methodology for the authorship analysis in a dataset generated from 168 plays from the Shakespearean era. Authorship study is a long-standing research activity, predating the availability of computers by many centuries. Interest in authorship follows from the magnitude of the consequences of assigning a work to one author or another, e.g. in separating canonical and apocryphal books in a religious tradition, establishing the authenticity of legal documents, or determining which works or parts of works are by celebrated literary authors. With the availability of digital text and computational tools, this activity is now predominantly quantitative. The Shakespearean canon has been a particular focus, with studies testing the claims of various possible authors other than William Shakespeare of Stratford to the works we know as Shakespeare’s [22], and debating possible additions to Shakespeare’s canon [23, 24]. The advent of quantitative methods has also renewed interest in the idea that authorship is objectively a property of texts, rather than largely the invention of readers and institutions, as was argued by poststructuralist theorists [25–28]. Here, we propose a new methodology for identifying groups in the large text corpus dataset. The methodology is easily generalizable and can be employed for analyzing any other dataset. This advanced clustering methodology brings together Jensen-Shannon divergence(JSD) as an information theoretic dissimilarity measure and combines it with a newly proposed memetic algorithm (iMA-Net) for graph clustering to yield a powerful strategy for segmenting the text corpus dataset. The proposed memetic algorithm maximizes the modularity quality function and it is an unsupervised algorithm, i.e. the method does not require any information about the data elements, such as authorship and genre of plays, it only uses the structure of the dataset to uncover clusters of similar plays. The outcomes are promising and show great authorial affinities in each cluster; a computational proof of the highly individual literary style of each author. Materials and Methods Dataset In this work, we utilized a text corpus dataset containing 168 plays from the Shakespearean era (16th and 17th centuries) with the unambiguous contribution of authorship of 39 authors (see Table 1). This dataset was generated by Craig and Whipp [29] using a software application called Intelligent Archive (IA) to count variant spellings of approximately 55,114 unique words across all 168 plays. The software tool IA is used to extract the word-play matrix from a collection of suitably edited texts, so that only the words of dialog are counted and other materials, such as prefaces, stage directions and speaker tags are omitted. In the texts, contractions such as “they’re” are expanded so this counts as one instance of they and one of are. The IA has the capacity to combine variant spelling forms so that multiple spellings of the same word are combined into a single count [29]. As a result, this dataset contains about 9.3 million word usage statistics (word frequencies) stored in the form of a (55,114 × 168) matrix. The word-play matrix is available in the supporting information of [30]. 10.1371/journal.pone.0157988.t001Table 1 Authors and their contributions in the 168 Shakespearean era plays dataset. Author # Plays Author # Plays Author # Plays Shakespeare 28 Shirley 3 Greville 1 Middleton 18 Webster 3 Davenant 1 Jonson 17 Rowley 2 Brome 1 Fletcher 15 Massinger 2 Day 1 Chapman 13 Haughton 2 Porter 1 Lyly 8 Kyd 2 Chettle 1 Ford 7 Marston 2 Edwards 1 Peele 5 Wilmot 1 Suckling 1 Dekker 5 Carey 1 Sidney 1 Marlowe 5 Daniel 1 Marmion 1 Heywood 5 Brandon 1 Beaumont 1 Greene 4 Lodge 1 Tourneur 1 Wilson 3 Goffe 1 Nashe 1 This dataset was analyzed by Marsden et al. [30] and they found measurably distinct literary styles by detecting the tendency towards overuse or avoidance of particular words for four key authors who account for the largest number of plays in the dataset. That study resulted in a set of marker words that can distinguish between plays written by Fletcher, Jonson, Middleton and Shakespeare compared to the remainder of the authors. Recently, Arefin et al. [31] used this dataset to test the performance of their new graph clustering method incorporating paraclique identification and a clustering method known as MST-kNN [32]. Method As mentioned earlier, to detect clusters of plays that are more similar in the given dataset, we propose an unsupervised graph-based clustering methodology that employs a new memetic algorithm which optimizes the modularity value in k-Nearest Neighbour (kNN) graphs derived from the dataset. Fig 1 outlines our methodology as a four-stage process. This clustering methodology is general; it can be used to detect clusters in a dataset of n elements where each element has m quantified features (the input dataset can be represented in a form of a (m × n) matrix). In the current study, n equals to 168 and refers to the number of plays and m equals 55,114 and refers to the words used in the plays. Details of each stage are described in the following section. The primary idea of this partitioning approach was proposed in [33] as a new method to analyse a complete network derived from field survey aimed at detecting consumer communities of trust and confidence in the Australian not-for-profit and charity sector. In this paper, we extended the primary idea by using the square root of Jensen-Shannon divergence metric (JSD) as the dissimilarity metric, enhancing the modularity optimization algorithm and finding clusters with higher quality. 10.1371/journal.pone.0157988.g001Fig 1 Flowchart of stages in the clustering methodology proposed in this study. Stage 1- Compute the Dissimilarity Matrix We start our method by computing the dissimilarity between all pairs of plays. We use the square root of Jensen-Shannon divergence (JSD). JSD is a metric measuring the dissimilarity between two probability distributions based on Jensen’s inequality [34] and the Shannon entropy [35]. We refer to Lin [36] for a discussion of the JSD. The JSD has been used in many studies, especially in bioinformatics [37–39]. The JSD is selected being the dissimilarity metric because it has the distance metric properties and satisfies the triangle inequality condition (this property is proved in [40, 41]). The JSD reflects the dissimilarity of two plays based on the frequency of different words appeared in each play. Generally, for two probability distributions, P and Q, JSD is computed as follows, JSD(P,Q)=HP+Q2-H(P)+H(Q)2(1) where, H(X) is Shannon information entropy for distribution X and is computed by Eq 2. H(X)=-∑xi∈Xxilog2xi.(2) Here, xi is the probability of occurrence of word i in the play X. The JSD is a non-negative and symmetric metric, therefore, JSD(P, Q) = JSD(Q, P). For two identical probability distributions (i.e. P = Q), JSD(P, Q) = 0, i.e. there is no dissimilarity between the two. Moreover, by using the log2 in Eq 2 the JSD is bounded by 1, thus 0≤JSD≤1. By computing the JSD for all pairs of plays, we obtain a symmetric 168 × 168 matrix representing the dissimilarity between all plays (see S1 Table). In this study, the JSD matrix is computed using the cibm.utils package developed in R. A graph is associated with the dissimilarity matrix, where each node represents a play and edges connect nodes where the weight of an edge is the dissimilarity between the plays. In the next stage, we will introduce a new method to analyze this graph. Stage 2- Generate k-Nearest Neighbour (kNN) Graphs In this stage we generate incomplete graphs by removing edges that are less important from the complete graph and preserving edges that represent stronger connections between nodes. For this purpose, we set up a series of kNN graphs starting with k = 1 to k = 10 (in total 10 graphs). To generate the kNN graphs from the dissimilarity matrix, a pair of nodes (a, b) is connected if either a is one of the k-nearest neighbors of b or if b is among the k nearest neighbors of a. The kNN graphs of this study are generated by the scalable software tool proposed by Arefin et al. [42]. Investigating 10 different kNN graphs gives us the opportunity to study the performance of our method under different cases. By increasing the value of k, the average degree of nodes is increased and the graph becomes denser. Therefore, identifying meaningful clusters in kNN graphs with a larger k is more computationally challenging in practice, as graph clustering methods identify clusters in the graph based on the graph’s edge structure, and more edges simply means more possible solutions that the method has to analyze. Table 2 shows the number of edges and the average degree for each of the kNN graphs. 10.1371/journal.pone.0157988.t002Table 2 Basic information of kNN graphs constructed from the complete dataset. All graphs have 168 nodes that are representing the 168 plays in the dataset. kNN Graphs # Edges Average Degree k = 1 145 1.7 k = 2 284 3.4 k = 3 441 5.3 k = 4 551 6.6 k = 5 687 8.2 k = 6 822 9.8 k = 7 957 11.4 k = 8 1,091 13.0 k = 9 1,225 14.6 k = 10 1,362 16.2 We then apply a clustering approach to the kNN graphs. To detect the clusters, we developed a memetic algorithm that optimizes modularity value as the evaluation criterion. Due to the fact that the memetic algorithm uses the edge weights to cluster nodes with strong connections, we have to modify the edge weights in kNN graphs in order to represent the ‘similarity’ between nodes so that the clusters will contain similar nodes. For this purpose, similarity is defined as one minus dissimilarity. Stage 3- Modularity optimization algorithm Modularity optimization is one of the most popular methods for detecting graph structure and hidden groups of nodes with many intra-connections and comparatively few inter-connections. As explained in the introduction, such a group of nodes in the graph is called a community or cluster. It is believed that nodes which belong to the same community (cluster) are more likely to play a similar role or share common functionality [18, 19]. Detecting the community structure of a given graph, or community detection, is a newer nomenclature for the graph clustering problem and both problems are similarly looking for groups of “related” nodes to label as a cluster or community [14]. It is worth emphasizing that detecting the community structure is practicable only if graphs are sparse [19]. If the number of edges is close to the maximal number of edges, then the graph is a dense graph, the nodes become too homogeneous and the community structure does not make sense turning the problem into data clustering, which has its own concepts and methods that are more suitable for high density graphs [5, 43]. As a variety of complex systems in biology, sociology, physics and computer science can be represented as graphs, community detection is one of the most attractive problems in many disciplines [44, 45]. Although different approaches and methods have been proposed to solve the community detection problem, it is still a difficult problem and has not been solved satisfactorily [19]. Modularity, which was first introduced by Newman and Girvan [17], has the ability to evaluate the quality of communities, therefore it has become the key objective function in most of the recent algorithms. The modularity function is formulated by computing the difference between the number of the edges within a community (i.e. sub-graph) and the number of edges expected to appear in the same sub-graph in the ‘null-model’. The null-model is a model of the graph which is used for comparison, to verify whether the graph has community structure. The most popular null-model is the randomised version of the original graph proposed by Newman and Girvan, where the same number of nodes are randomly assigned, under the constraint that the expected degree of each node matches the degree of the node in the original graph [19, 46]. The random graph is not expected to have a community structure, therefore the graph has a community structure if it is significantly different from the random graph. In other words, higher modularity shows greater difference between the edge density in a sub-graph and expected value in the random graph, revealing community structure. Given an edge-weighted graph G = (V, E, W) the modularity Q of a clustering solution (i.e. community structure) I is defined as follows: Q(I)=∑c∈IlcM-dc2M2(3) The summation runs over all the communities c contained in the solution I, and M refers to the total weight of G, i.e. M=∑eij∈Ewij,(4) where wij denotes the weight of the edge connecting node i and node j. In Eq 3, lc stands for the total weights of edges connecting two nodes in community c and dc is the sum of the weighted degrees of the nodes in community c, i.e. lc=∑i,j∈cwij(5) dc=∑i∈c∑∀jwij(6) A solution with a higher value of modularity is believed to indicate a better community structure or clustering solution within the given graph. Therefore, the maximization of modularity is by far the most popular approach for identifying the community structure of a graph [19]. Since Brandes et al. [47] proved that modularity maximization leads to a NP-complete problem, finding the community structure with the maximum modularity has become a great challenge for data scientists. Therefore, many heuristic algorithms have been developed to find near optimal and good solutions. Some of the most popular methods employed for modularity maximization are heuristics based on hierarchical or aggregation clustering methods [17, 48–50], genetic algorithms [51, 52], memetic algorithms [53–55], spectral methods [56, 57] and simulated annealing [58]. In the present study, we propose a new memetic algorithm specially designed for solving the community detection problems named “iMA-Net” that is an improved version of the MA-Net algorithm previously proposed in [59]. There are three main differences between MA-Net and iMA-Net. 1) MA-Net considers all graphs as unweighted graphs while iMA-Net detects communities in weighted graphs. 2) iMA-Net employs an enhanced local search algorithm to make the algorithm more stable. 3) While the population initialization procedure in both MA-Net and iMA-Net applies local search to each individual to improve its fitness, differing local searches make the initialization procedure performance different in MA-Net and iMA-Net. The framework and steps of iMA-Net are explained in detail in the following sections. In the last section of stage 3, the experimental results of the proposed algorithm (iMA-Net) in five real-world benchmark networks are given. To illustrate the performance reliability of iMA-Net, we compare results of iMA-Net with MA-Net [33] and those from five representative community detection algorithms. iMA-Net: A Memetic algorithm for Modularity optimization Due to the proven success of memetic algorithms and their effectiveness in dealing with many NP-hard combinatorial optimization problems [60], we chose the memetic framework for our modularity optimization algorithm, named iMA-Net. iMA-Net is a population-based algorithm in which each solution in the population represents a particular clustering of the given weighted and undirected graph. In the proposed algorithm, solutions are evolved using problem-specific genetic operators and a local search procedure employed to detect high-quality community structure with the optimal or near optimal modularity value. We used the string-coding solution representation in which a clustering solution of G is represented by a list of n integer numbers [C1, C2, …, Cn], where n is equal to the number of nodes in G and Ci refers to the community label of node vi. Therefore, if Ci = Cj then i and j are located in the same community. This way of representing clustering solutions has been used in previous algorithms, for instance [51, 53, 54]. An illustration example of string-coding representation is given in Fig 2. 10.1371/journal.pone.0157988.g002Fig 2 An illustration of the string-coding representation for a clustering solution. Left: a simple toy graph with two clusters shown using different colors, Right: an integer list encoding the solution, where the red and blue color clusters are labeled as 1 and 2, respectively. Algorithm 1 presents the overall framework of iMA-Net for finding the partition of G that achieves the maximum modularity (computed following Eq 3). Details of each step of the algorithm are explained in the following paragraphs. Algorithm 1. iMA-Net Framework Input: Graph G = (V, E, W) Parameters: Np: Population size, Pm0: Minimum mutation probability, Ns: Acceptable number of generations without improvement. Output: I* a partition of G with fitness value Q* pop = {I1, I2, …, INp} ← Initialization(G,Np); repeat  Inew ← Modularity Based Recombination(pop);  Imut ← Adaptive Mutation(Inew, Pm0, Ns);  Iimp ← Local Search(Imut);  pop ← Update(pop, Iimp); until Termination Condition(Ns); return I * = argmaxI ∈ pop{Q(I)}; Initialization: The initial population consists of Np feasible solutions, so the following process runs Np times to build the initial population. Firstly, each of the n nodes in G is given a random community label that can be any integer number in the range [1,n]. Then to improve the quality of the solutions in the initial population, the local search procedure, as explained later in this section, is applied on each solution and made changes on community labels of nodes to find a neighbor solution with higher modularity. As the community label is selected randomly from a wide range of [1,n], there is not any constraint on the number of the different labels or the number of clusters in a solution. Applying the local search procedure on each solution, improves the quality of the initial population and speeds up the convergence of the algorithm. Modularity-based recombination operator: Traditional operators including uniform, one-point and two-point recombination operators are not suitable for the string based representation as they do not convey characteristics of parent solutions to the offspring. iMA-Net uses a problem-specific recombination operator for generating new solutions that inherit the best characteristics from the parent solutions which is named the modularity-based recombination operator. This recombination operator was first proposed for MA-Net [59]. This recombination method effectively transfers the best characteristics of the parents to the descendants and it works as follows. First, two solutions Ia and Ib are selected uniformly at random from the population as parents. The initialization procedure allows us to safely assume that all population members have relatively good quality and the random selection strategy helps maintain the population diversity. Next, a priority list PL is generated by sorting all communities in both parents based on their fitness values, as computed by Eq 3. Then, the best community with the highest priority from PL is selected and the same community is formed in the offspring. Then, for the next community in the priority list, form the most similar community in the offspring with the nodes that are not assigned to any community yet. This procedure runs until all the nodes in the offspring solution have been assigned to a community. The offspring generated by the modularity-based recombination operator inherits the best community structure from their parents [33]. One of the limitations of modularity optimization algorithms is their difficulty in detecting small size communities and their tendency to merge small communities and form large communities [61]. The modularity-based recombination operator that we proposed has the advantage of generating offspring with a larger number of communities than its parents. This advantage helps the algorithm to explore the search space more deeply and not avoid solutions with small communities. Adaptive mutation operator: This operator changes the community’s label of nodes with the probability of Pm. This means that in each run of the adaptive mutation operator ⌈Pm * n⌉ nodes are randomly moved to one of their neighbors’ communities. Thus, this mutation operator is neighbor-based and considers only effective changes. The mutation probability (Pm) is adaptive and it linearly grows from Pm0 to 2Pm0, as the algorithm approaches the termination condition. As shown in Algorithm 1, the stopping condition of iMA-Net is Ns generations without any improvement in the best solution found so far. By increasing the mutation probability (Pm), more changes will occur in the solution, thus diversity of the population will grow and the search area will expand. For instance, when the algorithm’s parameters are set to be Pm0=0.05 and Ns = 30 and the algorithm runs 15 generations without improvement in the best solution found, the mutation probability will grow to 1530Pm0+Pm0 and it will be Pm = 0.075. Local search: The local search procedure works on every new solution developed by the adaptive mutation operator, and it aims to improve the fitness of the solution before adding it to the population. iMA-Net uses the Vertex Movement (VM) heuristic [62] together with a stochastic hill climbing strategy to search the neighborhood to find a better solution. The local search procedure is shown in 2. Given a graph G with n nodes, the local search procedure works as follows. Firstly, L, a random sequence of the nodes is generated. Then, for each node vi chosen according to L, vi’s label is changed to that of a randomly selected neighbor, if the change improves the fitness of the solution. If the movement of vi from its community to none of its neighbors’ community increases the fitness (modularity), we leave vi in its place and check the next node in L. The above process runs until there is no change for any of the nodes in L that improves the fitness of the solution (see Algorithm 2). Algorithm 2. Local search procedure Input: Graph G with n nodes; A clustering solution I = [C1, …, Cn]. Output: I′=[C1′,...,Cn′] a solution with better fitness than I. I′ ← I; repeat  t = 0 (t shows the number of nodes that cannot move);  L ← a random sequence of n nodes;  for each node vi in L do   N ← a list of vi’s neighbours located in other communities;   ΔQ = 0;   while ΔQ ≤ 0 do    vj ← a neighbour node selected from N;    ΔQ ← fitness change due to the movement of vi to the community of vj;   end   if ΔQ > 0 then    Ci′=Cj (vi is moved to the community of vj);   else    t = t + 1 (when vi is not moved);   end  end until t = n; return I′; The local search procedure is designed to accept the first random movement of the node that improves the fitness value. This stochastic behavior in selecting the neighbor helps the algorithm to avoid one of the known limitations of deterministic hill climbing strategies which is getting stuck in the local optima. The local search procedure operation is effective and we reduced the computational cost by using the method proposed in [50] for computing the change to the modularity ΔQ incurred moving node vi into the community of node vj. Elitist population updating strategy: As with MA-Net [33], in iMA-Net the population updating strategy is elitist. New solutions generated by the local search procedure are added to the population and the least fit solutions are removed from the population. The elitist strategy enhances the algorithm convergence by retaining the better solutions in the population, making them eligible to be parents in the subsequent generations. Performance results of iMA-Net on benchmark networks Before applying the iMA-Net on the kNN graphs produced in Stage 2, we tested the performance of iMA-Net on five real-world networks with similar size to the kNN graphs. These networks include the Zachary’s Karate Club network (karate) [63], the Bottlenose Dolphins network (dolphin) [64], the American College Football network (football) [18], the Political Books network (polbooks) [56] and the Jazz Musicians network (jazz) [65]. Basic information on these benchmark networks is shown in Table 3. All of these networks are unweighted and undirected graphs. 10.1371/journal.pone.0157988.t003Table 3 Basic information of the real-world networks. Network # Nodes # Edges Average Degree 1 karate 34 78 4.59 2 dolphin 62 159 5.13 3 polbooks 105 441 8.40 4 football 115 613 10.66 5 jazz 198 2742 27.70 The comparison between iMA-Net and MA-Net in Table 4, is made to illustrate the effectiveness of iMA-Net. Moreover, to investigate the performance of iMA-Net, we compared the iMA-Net results with a set of well-known community detection algorithms with different approaches, frameworks and objective functions. The benchmark algorithms include LPAm [66], Meme-Net [67], MOGA-Net [68], MODPSO [69] and MLCD [70]. 10.1371/journal.pone.0157988.t004Table 4 Experimental results on five real-world benchmark networks. The maximum, average and standard deviation of modularity values (Qmax,Qavg,Qstd) obtained by LPAm, Meme-Net, Moga-Net, MODPSO, MLCD, MA-Net and iMA-Net. Criterion LPAm Meme-Net Moga-Net MODPSO MLCD MA-Net iMA-Net karate Qmax 0.4052 0.4020 0.4159 0.4198 0.4198 0.4198 0.4198 Qavg 0.3564 0.4020 0.3945 0.4182 0.4198 0.4195 0.4198 Qstd 0.0285 0 0.0089 0.0079 0 0.0022 0 dolphin Qmax 0.5071 0.5185 0.5034 0.5264 0.5285 0.5285 0.5285 Qavg 0.4938 0.5096 0.4584 0.5255 0.5285 0.5247 0.5252 Qstd 0.0114 0.0061 0.0163 0.0070 0 0.0032 0.0026 polbooks Qmax 0.5145 0.5232 0.4993 0.5264 0.5272 0.5272 0.5272 Qavg 0.4976 0.5218 0.4618 0.5263 0.5272 0.5255 0.5270 Qstd 0.0158 0.0031 0.0129 0.0007 0 0.0029 0.0004 football Qmax 0.6032 0.6044 0.4325 0.6046 0.6046 0.6046 0.6046 Qavg 0.5777 0.6023 0.3906 0.6038 0.6046 0.5984 0.6042 Qstd 0.0199 0.0015 0.0179 0.0011 0.0000 0.0051 0.0006 jazz Qmax 0.4448 0.4376 0.2929 0.4421 0.4451 0.4451 0.4451 Qavg 0.4360 0.4330 0.2952 0.4419 0.4451 0.4448 0.4450 Qstd 0.0098 0.0011 0.0084 0.0001 0.0000 0.0002 0.0001 LPAm presented by Barber and Clark [66] is a label propagation algorithm which reformulated the original label propagation algorithm (LPA) [71] to overcome its drawbacks. LPAm employs a modularity-specific rule for label propagation and updates the community label of nodes repeatedly till no possible improvement can be found. Meme-Net [67] is a memetic algorithm optimizing the modularity density to discover communities in the graph. MOGA-Net [68] and MODPSO [69] are two multi-objective optimization algorithms for community detection. MOGA-Net maximizes the intra-community links and minimizes the inter-community links, while MODPSO maximizes the internal link density and minimizes the external link density. MLCD is a novel memetic algorithm presented by Ma et. al [70]. MLCD has a multi-level learning framework for detecting the community structure by optimizing the modularity with the aid of a special hybrid global-local heuristic search procedure. In a comprehensive study [70], Ma et. al. ran all of the selected benchmark algorithms 50 times each, with the same key parameters, on several real-world and computer synthesized benchmark networks, and recorded the maximum, average and standard deviation of the modularity values (Qmax,Qavg,Qstd). In this study, we refer to the reported results in [70] and compare results obtained by 50 independent runs of MA-Net and iMA-Net with the benchmark algorithms. MA-Net and iMA-Net are implemented in Python 2.7 and executed on a PC with Intel ® Xeon ® CPU E5-1620 at a clock speed of 3.6 GHz (4 cores and 8 logical processors) and 16 GB of memory. We tuned the parameters of MA-Net and iMA-Net as follows: population size, Np, is set to 40, the minimum mutation probability, Pm0, is set to 0.05 and the termination condition is set to 30 generations without improvement (Ns = 30). The comparison results are shown in Table 4. Firstly, the comparison results in Table 4 show that in all benchmark networks MLCD, MA-Net and iMA-Net obtained the largest Qmax. While both MA-Net and iMA-Net achieved the maximum possible modularity in all networks, iMA-Net has more reliable performance because it improves the Qavg. Meanwhile, comparing Qstd values obtained by MA-Net and iMA-Net show that in all networks iMA-Net has smaller deviation, indicating that the improvements in iMA-Net enhance the stability and accuracy of the algorithm. Comparisons between the Qavg values obtained by iMA-Net and the other algorithms show that in all cases iMA-Net achieved higher Qavg and performed better than four representative algorithms, LPAm, Meme-Net, Moga-Net and MODPSO. Only MLCD, which is a novel multi-level learning memetic algorithm has a better performance than iMA-Net in benchmark networks. As the MLCD method is not publicly available and iMA-Net shows outstanding performance in detecting the high quality community structure with great stability, in this study we use the iMA-Net as the community detection algorithm. It is important to note that the clustering methodology is a generic method and any community detection algorithm can be employed in the third stage. Stage 4- Partition Quality Evaluation To evaluate the quality of the solution obtained by our clustering method, we can compare the solution with the known authorship label of each play. In other words, we expect that a good solution would put the plays written by the same author in one group. Thus, we are considering the authorship as the true label of each play and to evaluate the solution quality we used two well-known clustering similarity measures: Normalized Mutual Information (NMI) and Adjusted Rand Index (ARI). NMI: A symmetric index computing the similarity between two clustering solutions based on the confusion matrix (also referred to as the contingency matrix). As defined in [72], for two clustering solutions of a given graph A and B, the NMI is defined as follows: NMI(A,B)=-2∑i=1cA∑j=1cBFijlogFij.nFirow·Fjcol∑i=1cAFirowlog(Firown)+∑j=1cBFjcollog(Fjcoln)(7) where F is the confusion matrix for solutions A and B in which rows and columns correspond to the clusters in A and B, respectively. Fij is the node-overlap ith cluster of A and jth cluster of B. Firow is the sum of the elements in the row i of F and Fjcol denotes the sum of elements in the column j of F. cA and cB are the number of clusters in A and B, respectively. The NMI measures the similarity of two solutions, therefore if A = B, then NMI(A,B) = 1 and if there is no similarity between A and B then NMI(A,B) = 0. ARI: The adjusted version of the Rand Index(RI), first introduced in [73], compares two clusterings based on the number of cluster membership agreements and disarrangements between them. It shows the ratio of the number of node pairs similarly classified in both solutions, divided by the total number of pairs. RI is defined as RI(A,B)=α+βn2,(8) where n is the number of nodes, α refers to the number of pairs of nodes which are in the same cluster in solution A and in the same cluster in solution B and β is the number of pairs of nodes that are classified in different clusters in both solutions. The RI lies between 0 and 1 and it is equal to 1 when two solutions are exactly the same. The ARI was proposed by Hubert and Arabie [74]. To adjust the RI they assumed the generalized hypergeometric distribution as a null model and defined ARI as follows: ARI(A,B)=∑i,jFij2-∑iFirow2∑jFjcol2/n212∑iFirow2+∑jFjcol2-∑iFirow2∑jFjcol2/n2(9) where, similar to Eq 7, Fij, Firow and Fjcol are extracted from the confusion matrix F. Like the NMI, a larger ARI shows that two solutions are more similar. The ARI attracted more attention than RI as it is in the range [−1, +1], the wider range of values increasing the sensitivity of the index. To combine the effectiveness of NMI and ARI, we also used their product, NMI × ARI, for evaluation and comparing different clusterings. Results We applied iMA-Net to a series of ten kNN graphs (for k = 1 to k = 10) to optimize the modularity value and find the best solution that partitions the set of 168 plays written by 39 authors. The computer used for these experiments has the same configuration as stated in the section ‘Performance results of iMA-Net on real-world networks’. We ran iMA-Net 30 times for each graph with the following parameters: population size Np is set to 40, the termination criterion is set to 30 generations without improvement (Ns = 30) and the minimum mutation probability is equal to 0.05 (Pm0). The best solution found in 30 runs of the algorithm, which has the highest value of modularity, is recorded in Table 5. Quality measures including NMI, ARI and NMI × ARI, representing the similarity between best solutions and the true clustering solution obtained from the authorship labels, are shown in Table 5. 10.1371/journal.pone.0157988.t005Table 5 The best solution found by iMA-Net in ten kNN graphs derived from 168 Shakespearean era play dataset. kNN graph Q # Clusters NMI ARI NMI×ARI k = 1 0.898 24 0.686 0.317 0.217 k = 2 0.751 18 0.742 0.525 0.390 k = 3 0.713 10 0.690 0.457 0.315 k = 4 0.670 9 0.673 0.441 0.296 k = 5 0.623 9 0.663 0.431 0.286 k = 6 0.579 9 0.670 0.429 0.288 k = 7 0.542 9 0.672 0.434 0.292 k = 8 0.509 8 0.639 0.396 0.253 k = 9 0.482 8 0.639 0.401 0.256 k = 10 0.455 8 0.623 0.392 0.244 Q is the modularity value used as a fitness value of the clustering. NMI, ARI and NMI×ARI are quality measures used to compare with the true solution of the dataset based on the authorship of plays. The results in Table 5 demonstrate that by increasing the value of k and adding more connections to the graph, the number of the detected clusters and the solution modularity (Q) is reduced. Because of the dependency of the highest value of Q on the structure of the graph, it cannot be used for comparing solutions from different kNN graphs. But we can refer to quality measures (NMI and ARI) to find in which graphs iMA-Net finds the solution that better matches authorship labels. The results show that the best-matched solution with the highest quality measures, i.e. NMI = 0.742 and ARI = 0.525, is obtained by analyzing the 2-nearest neighbor (2NN) graph and contains 18 clusters, as is shown in Fig 3. 10.1371/journal.pone.0157988.g003Fig 3 Best clustering outcome of iMA-Net with the highest NMI and ARI. 18 clusters are detected in 2-nearest neighbor graph shown by different colours. Node size is proportional to the node’s total degree. Nodes are labeled by the play.author. Fig 3 presents the best solution in 2NN graph where each node represents a unique play. The play’s author is shown as the label of the node and the detected clusters are identified by different colors. Lists of the plays located in each cluster of this clustering solution together with the confusion matrix of the solution and the true solution is available in S2 and S3 Tables. Moreover, Fig 3 shows that the solution contains five clusters dedicated to only one author, where all plays in these clusters are written by the same author. The five finely detected authors are Dekker (Cluster 3), Ford (Cluster 6), Lyly (Cluster 16), Chapman (Cluster 11) and Webster (Cluster 14). It also contains a good classification of authors with more contributions in the dataset; for instance the biggest cluster (Cluster 5) with 23 plays includes 20 plays attributed to Shakespeare together with three plays from authors that have only one contribution in the dataset. Moreover, we have a very good clustering of the plays written by four authors Fletcher (Cluster 1), Middleton (Cluster 13), Jonson (Cluster 15) and Chapman (Cluster 17). These authors are separately clustered excepting only one play by another author included in each cluster. Furthermore, considering authors with limited contributions in this dataset and the way these works grouped with other plays reveals some interesting insights about their literary style. For instance, the single Tourneur play in the dataset is classified in Cluster 5 with 20 plays of Shakespeare. Commentators have seen Tourneur as a follower of Shakespeare [75], and it is likely that in this case the influence of Shakespeare’s style explains the placement of the play within the cluster. Another example is the clustering of the play of Marmion and the single play of Brome in the same group (Cluster 12) with 10 plays attributed to Middleton that illustrate the similarity between their word usage. Reduced Datasets Referring to Table 1, the complete dataset contains 19 authors with only one attributed play. iMA-Net classified these plays with the most similar plays, maximizing the modularity value. However, this behavior reduces the quality measures (NMI and ARI). To better understand the performance of the proposed methodology, we applied the same stages on reduced datasets resulting from removing some plays from the original dataset. The five reduced datasets generated by removing plays written by authors who have 1, 2, 3, 4 and 5 contributions are as follows: G1: Plays from authors with more than one contribution. G2: Plays from authors with more than two contributions. G3: Plays from authors with more than three contributions. G4: Plays from authors with more than four contributions. G5: Plays from authors with more than five contributions. Table 6 shows the number of plays and authors that remained in each of the reduced datasets. We applied our four-stage method to the five reduced datasets (G1-G5) and the results are given in Table 7. 10.1371/journal.pone.0157988.t006Table 6 Configuration of reduced datasets. Reduced dataset # Plays # Authors G1 149 20 G2 139 15 G3 130 12 G4 126 11 G5 106 7 10.1371/journal.pone.0157988.t007Table 7 Best solutions found by iMA-Net in 10 kNN graphs derived from each reduced dataset (G1-G5). The highest values of NMI, ARI and NMI×ARI in each dataset are denoted in bold. Reduced dataset kNN graph Q # Clusters NMI ARI NMI×ARI G1 k = 1 0.907 23 0.696 0.371 0.258 k = 2 0.753 16 0.738 0.505 0.373 k = 3 0.713 10 0.659 0.451 0.297 k = 4 0.672 9 0.701 0.505 0.354 k = 5 0.625 10 0.711 0.542 0.386 k = 6 0.583 9 0.669 0.457 0.306 k = 7 0.541 9 0.676 0.463 0.313 k = 8 0.511 8 0.628 0.415 0.261 k = 9 0.476 7 0.642 0.438 0.281 k = 10 0.447 8 0.629 0.424 0.267 G2 k = 1 0.903 22 0.707 0.393 0.278 k = 2 0.765 16 0.741 0.499 0.370 k = 3 0.722 11 0.717 0.522 0.374 k = 4 0.680 9 0.702 0.520 0.365 k = 5 0.630 10 0.717 0.564 0.404 k = 6 0.588 9 0.683 0.495 0.338 k = 7 0.552 9 0.672 0.492 0.331 k = 8 0.506 8 0.649 0.454 0.295 k = 9 0.476 8 0.683 0.501 0.342 k = 10 0.453 8 0.672 0.464 0.312 G3 k = 1 0.892 18 0.660 0.377 0.249 k = 2 0.767 15 0.773 0.542 0.419 k = 3 0.729 10 0.711 0.552 0.392 k = 4 0.682 11 0.740 0.560 0.414 k = 5 0.636 10 0.744 0.617 0.459 k = 6 0.599 9 0.694 0.542 0.376 k = 7 0.561 9 0.687 0.526 0.361 k = 8 0.520 9 0.678 0.509 0.345 k = 9 0.488 8 0.690 0.536 0.370 K = 10 0.460 7 0.632 0.448 0.283 G4 k = 1 0.893 19 0.646 0.343 0.222 k = 2 0.766 15 0.777 0.554 0.431 k = 3 0.762 16 0.772 0.552 0.426 k = 4 0.681 11 0.731 0.554 0.405 k = 5 0.639 10 0.740 0.628 0.465 k = 6 0.604 9 0.708 0.595 0.421 k = 7 0.563 9 0.689 0.536 0.369 k = 8 0.523 9 0.680 0.522 0.355 k = 9 0.490 7 0.631 0.431 0.272 k = 10 0.464 8 0.663 0.498 0.330 G5 k = 1 0.895 17 0.733 0.451 0.331 k = 2 0.781 11 0.826 0.618 0.511 k = 3 0.751 9 0.758 0.587 0.445 k = 4 0.701 8 0.843 0.730 0.615 k = 5 0.656 7 0.856 0.810 0.693 k = 6 0.627 7 0.847 0.814 0.690 k = 7 0.584 7 0.847 0.814 0.690 k = 8 0.544 7 0.837 0.787 0.659 k = 9 0.514 6 0.799 0.690 0.551 k = 10 0.482 6 0.808 0.706 0.570 From the higher values of NMI and ARI in Table 7, compared to their values in the original dataset, Table 5, we can conclude that by removing authors with fewer contributions from the original dataset, the proposed clustering method found solutions with higher quality with clusters better matched to the authors’ labels. More precisely, while the average NMI and average ARI in the original dataset are 0.670 and 0.422, respectively. The average NMI increased to 0.718 and the average value of ARI to 0.538 in the reduced datasets. On the other hand, the results in Table 7 show that NMI and ARI are different quality measures and they do not always agree on the best solution among the ten solutions found in the kNN graphs. Thus, we consider NMI × ARI to combine the effect of two measure into a single index. Interestingly, NMI×ARI always attains the highest value in the 5-nearest neighbour (5NN) graphs in all of the five reduced datasets. The best solution according to NMI×ARI is found in the G5 dataset and in the 5NN graph where the highest NMI×ARI is equal to 0.693. Table 8 shows the confusion matrix of the true solution and the solution with the highest quality measures for NMI and NMI×ARI and Fig 4 represents this best clustering solution in the 5NN graph. For details of plays in each cluster refer to the S4 Table. 10.1371/journal.pone.0157988.t008Table 8 Confusion matrix of the true authorship and the clustering solution obtained by the 5NN graph with NMI × ARI = 0.693. The confusion matrix shows how 106 plays by 7 authors are distributed into 7 clusters. As expected, a good separation occurred in clusters 2, 4, 6 and 7, which are formed by plays of one specific author. Author # Plays Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6 Cluster 7 1 Shakespeare 28 28 0 0 0 0 0 0 2 Middleton 18 0 18 0 0 0 0 0 3 Jonson 17 0 0 15 0 2 0 0 4 Fletcher 15 1 0 0 14 0 0 0 5 Chapman 13 1 0 7 0 5 0 0 6 Lyly 8 1 0 0 0 0 7 0 7 Ford 7 0 0 0 0 0 0 7 Total 106 31 18 22 14 7 7 7 10.1371/journal.pone.0157988.g004Fig 4 The best clustering outcome of iMA-Net with the highest NMI × ARI in G5. Seven clusters are shown by different colours in the 5-nearest neighbour graph constructed from similarity between 106 plays from 7 authors. Nodes sizes are proportioned to their degree. Both Table 8 and Fig 4 show that plays written by Middleton (Cluster 2), Fletcher (Cluster 4), Lyly (Cluster 6) and Ford (Cluster 7) are classified in completely separated clusters. All 28 plays of Shakespeare in this dataset are in the biggest group, Cluster 1, together with three plays from different authors which are the comedy ‘Blind Beggar of Alexandria’ written by Chapman, the pastoral play ‘Faithful Shepherdess’ by Fletcher and a comedy from Lyly named ‘Woman in the Moon’. So far as we are aware no commentators have suggested previously that these three non-Shakespeare plays in Cluster 1 are particularly Shakespearean, but they are sometimes noted as unusual for their authors. In the present analysis ‘The Woman in the Moon’ is among the five nearest neighbors of ‘The Faithful Shepherdess’ and ‘The Blind Beggar of Alexandria’, thus these three plays are themselves connected in G5. The ‘Woman in the Moon’ is a departure for Lyly in that he moves to verse after a career writing prose comedies. It is his last-performed play. Critics note other important differences from Lyly’s normal practice as well [76]. None of the other seven Lyly plays in this dataset appear among the five plays with the highest similarity score and connected to ‘Woman in the Moon’. Moreover, though this play is normally classified as a comedy [77] the closest play according to JSD measure is a Shakespeare tragedy, ‘Romeo and Juliet’, and other closest is another tragedy, ‘Tis Pity She’s a Whore’, by John Ford, from a much later phase of drama—the first production of Ford’s play was in 1632, compared to 1593 for the Lyly play. In her edition of the play Leah Scragg suggests that the play has more similarities to the rest of the Lyly canon than critics generally acknowledge, and has links to two other Lyly plays, ‘Endymion’ and ‘Mother Bombie’, in particular [76]; the results of the present analysis come down on the side of change rather than continuity in this play. In the context of a clustering in which author and genre likenesses come through strongly, and consistently, overall, the cross-author and cross-genre links for this play are worthy of further investigation, though that is beyond the scope of the present paper. ‘The Faithful Shepherdess’ is a pastoral, quite unlike Fletcher’s usual work in comedy and tragicomedy. In his studies of the shares of Fletcher and Beaumont in their collaborative plays Cyrus Hoy excludes ‘The Faithful Shepherdess’ from Fletcher baselines for that reason [78]. Jonathan Hope notes Fletcher’s departure from his usual style in ‘The Faithful Shepherdess’, which he calls “a consciously literary and deliberately archaic piece” [79]. Thus it is not surprising to find it clustered away from the main Fletcher cluster. Two Fletcher plays appear in the nearest neighbors of ‘The Faithful Shepherdess’, suggesting that the analysis has uncovered some authorial characteristics in the play, even if they are not sufficient to have it join the Fletcher cluster (Cluster 4). Something similar may be said about the ‘Blind Beggar of Alexandria’, which appears with two other Chapman comedies in the list of plays with the highest similarity score. The ‘Blind Beggar of Alexandria’ is Chapman’s first performed play and has often been viewed negatively as overly farcical with an underdeveloped romantic plot [80]. It is most likely that these three plays are clustered with Shakespeare because they have weak links to their authorial canons, and the Shakespeare cluster is large and therefore likely to be diverse, with many options for links. Two other clusters (Cluster 3 and Cluster 5) are formed by mixing all 17 of Jonson’s plays and 12 plays written by Chapman. A deeper look at the plays in these two clusters shows that the genre of the plays has a great role in putting similar works in the same cluster. The larger cluster (Cluster 3) contains 21 comedies of both Jonson and Chapman together with the only pastoral play written by Jonson named (‘Sad Shepherd’). However, Cluster 5 contains all of six tragedy plays written by these two authors and a historical play of Chapman which is called ‘Caesar and Pompey’. Both Jonson and Chapman would be recognized as having drastically different styles in their own comedies compared to their own tragedies, so this would fit with general critical opinion. Taking all the clusters found by our methodology into account, authorship is generally very strong on these measures, but a few authors have either an aberrant play or a consistent internal division by genre. Chapman has both of these attributes. The clusters identified by the proposed unsupervised method show considerable associations with the authors’ writing styles. Nevertheless and in order to validate the effectiveness of our method, we applied two available clustering methods on the dataset and compared the results. Performance comparison with other clustering methods The proposed method provides solutions for the clustering of plays in the dataset based on the similarity of word frequencies in each play. To compare our method with other clustering methods, we applied six well-known distance-based clustering techniques on the Complete dataset and the five reduced datasets (G1-G5). The selected clustering methods for comparisons are: a well-known implementation of the k-means method named k-means++ [81], one unsupervised graph-based clustering algorithm named MST-kNN [32], and four popular hierarchical clustering methods: 1)Complete-linkage, 2)Average-linkage, 3)Single-linkage and 4)Ward’s algorithm. k-means is by far the most popular clustering method used in many scientific and industrial applications [82, 83]. k-means++ developed by Arthur [81], is a novel k-means algorithm with an enhanced seeding technique for outperforming the standard k-means. Due to the popularity of k-means methods, k-means++ is selected for comparison. k-means++ is implemented in the Scikit-learn(0.16.1) clustering package [84] in Python and is publicly available. It is worth noting that k-means++, according to the original method, uses the Euclidean distance and identifies clusters by minimizing the average square distance between elements in the same clusters [81]. This method is a supervised method, in the sense that the number of clusters (K) is known in advance; therefore, we run the algorithm for K = 2 to K = 20 and report the best solution based on the quality measure NMI×ARI. MST-kNN is an unsupervised graph clustering technique proposed by Inostroza-Ponta et. al [32]. This method works by partitioning the minimum spanning tree using the k-nearest neighbour graph with an adaptive determination of the number of clusters. The MST-kNN is chosen for comparison because for two main reasons: 1) it can use the JSD matrix and compute the clusters based on this distance (dissimilarity) matrix, 2) MST-kNN also uses a similar approach of using graph structure for data clustering. This method has been employed in several studies [85–87] and it has shown remarkable results in data analysis. While MST-kNN is unsupervised, it is applied directly on JSD matrices and the obtained clusters is evaluated regarding the true partitions. Hierarchical clustering methods seek the hierarchical structure of the graph by recursively partitioning nodes into clusters either in top-down fashion by dividing clusters into smaller clusters (divisive hierarchical clustering) or bottom-up fashion by merging clusters into larger cluster (agglomerative hierarchical clustering). Dissimilarity measures and linkage criteria are used in order to select which cluster should be merged or divided. The linkage criterion refers to the manner in which the distance between two clusters is calculated. The most well-known linkage criteria are complete-linkage, average-linkage and single-linkage [88]. The distance between two cluster according to the complete-linkage criterion is equal to the longest distance between any pair of nodes from the two clusters [89], but based on the single-linkage criteria it is equal to the shortest distance between any pair of nodes from the two clusters [90] and in average-linkage methods the distance between two cluster is defined as the average distance of any member of one cluster to any member of the other cluster [91]. In this study, we use the stats (3.3.0) package (hclust function) provided in R. The JSD is employed as the dissimilarity measure and for the linkage criterion four methods have been used: Complete-linkage, Average-linkage, Single-linkage and Ward’s method [92, 93]. Ward’s method is a modified average-linkage method that works based on the squared dissimilarity. Hierarchical methods result in dendrograms, representing the nested grouping of nodes. The clustering solution of the data elements is obtained by cutting the dendrogram at the desired level. Therefore in this study to compare the clustering result of hierarchical methods with the true partitions and other methods, the obtained dendrogram was cut to get the desired number of clusters. To give the most possible chance to the dendrogram to provide the best solution the number of clusters was varied from K = 2 to K = 20. Table 9 shows the best and worst result of the proposed clustering methods together with the best results obtained by k-means++, MST-kNN and four hierarchical clustering methods (i.e. complete-linkage, average-linkage, single-linkage and Ward’s method). The quality of the clustering results are evaluated by NMI, ARI and NMI×ARI which are computed by comparing the clusters with the true authorship label of the plays in each dataset. The Best and the Worst results are repetitions from Table 5 (for Complete dataset) and Table 7 (for reduced datasets). Methods are ranked in each dataset based on NMI × ARI that shows how successfully the method obtained the true partition. 10.1371/journal.pone.0157988.t009Table 9 The best clustering solutions obtained by benchmark methods (k-means++, MST-kNN, Complete-linkage, Average-linkage, Single-linkage and Ward’s method) together with the best (Best) and the worst (Worst) clustering result obtained by the proposed method in Complete dataset and five reduced datasets (G1-G5). The Rank column is based on the value of NMI × ARI. Dataset Method # Clusters NMI ARI NMI × ARI Rank Complete Best 18 0.742 0.525 0.390 2 Worst 24 0.686 0.317 0.217 3 k-means++ 13 0.617 0.339 0.209 4 MST-kNN 2 0.292 0.026 0.008 8 Complete-linkage 20 0.623 0.324 0.202 5 Average-linkage 20 0.440 0.052 0.023 6 Single-linkage 20 0.384 0.033 0.013 7 Ward’s method 20 0.772 0.568 0.438 1 G1 Best 10 0.711 0.542 0.386 2 Worst 8 0.628 0.415 0.261 3 k-means++ 18 0.638 0.354 0.226 4 MST-kNN 3 0.435 0.090 0.039 6 Complete-linkage 20 0.591 0.181 0.107 5 Average-linkage 20 0.434 0.065 0.028 7 Single-linkage 19 0.388 0.052 0.020 8 Ward’s method 19 0.775 0.570 0.442 1 G2 Best 10 0.717 0.564 0.404 1 Worst 8 0.649 0.454 0.295 3 k-means++ 16 0.628 0.395 0.248 4 MST-kNN 3 0.469 0.132 0.062 6 Complete-linkage 19 0.562 0.191 0.107 5 Average-linkage 20 0.467 0.098 0.046 7 Single-linkage 20 0.392 0.064 0.025 8 Ward’s method 18 0.765 0.507 0.387 2 G3 Best 10 0.744 0.617 0.459 1 Worst 18 0.660 0.377 0.249 4 k-means++ 12 0.626 0.448 0.280 3 MST-kNN 3 0.476 0.147 0.070 6 Complete-linkage 19 0.574 0.224 0.129 5 Average-linkage 20 0.491 0.125 0.061 7 Single-linkage 20 0.412 0.081 0.033 8 Ward’s method 14 0.740 0.535 0.396 2 G4 Best 10 0.740 0.628 0.465 2 Worst 7 0.631 0.431 0.222 4 k-means++ 17 0.640 0.389 0.249 3 MST-kNN 3 0.486 0.162 0.079 6 Complete-linkage 20 0.612 0.264 0.162 5 Average-linkage 18 0.400 0.080 0.032 8 Single-linkage 20 0.517 0.142 0.074 7 Ward’s method 13 0.752 0.619 0.465 1 G5 Best 7 0.856 0.810 0.693 1 Worst 17 0.733 0.451 0.331 3 k-means++ 10 0.600 0.422 0.253 5 MST-kNN 3 0.540 0.237 0.128 6 Complete-linkage 15 0.665 0.465 0.310 4 Average.linkage 10 0.449 0.124 0.056 7 Single.linkage 16 0.362 0.075 0.027 8 Ward’s method 9 0.830 0.797 0.661 2 Firstly, the best results of the proposed method (Best), in the three of six datasets, i.e. G2, G3 and G5 datasets, are ranked first and outperform the other methods. In these three datasets the results of Ward’s method are in the second rank. In the other three datasets, i.e. Complete, G2 and G4 datasets, Ward’s method is ranked first and Best is in the second place. Secondly, comparison between the benchmark methods shows that MST-kNN has an obvious tendency to merge small clusters. Hence, it results in a solution with fewer clusters. On the other hand, the three hierarchical clustering methods, i.e. Complete-linkage, Average-linkage and Single-linkage, result in solutions with more clusters. However, among the three traditional hierarchical clustering methods the Complete-linkage method obtained better results than Average-linkage and Single-linkage in all datasets. Finally, comparison between the worst results of the proposed method (Worst) with other benchmark methods indicates that in four of the datasets, i.e. Complete, G1, G2, G3 and G5, Worst is ranked 3 after Best and Ward’s. Another striking observation about the Worst results is that in all datasets, it is better than MST-kNN, Complete-linkage, Average-linkage and Single-linkage. Furthermore, comparison between the performance of the popular method k-means++ with Worst illustrates that in four datasets, Worst is ranked better than k-means++, and only in G3 and G4 does k-means++ show slightly better performance than Worst. Discussion Significance and Contribution In this paper, we proposed a new four-stage methodology for data clustering. The methodology is unsupervised and it is able to identify groups of data elements which have a meaningful similarity. Though several methods have been proposed for two problems of data clustering and community detection, in this study, we introduced a new general methodology to convert the data clustering problem to the community detection problem by borrowing fundamental concepts of information theory and graph theory. This methodology aims to identify novel clusters from data. To evaluate the performance of our methodology, we have conducted several experiments on an interesting corpus dataset generated by counting word frequencies in 168 Shakespearean plays. Clusters obtained from datasets (Complete and reduced datasets) revealed stimulating findings about author’s literary styles. Moreover, the memetic algorithm (iMA-Net) which is proposed optimizing the modularity value shows efficient performance in discovering solutions in benchmark networks. Also, in the kNN graphs generated for the Complete dataset and reduced datasets, the iMA-Net obtained good clustering results that are highly matched with the true partitions of the datasets. Furthermore, comparison with six well-known clustering techniques (Table 9) illustrates the great ability of the developed methodology in detecting superior clustering solutions in the studied datasets. Finally, the proposed methodology is a generic method and in each of the four stages a variety of other methods can be employed. For instance, in the third stage of identifying the community structure of the kNN graphs, variety of community detection methods might be applicable. Limitations and future research Although the proposed methodology obtains remarkable results in studied datasets, there are a few limitations, which are possible areas for future work. Firstly, we utilized the square root of Jensen-Shannon divergence (JSD) to measure the dissimilarity between elements and then convert the dissimilarity (distance) in kNN graphs to the similarity by subtracting it from one; it performs competently in the studied datasets, but more studies on different datasets and comparison with other measures are required to prove that JSD is the most suitable measure for constructing the kNN graphs. Secondly, there is a limitation regarding the number of kNN graphs that should be analyzed. In the second stage of the proposed methodology, the kNN graphs were generated from the dissimilarity matrix and we set the value of k to be in the range of 1 to 10. As mentioned before, comparison between the best found clustering in the five reduced datasets (Table 7) showed that solutions with the highest value of NMI×ARI were always obtained from the 5NN graph. However more investigations are required to justify that analyzing 5NN graphs has an advantage over the other kNN graphs. One of the future research directions would be finding a proper procedure for tuning the value of k in different datasets. Note that the proposed clustering methodology is a four-stage procedure and for each stage different tools are applicable. This has its advantages and disadvantages. While the main advantage is the flexibility of the method to employ different tools, the disadvantage is its complexity in combining many different tools. Future work will aim to develop a single software tool for the proposed clustering approach that can apply all four stages in an optimized and scalable way. Another future research direction is to employ multi-objective community detection algorithms that optimize not only modularity value but also other quality functions. As an example, we can refer to the multi-objective community detection method proposed by Shi et al. [94]. Multi-objective algorithms return a set of non-dominated solutions and give the opportunity for the user to select the most appropriate solution from a few. Furthermore, applying overlapping community detection methods in the third stage is another important future research direction, which will enhance the ability of the proposed methodology to identify overlapping clusters; the latter has several real-world applications. To conclude, we proposed a novel four-stage methodology for clustering based on information theory and modularity optimization. To demonstrate the effectiveness of this methodology, we conducted experiments on clustering a large text corpus dataset. We discovered remarkable results which lead to the authors’ literary styles in different genres. As discussed, we envision that the proposed efficient methodology might be adopted in various fields for the purpose of investigating the clustering structure of datasets. Supporting Information S1 Table The 168×168 matrix contains the JSD between each pair of 168 plays. (XLSX) Click here for additional data file. S2 Table Confusion Matrix of the best solution that is demonstrated in Fig 3. (XLSX) Click here for additional data file. S3 Table List of plays located in each cluster of the solution demonstrated in Fig 3. (XLSX) Click here for additional data file. S4 Table List of plays located in each cluster of the best clustering solution of G5 which is demonstrated in Fig 4. (XLSX) Click here for additional data file. We would like to thank Renato Vimieiro and Carlos Riveros for providing the R package for JSD matrix computation, and to Luke Mathieson for his extensive proof-reading and constructive comments. We also thank the four anonymous reviewers for their valuable comments. ==== Refs References 1 Baldi P , Hatfield GW . DNA microarrays and gene expression: from experiments to data analysis and modeling . Cambridge University Press ; 2002 . 2 Clark MB , Johnston RL , Inostroza-Ponta M , Fox AH , Fortini E , Moscato P , et al Genome-wide analysis of long noncoding RNA stability . Genome research . 2012 ;22 (5 ):885 –898 . 10.1101/gr.131037.111 22406755 3 Arabie P . Cluster analysis in marketing research Advanced methods in marketing research . 1994 ;p. 160 –189 . 4 de Vries NJ , Carlson J , Moscato P . A Data-Driven Approach to Reverse Engineering Customer Engagement Models: Towards Functional Constructs . PLoS ONE . 2014 7 ;9 (7 ):1 –19 . 10.1371/journal.pone.0102768 5 Aggarwal CC , Reddy CK . Data clustering: algorithms and applications . CRC ; 2013 . 6 Han J , Kamber M , Pei J . Data mining: concepts and techniques . Elsevier ; 2011 . 7 Smyth P . Model selection for probabilistic clustering using cross-validated likelihood . Statistics and Computing . 2000 ;10 (1 ):63 –72 . 10.1023/A:1008940618127 8 Hathaway RJ , Bezdek JC . Extending fuzzy and probabilistic clustering to very large data sets . Computational Statistics & Data Analysis . 2006 ;51 (1 ):215 –234 . 10.1016/j.csda.2006.02.008 9 Jain AK . Data clustering: 50 years beyond K-means . Pattern Recognition Letters . 2010 ;31 (8 ):651 –666 . 10.1016/j.patrec.2009.09.011 10 Ester M, Kriegel HP, Sander J, Xu X. A density-based algorithm for discovering clusters in large spatial databases with noise. In: KDD. vol. 96; 1996. p. 226–231. 11 Amini A, Wah TY, Saybani MR, Yazdi SRAS. A study of density-grid based clustering algorithms on data streams. In: Fuzzy Systems and Knowledge Discovery (FSKD), 2011 Eighth International Conference on. vol. 3. IEEE; 2011. p. 1652–1656. 12 Ng AY , Jordan MI , Weiss Y , et al On spectral clustering: Analysis and an algorithm . Advances in neural information processing systems . 2002 ;2 :849 –856 . 13 Von Luxburg U . A tutorial on spectral clustering . Statistics and Computing . 2007 ;17 (4 ):395 –416 . 10.1007/s11222-007-9033-z 14 Schaeffer SE . Graph clustering . Computer Science Review . 2007 ;1 (1 ):27 –64 . 10.1016/j.cosrev.2007.05.001 15 Chan PK , Schlag MD , Zien JY . Spectral k-way ratio-cut partitioning and clustering . Computer-Aided Design of Integrated Circuits and Systems, IEEE Transactions on . 1994 ;13 (9 ):1088 –1096 . 10.1109/43.310898 16 Shi J , Malik J . Normalized cuts and image segmentation . Pattern Analysis and Machine Intelligence, IEEE Transactions on . 2000 ;22 (8 ):888 –905 . 10.1109/34.868688 17 Newman ME , Girvan M . Finding and evaluating community structure in networks . Physical Review E . 2004 ;69 (2 ):026113 10.1103/PhysRevE.69.026113 18 Girvan M , Newman ME . Community structure in social and biological networks . Proceedings of the National Academy of Sciences . 2002 ;99 (12 ):7821 –7826 . 10.1073/pnas.122653799 19 Fortunato S . Community detection in graphs . Physics Reports . 2010 ;486 (3 ):75 –174 . 10.1016/j.physrep.2009.11.002 20 Andreev K, Räcke H. Balanced Graph Partitioning. In: Proceedings of the Sixteenth Annual ACM Symposium on Parallelism in Algorithms and Architectures. ACM; 2004. p. 120–124. 21 Feldmann AE . Fast balanced partitioning is hard even on grids and trees In: Mathematical Foundations of Computer Science . Springer ; 2012 p. 372 –382 . 22 Elliott WE , Valenza RJ . Oxford by the Numbers: What Are the Odds that the Earl of Oxford Could Have Written Shakespeare’s Poems and Plays . Tennessee Law Review . 2004 ;72 :323 . 23 Vickers B . ‘Counterfeiting’ Shakespeare: Evidence, Authorship and John Ford’s Funerall Elegye . Cambridge University Press ; 2002 . 24 Jackson MP . Determining the Shakespeare Canon: Arden of Faversham and A Lover’s Complaint . Oxford University Press ; 2014 . 25 Foucault M . What is an author? In: Faubion JD , trans Harari Josué V , editors. Aesthetics, Method, and Epistemology . The New Press ; 1994 p. 205 –21 . 26 Barthes R . The Death of the Author Image / Music / Text . 1977 ;p. 142 –148 . 27 Craig H . Style, statistics, and new models of authorship . Early Modern Literary Studies . 2009 ;15 (1 ):1 –42 . 28 Egan G. What Is Not Collaborative About Early Modern Drama In Performance And Print? In: Holland P, editor. Shakespeare Survey. vol. 67. Cambridge University Press; 2014. p. 18–28. Shakespeare Survey Online. 29 Craig H , Whipp R . Old spellings, new methods: automated procedures for indeterminate linguistic data . Literary and Linguistic Computing . 2010 ;25 (1 ):37 –52 . 10.1093/llc/fqp033 30 Marsden J , Budden D , Craig H , Moscato P . Language individuation and marker words: Shakespeare and his Maxwell’s demon . PLoS ONE . 2013 ;8 (6 ):e66813 10.1371/journal.pone.0066813 23826143 31 Arefin AS , Riveros C , Berretta R , Moscato P . The MST-kNN with Paracliques In: Artificial Life and Computational Intelligence . Springer ; 2015 p. 373 –386 . 32 Inostroza-Ponta M , Mendes A , Berretta R , Moscato P . An integrated QAP-based approach to visualize patterns of gene expression similarity In: Progress in Artificial Life . Springer ; 2007 p. 156 –167 . 33 Naeni LM, de Vries NJ, Reis R, Arefin AS, Berretta R, Moscato P. Identifying Communities of Trust and Confidence in the Charity and Not-for-profit Sector: A Memetic Algorithm Approach. In: Big Data and Cloud Computing (BdCloud), 2014 IEEE Fourth International Conference on; 2015. p. 500–507. 34 Rao CR , Nayak TK . Cross entropy, dissimilarity measures, and characterizations of quadratic entropy . Information Theory, IEEE Transactions on . 1985 ;31 (5 ):589 –593 . 10.1109/TIT.1985.1057082 35 Shannon CE . A mathematical theory of communication . Bell System Technical Journal . 1948 7 ;27 (3 ):379 –423 . 10.1002/j.1538-7305.1948.tb01338.x 36 Lin J . Divergence measures based on the Shannon entropy . Information Theory, IEEE Transactions on . 1991 ;37 (1 ):145 –151 . 10.1109/18.61115 37 Berretta R , Moscato P . Cancer biomarker discovery: the entropic hallmark . PLoS ONE . 2010 ;5 (8 ):e12262 10.1371/journal.pone.0012262 20805891 38 Grosse I , Bernaola-Galván P , Carpena P , Román-Roldán R , Oliver J , Stanley HE . Analysis of symbolic sequences using the Jensen-Shannon divergence . Physical Review E . 2002 ;65 (4 ):041905 10.1103/PhysRevE.65.041905 39 Sims GE , Jun SR , Wu GA , Kim SH . Alignment-free genome comparison with feature frequency profiles (FFP) and optimal resolutions . Proceedings of the National Academy of Sciences . 2009 ;106 (8 ):2677 –2682 . 10.1073/pnas.0813249106 40 Endres DM , Schindelin JE . A new metric for probability distributions . IEEE Transactions on Information theory . 2003 ;. 10.1109/TIT.2003.813506 41 Österreicher F , Vajda I . A new class of metric divergences on probability spaces and its applicability in statistics . Annals of the Institute of Statistical Mathematics . 2003 ;55 (3 ):639 –653 . 10.1007/BF02517812 42 Arefin AS , Riveros C , Berretta R , Moscato P . GPU-FS-k NN: A Software Tool for Fast and Scalable k NN Computation Using GPUs . PLoS ONE . 2012 ;7 (8 ):e44000 10.1371/journal.pone.0044000 22937144 43 Gan G , Ma C , Wu J . Data clustering: theory, algorithms, and applications . vol. 20 Siam ; 2007 . 44 Watts DJ . A twenty-first century science . Nature . 2007 ;445 (7127 ):489 –489 . 10.1038/445489a 17268455 45 Lazer D , Pentland AS , Adamic L , Aral S , Barabasi AL , Brewer D , et al Life in the network: the coming age of computational social science . Science . 2009 ;323 (5915 ):721 10.1126/science.1167742 19197046 46 Newman ME , Girvan M . Finding and evaluating community structure in networks . Physical Review E . 2004 ;69 (2 ):026113 10.1103/PhysRevE.69.026113 47 Brandes U, Delling D, Gaertler M, Görke R, Hoefer M, Nikoloski Z, et al. Maximizing modularity is hard. arXiv preprint physics/0608255. 2006;. 48 Clauset A , Newman ME , Moore C . Finding community structure in very large networks . Physical Review E . 2004 ;70 (6 ). 10.1103/PhysRevE.70.066111 49 Wakita K, Tsurumi T. Finding community structure in mega-scale social networks. In: Proceedings of the 16th international conference on World Wide Web. ACM; 2007. p. 1275–1276. 50 Blondel VD , Guillaume JL , Lambiotte R , Lefebvre E . Fast unfolding of communities in large networks . Journal of Statistical Mechanics: Theory and Experiment . 2008 ;2008 (10 ):P10008 10.1088/1742-5468/2008/10/P10008 51 Tasgin M, Herdagdelen A, Bingol H. Community detection in complex networks using genetic algorithms. arXiv preprint arXiv:07110491. 2007;. 52 Li J , Song Y . Community detection in complex networks using extended compact genetic algorithm . Soft Computing . 2013 ;17 (6 ):925 –937 . 10.1007/s00500-012-0942-1 53 Gach O , Hao JK . In: Coello CAC , Cutello V , Deb K , Forrest S , Nicosia G , Pavone M , editors. A memetic algorithm for community detection in complex networks . Springer Berlin Heidelberg ; 2012 p. 327 –336 . 54 Shang R , Bai J , Jiao L , Jin C . Community detection based on modularity and an improved genetic algorithm . Physica A: Statistical Mechanics and its Applications . 2013 ;392 (5 ):1215 –1231 . 10.1016/j.physa.2012.11.003 55 Liu D , Jin D , Baquero C , He D , Yang B , Yu Q . Genetic algorithm with a local search strategy for discovering communities in complex networks . International Journal of Computational Intelligence Systems . 2013 ;6 (2 ):354 –369 . 10.1080/18756891.2013.773175 56 Newman ME . Modularity and community structure in networks . Proceedings of the National Academy of Sciences . 2006 ;103 (23 ):8577 –8582 . 10.1073/pnas.0601602103 57 Leicht EA , Newman ME . Community structure in directed networks . Physical Review Letters . 2008 ;100 (11 ):118703 10.1103/PhysRevLett.100.118703 18517839 58 Liu J , Liu T . Detecting community structure in complex networks using simulated annealing with k-means algorithms . Physica A: Statistical Mechanics and its Applications . 2010 ;389 (11 ):2300 –2309 . 10.1016/j.physa.2010.01.042 59 Naeni LM , Berretta R , Moscato P . In: Handa H , Ishibuchi H , Ong YS , Tan CK , editors. MA-Net: A Reliable Memetic Algorithm for Community Detection by Modularity Optimization . Cham : Springer International Publishing ; 2015 p. 311 –323 . 60 Neri F , Cotta C , Moscato P . Handbook of memetic algorithms . vol. 379 Springer ; 2012 . 61 Fortunato S , Barthélemy M . Resolution limit in community detection . Proceedings of the National Academy of Sciences . 2007 ;104 (1 ):36 –41 . 10.1073/pnas.0605965104 62 Schuetz P , Caflisch A . Efficient modularity optimization by multistep greedy algorithm and vertex mover refinement . Physical Review E . 2008 ;77 (4 ):046112 10.1103/PhysRevE.77.046112 63 Zachary WW . An information flow model for conflict and fission in small groups . Journal of Anthropological Research . 1977 ;p. 452 –473 . 10.1086/jar.33.4.3629752 64 Lusseau D , Schneider K , Boisseau OJ , Haase P , Slooten E , Dawson SM . The bottlenose dolphin community of Doubtful Sound features a large proportion of long-lasting associations . Behavioral Ecology and Sociobiology . 2003 ;54 (4 ):396 –405 . 10.1007/s00265-003-0651-y 65 Gleiser PM , Danon L . Community structure in jazz . Advances in Complex Systems . 2003 ;6 (04 ):565 –573 . 10.1142/S0219525903001067 66 Barber MJ , Clark JW . Detecting network communities by propagating labels under constraints . Physical Review E . 2009 ;80 (2 ):026129 10.1103/PhysRevE.80.026129 67 Gong M , Fu B , Jiao L , Du H . Memetic algorithm for community detection in networks . Physical Review E . 2011 ;84 (5 ):056101 10.1103/PhysRevE.84.056101 68 Pizzuti C . A multiobjective genetic algorithm to find communities in complex networks . Evolutionary Computation, IEEE Transactions on . 2012 ;16 (3 ):418 –430 . 10.1109/TEVC.2011.2161090 69 Gong M , Cai Q , Chen X , Ma L . Complex network clustering by multiobjective discrete particle swarm optimization based on decomposition . IEEE Transactions on Evolutionary Computation . 2014 ;18 (1 ):82 –97 . 10.1109/TEVC.2013.2260862 70 Ma L , Gong M , Liu J , Cai Q , Jiao L . Multi-level learning based memetic algorithm for community detection . Applied Soft Computing . 2014 ;19 :121 –133 . 10.1016/j.asoc.2014.02.003 71 Raghavan UN , Albert R , Kumara S . Near linear time algorithm to detect community structures in large-scale networks . Physical Review E . 2007 ;76 (3 ):036106 10.1103/PhysRevE.76.036106 72 Ana L, Jain AK. Robust data clustering. In: Computer Vision and Pattern Recognition, 2003. Proceedings. 2003 IEEE Computer Society Conference on. vol. 2. IEEE; 2003. p. II–128. 73 Rand WM . Objective criteria for the evaluation of clustering methods . Journal of the American Statistical Association . 1971 ;66 (336 ):846 –850 . 10.1080/01621459.1971.10482356 74 Hubert L , Arabie P . Comparing partitions . Journal of classification . 1985 ;2 (1 ):193 –218 . 10.1007/BF01908075 75 Eliot S T . Eassys on Elizabethan Drama . Harcourt, Brace and Company, Inc. ; 1956 . 76 Lyly J , Scragg L . The Woman in the Moon: By John Lyly. The Revels Plays . Manchester University Press ; 2006 . 77 Harbage A , Schoenbaum S . Annals of English Drama, 975–1700: An Analytical Record of All Plays, Extant Or Lost, Chronologically Arranged and Indexed by Authors, Titles, Dramatic Companies & C . 2nd ed University of Pennsylvania ; 1964 . 78 Hoy C . The Shares of Fletcher and His Collaborators in the Beaumont and Fletcher Canon (I) . Studies in Bibliography . 1956 ;8 :129 –146 . 79 Hope J . The authorship of Shakespeare’s plays: a socio-linguistic study . Cambridge University Press ; 1994 . 80 Kaufman HA . The Blind Beggar of Alexandria: A Reappraisal . Philological Quarterly . 1959 ;38 :101 . 81 Arthur D, Vassilvitskii S. k-means++: The advantages of careful seeding. In: Proceedings of the eighteenth annual ACM-SIAM symposium on Discrete algorithms. Society for Industrial and Applied Mathematics; 2007. p. 1027–1035. 82 Berkhin P . A survey of clustering data mining techniques In: Grouping multidimensional data . Springer ; 2006 p. 25 –71 . 83 Jacques J , Preda C . Functional data clustering: a survey . Advances in Data Analysis and Classification . 2014 ;8 (3 ):231 –255 . 10.1007/s11634-013-0158-y 84 Pedregosa F , Varoquaux G , Gramfort A , Michel V , Thirion B , Grisel O , et al Scikit-learn: Machine Learning in Python . Journal of Machine Learning Research . 2011 ;12 :2825 –2830 . 85 Arefin AS , Mathieson L , Johnstone D , Berretta R , Moscato P . Unveiling clusters of RNA transcript pairs associated with markers of Alzheimer’s disease progression . PLoS ONE . 2012 ;7 (9 ):e45535 10.1371/journal.pone.0045535 23029078 86 Capp A , Inostroza-Ponta M , Bill D , Moscato P , Lai C , Christie D , et al Is there more than one proctitis syndrome? A revisitation using data from the TROG 96.01 trial . Radiotherapy and Oncology . 2009 ;90 (3 ):400 –407 . 10.1016/j.radonc.2008.09.019 18952309 87 Inostroza-Ponta M , Berretta R , Moscato P . QAPgrid: A two level QAP-based approach for large-scale data analysis and visualization . PLoS ONE . 2011 ;6 (1 ):e14468 10.1371/journal.pone.0014468 21267077 88 Rokach L , Maimon O . Clustering methods In: Data mining and knowledge discovery handbook . Springer US ; 2005 p. 321 –352 . 89 King B . Step-wise clustering procedures . Journal of the American Statistical Association . 1967 ;62 (317 ):86 –101 . 10.1080/01621459.1967.10482890 90 Sneath PH , Sokal RR . Numerical taxonomy The principles and practice of numerical classification . W. H. Freeman ; 1973 . 91 Murtagh F. Multidimensional clustering algorithms. Compstat Lectures; 1985. 92 Ward JH . Hierarchical grouping to optimize an objective function . Journal of the American Statistical Association . 1963 ;58 (301 ):236 –244 . 10.1080/01621459.1963.10500845 93 Murtagh F , Legendre P . Ward’s hierarchical agglomerative clustering method: which algorithms implement Ward’s criterion? Journal of Classification . 2014 ;31 (3 ):274 –295 . 10.1007/s00357-014-9161-z 94 Shi C , Yan Z , Cai Y , Wu B . Multi-objective community detection in complex networks . Applied Soft Computing . 2012 ;12 (2 ):850 –859 . 10.1016/j.asoc.2011.10.005

PMC005xxxxxx/PMC5003343.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757152210.1371/journal.pone.0161649PONE-D-16-05039Research ArticleResearch and Analysis MethodsMathematical and Statistical TechniquesStatistical MethodsMeta-AnalysisPhysical SciencesMathematicsStatistics (Mathematics)Statistical MethodsMeta-AnalysisBiology and Life SciencesAnatomyBody FluidsBileBilirubinMedicine and Health SciencesAnatomyBody FluidsBileBilirubinBiology and Life SciencesPhysiologyBody FluidsBileBilirubinMedicine and Health SciencesPhysiologyBody FluidsBileBilirubinResearch and Analysis MethodsDatabase and Informatics MethodsDatabase SearchingMedicine and Health SciencesEndocrinologyEndocrine DisordersDiabetes MellitusMedicine and Health SciencesMetabolic DisordersDiabetes MellitusBiology and Life SciencesCell BiologyOxidative StressResearch and Analysis MethodsResearch DesignCohort StudiesMedicine and Health SciencesOphthalmologyRetinal DisordersRetinopathyDiabetic RetinopathyResearch and Analysis MethodsDatabase and Informatics MethodsThe Negative Relationship between Bilirubin Level and Diabetic Retinopathy: A Meta-Analysis The Relationship between Bilirubin Level and Diabetic Retinopathyhttp://orcid.org/0000-0002-1062-1295Zhu Bo 1*Wu Xiaomei 2Ning Kang 3Jiang Feng 4Zhang Lu 11 Department of Cancer Prevention and Treatment, Cancer Hospital of China Medical University, Liaoning Cancer Hospital & Institute, Shenyang, People’s Republic of China2 Department of Clinical Epidemiology and Evidence Medicine, The First Hospital of China Medical University, Shenyang, People’s Republic of China3 Department of Occupational Health, Liaoning Disease Prevention and Control Center, Shenyang, People’s Republic of China4 Center of Health Management, Shenyang 242 Hospital, Shenyang, People’s Republic of ChinaVavvas Demetrios G. EditorMassachusetts Eye & Ear Infirmary, Harvard Medical School, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: BZ XMW. Formal analysis: BZ XMW LZ. Funding acquisition: XMW. Investigation: BZ XMW KN. Methodology: BZ XMW. Project administration: XMW. Resources: XMW. Software: BZ. Validation: XMW FJ. Visualization: BZ LZ. Writing – original draft: BZ XMW. Writing – review & editing: BZ XMW. * E-mail: 15998896991@163.com29 8 2016 2016 11 8 e01616494 2 2016 9 8 2016 © 2016 Zhu et al2016Zhu et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Objectives Findings on the relationship between total bilirubin level (TBL) and diabetic retinopathy (DR) are inconsistent. Thus, we carried out a meta-analysis to investigate the relationship between TBL and the risk of DR. Methods Relevant studies were selected from six databases up to 31 May 2016 using a search strategy. The relevant data were extracted from the included studies according to the inclusion and exclusion criteria, and the mean value with standard errors or odds ratio (OR) with 95% confidence intervals (CIs) were calculated. We compared TBL in patients with DR with that in patients with diabetes but without retinopathy (NDR), and analyzed the dose-response relationship between TBL and the risk of DR. Results Twenty-four studies were selected in this meta-analysis. Twenty studies were included to calculate the pooled SMD, and the results showed that TBL in the DR group was lower than that in the NDR group (SMD: –0.52, 95% CI: –0.67, –0.38). Nine studies were included to calculate the pooled ORs, and the results showed that there was a significant negative relationship between TBL and the risk of DR (OR: 0.19, 95% CI: 0.14, 0.25). Six studies were included to investigate the dose-response relationship between TBL and the risk of DR, and we found a nonlinear relationship between TBL and the risk of DR. The results of our meta-analysis were found to be reliable using subgroup and sensitivity analyses. Conclusions The results of our meta-analysis indicate that higher TBL may be protective against DR in subjects with diabetes, and TBL could be used as a biomarker to predict the risk of DR. The Science and Technology General Project of Liaoning Education DepartmentL2015592Wu Xiaomei Our study was supported by the Science and Technology General Project of Liaoning Education Department (L2015592). Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction The incidence of type 2 diabetes mellitus (T2DM) is increasing rapidly. T2DM is a serious disease affecting quality of life and has become a health problem worldwide [1]. Diabetic retinopathy (DR) is a common microvascular complication of diabetes mellitus in people aged 36–69 years, and can lead to acquired blindness [2]. The incidence of DR in population-based studies ranges from 10 to 55%, and in clinic-based studies ranges from 11 to 65% [3]. Approximately 28.5% of blindness is attributed to DR in subjects aged 40 years and older [4]. More and more evidence has shown that oxidative stress is a critical risk factor in the pathogenesis of DR [5]. Oxidative stress is a pathologic state and reflects an imbalance between oxidation and antioxidation. Potent free radicals, such as superoxide radical, hydroxyl radical, and hydrogen peroxide radical, are commonly termed reactive oxygen species (ROS) and are regarded as extremely toxic [5]. The retina is a source of free radical production in T2DM, and the levels and types of oxidative stress products can predict the severity of retinopathy. The serum level of 8-OHdG in DR subjects is significantly increased compared to that in subjects with diabetes but without retinopathy (NDR) [6]. Bilirubin is mainly generated from heme degradation, and has strong antioxidant and anti-inflammatory effects on the microvasculature [7]. Bilirubin can covalently bind to albumin with a combination rate of more than 99%. It not only scavenges superoxide radical and peroxide radical, but also prevents oxidation modifications of low density lipoprotein and lipid oxidation. Animal studies have shown that bilirubin can accelerate wound healing by reducing the oxidant status of wounds in diabetic rats [6]. Some prospective studies have suggested that there is a negative relationship between total bilirubin level (TBL) and cardiovascular disease [8, 9]. A meta-analysis on the relationship between TBL and cardiovascular disease also confirmed this inverse relationship [10]. However, several studies have reported that there was no association between TBL and retinopathy of prematurity [11, 12]. In recent years, the number of studies on the relationship between TBL and DR has increased, but the effect of bilirubin on DR is unclear. Some studies showed that TBL in DR was higher than that in NDR [13–16], but other studies found no significant increase in patients with DR compared to those with NDR [17–19]. Therefore, we carried out a meta-analysis to analyze the effect of TBL on DR and evaluated the relationship between TBL and DR. To date, no meta-analyses on the relationship between TBL and DR have been carried out. Methods Search strategy and study selection We selected appropriate studies from multiple databases (PubMed, Web of Science, CNKI (China National Knowledge Infrastructure), Wanfang database, CBM (China Biological Medicine Database), and VIP) up to 31 May, 2016 using a search strategy. The search terms used were as follows: “bilirubin” and (“retinopathy” or “retina” or “eye disease” (The filter: bilirubin [Text Word]) AND (retinopathy [Text Word]) OR retina [Text Word]) OR eye disease [Text Word]). In order to select as many appropriate articles as possible, we also reviewed reference lists in all relevant original research and review articles to identify additional potentially eligible studies. DR was defined by an ophthalmologist following an eye examination. Fundus photographs taken for both eyes were analyzed and graded to confirm the presence and severity of DR. DR mainly included non-proliferative and proliferative types. Non-proliferative DR showed one or more of the following symptoms: microaneurysm, hemorrhage, exudates, or microvascular abnormalities; proliferative DR showed the generation of new vessels and fibrosis. Eligible studies were included using the following criteria: (a) the study was observational (case-control, cross-sectional or cohort study) and investigated the relationship between TBL and DR; (b) the study reported either the incidence rate of DR in T2DM or TBL in DR and NDR subjects; (c) the study provided the mean value with standard errors or odds ratio (OR) with 95% confidence intervals (CIs), or the necessary data to calculate these parameters. In addition, if more than two studies investigated the same population, the latest or highest-quality study was selected. We contacted the author if the study did not provide sufficient data. We excluded studies using the following criteria: (a) the study was not original research; (b) the study did not involve human subjects; (c) the study was a duplicate study or the data were duplicated; (d) the study did not report the relationship between TBL and DR; (e) the study did not contain sufficient data to calculate the mean value with standard errors or OR with 95% CIs. The process of study selection and exclusion was carried out by two independent reviewers (Xiaomei, Wu and Bo, Zhu), and any discrepancies were resolved by discussion or consultation with a third reviewer (Kang, Ning). Data extraction and conversion Two authors (Xiaomei, Wu and Bo, Zhu) independently extracted relevant information from the included studies according to the inclusion and exclusion criteria, and calculated the mean value with standard errors or OR with 95%CIs by obtaining data from the study or author. This information included the following: first author’s name, year of publication, study design, number of subjects, gender, age, blood components used for detection, and the corresponding estimate of the relationship between TBL and DR. To reduce the effects of the relevant factors on the results, we extracted and analyzed the adjusted OR to unadjusted OR. If the study had several ORs which were adjusted for different combinations of relevant factors, we extracted the OR which was adjusted for the greatest number of relevant factors. The method used to measure TBL in the included studies was mainly by automatic biochemical analyzer. As the units for TBL were often μmol/L and mg/dL, in order to carry out a statistical analysis, μmol/L was converted to mg/dL by dividing by 17.1. We also determined whether there was a dose-response relationship between the risk of DR and TBL, therefore, the following information was also required: (a) the risk estimates and their 95%CIs for at least three exposure categories; (b) the median or mean TBL in each category. We extracted the exposure category and ORs with their 95%CIs from the studies, which contained the above information. Quality Assessment The included studies in our meta-analysis were independently assessed by two authors (Xiaomei, Wu and Bo, Zhu). Seven items were selected from the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement [20], as follows: Objectives: state specific objectives, including any prespecified hypotheses; Study design: present key elements of study design early in the paper; Setting: describe the setting, locations, and relevant dates, exposure, follow-up, and data collection; Participants: Cohort study—Give the eligibility criteria, and the sources and methods of selection of participants. Describe methods of follow-up. Case-control study—Give the eligibility criteria, and the sources and methods of case ascertainment and control selection. Give the rationale for the choice of cases and controls. Cross-sectional study—Give the eligibility criteria, and the sources and methods of selection of participants; Variables: clearly define all outcomes, exposures, give diagnostic criteria; Statistical methods: Describe all statistical methods; Main results: Report numbers of outcome events or summary measures and category boundaries when continuous variables were categorized. Give unadjusted estimates and, if applicable, confounder-adjusted estimates and their precision (e.g., 95% confidence interval). The high quality studies met 7 items, the low quality studies met less than 5 items, and other studies were moderate quality. Statistical analysis This meta-analysis was conducted using the Stata software package (Version 12.0; Stata Corp., College Station, TX). The standardized mean difference (SMD) and pooled odds ratios (ORs) were used to compare between DR and NDR group. The Chi-square-based Q-test was used to evaluate the heterogeneity among the individual studies. Heterogeneity was quantified based on I2, which ranged from 0% to 100% (I2 = 0% to 25%, no heterogeneity; I2 = 25% to 50%, moderate heterogeneity; I2 = 50% to 75%, large heterogeneity; I2 = 75% to 100%, extreme heterogeneity). When I2 was larger than 50%, a random effects model was used; otherwise, the fixed model was used. We conducted the dose—response meta-analysis to calculate study-specific slopes (i.e., linear trends) and 95%CIs, which proposed by Greenland S. et al and Orsini N. et al [21, 22]. If the study reported exposure category by a range, the midpoint was calculated by averaging the lower and upper bound; if the lowest category was open-ended, the lowest boundary was considered to be zero; if the highest exposure category was open-ended, the width of the open-ended interval was taken to be the same as the adjacent interval. If there was high heterogeneity between studies, we used subgroup analysis on year of publication, study design, number of subjects, gender, age and quality of study to find the source of heterogeneity. We used the sensitivity analysis to evaluate the robustness of the results in our meta-analysis. In the sensitivity analysis, we excluded each study in turn and obtained the pooled estimates from the remaining studies. The purpose of sensitivity analysis was to evaluate the effect of a single study on the overall pooled estimates. The possibility of publication bias was assessed using visual inspection of funnel plots and Egger's test, Egger's test was used to detect the asymmetry test for the funnel plot. A two-sided P value <0.05 in statistical process was considered significantly different. Results Literature search and study characteristics We identified 760 potential studies in six electronic databases (PubMed: 126, Web of Science: 476, CNKI: 70, Wanfang database: 52, CBM: 21, and VIP: 15) using our search strategy. We did not find any new eligible studies during a manual search of the reference lists within these studies. After reading the title and abstract, we excluded 632 studies: 167 duplicate studies, 98 reviews, editorials and commentaries, 11 case reports and 356 animal, chemistry-based and cell-line studies. After reading the full text, we excluded 92 studies, which did not report the relationship between TBL and DR or contained data from the same study population. When we extracted the data, 12 studies did not provide sufficient information. Thus, 24 studies [13–19, 23–39] were finally selected for our meta-analysis. The flow chart of the screening process is shown in Fig 1. 10.1371/journal.pone.0161649.g001Fig 1 The process of study selection in our meta-analysis. Of the 24 studies included, there were 16 case-control studies, 7 cross-sectional studies, and 1 cohort study. In total, 20,913 subjects were enrolled in our meta-analysis. Twenty studies included mean values and their standard errors and9 studies included ORs and their 95%CIs. Six studies provided the exposure category and ORs with their 95%CIs. The characteristics of the included studies are shown in Table 1. 10.1371/journal.pone.0161649.t001Table 1 The characteristics of the included studies in the meta-analysis. First Author Year Design Number of subjects Gender(Male/Female) Age, years(mean± SD) Blood The criterion of DR TBL (mean±SD, mg/dl) Exposure categories (mg/dl) OR (95%CI)b DR NDR OR (95%CI)a Dan, Zhang[23] 2015 Case-control 553 271/282 59.0 ± 11.3 Serum Funduscopic examination 0.71 ± 0.31 0.93 ± 0.31 0.20 (0.10, 0.37)# - - Wei, Wei[15] 2015 Case-control 100 56/44 53.9 ± 9.9 Serum Funduscopic examination 0.88 ± 0.45 1.36 ± 0.68 0.05 (0.00, 0.81) ≤0.65 1.00 0.66–1.23 0.78 (0.50, 1.21) 1.24–1.59 0.69 (0.42, 1.12) ≥1.60 0.20 (0.08, 0.51) Fang, Chen[24] 2015 Case-control 237 134/103 58.7 ± 14.8 - Funduscopic examination 0.59 ± 0.25 0.84 ± 0.43 0.07 (0.01, 0.33)# - - Junwei, Cai[19] 2015 Case-control 118 49/69 54.9 ± 6.0 Serum Funduscopic examination 0.83 ± 0.06 0.83 ± 0.06 - - - Sekioka, Risa[16] 2015 Cross-sectional 674 446/228 64.7 ± 13.9 Serum Funduscopic examination 0.65 ± 0.30 0.80 ± 0.39 0.26 (0.14, 0.49)# - - Hamamoto, S[25] 2015 Cross-sectional 523 312/211 60.5 ± 12.3 - Funduscopic examination 0.53 ± 0.22 0.62 ± 0.23 0.10 (0.03, 0.27)# - - Dave, Apoorva[26] 2015 Case-control 80 46/34 53.9 ± 9.9 Serum Funduscopic examination 0.88 ± 0.45 1.36 ± 0.68 0.05 (0.00, 0.81)# ≤0.65 1.00 0.66–1.23 0.73 (0.45, 1.20) 1.24–1.59 0.61 (0.35, 1.08) ≥1.60 0.17 (0.06, 0.51) Xiaojing, Shang[17] 2014 Case-control 95 52/43 58.4 ± 1.7 Serum Funduscopic examination 0.61 ± 0.36 0.67 ± 0.30 - - - Wei, Peng[27] 2014 Case-control 220 93/127 56.9 Serum Funduscopic examination 0.50 ± 0.17 0.68 ± 0.26 - - - Najam, Syeda Sadia[28] 2014 Cross-sectional 1761 755/1006 61.2 ± 9.3 Serum Funduscopic examination - - - <0.6 1.00 0.6–0.76 1.06 (0.69, 1.64) 0.77–0.99 0.67 (0.42, 1.09) >0.99 0.55 (0.33, 0.91) Kim, Eun Sook[29] 2014 Cross-sectional 1711 808/903 57.1 ± 10.5 Serum n/r - - - Male:<0.70 1.00 0.70–0.90 0.64 (0.50, 0.82) >0.90 0.45 (0.34, 0.60) Female:<0.60 1.00 0.60–0.70 0.49 (0.37, 0.63) >0.70 0.59 (0.46, 0.76) Xiaozhong, Yu[14] 2013 Case-control 373 190/183 59.6 ± 10.0 Serum Funduscopic examination 0.44 ± 0.31 0.63 ± 0.28 - - - Jie, Lai[30] 2013 Case-control 495 249/246 62.9 ± 10.7 Serum Funduscopic examination 0.76 ± 0.26 0.82 ± 0.29 - - - Chan, K. H[31] 2013 Cross-sectional 9795 6045/3750 62.0 ± 7.0 Plasma n/r - - - Male: ≤0.35 1.00 0.41–0.47 0.87 (0.68, 1.12) 0.53–0.58 0.80 (0.62, 1.03) 0.64–0.70 0.81 (0.61, 1.06) ≥0.76 0.90 (0.70, 1.15) Female: ≤0.35 1.00 0.41–0.47 0.86 (0.64, 1.15) 0.53–0.58 1.12 (0.81, 1.56) 0.64–0.70 1.25 (0.82, 1.91) ≥0.76 0.60 (0.36, 1.02) Yajing, Luo[32] 2013 Cross-sectional 246 118/128 62.1 ± 11.1 Serum Funduscopic examination 0.65 ± 0.19 0.76 ± 0.20 - - - Yuan, He[33] 2012 Case-control 708 355/352 56.7 ± 11.4 - Funduscopic examination 0.70 ± 0.35 0.76 ± 0.35 - - - Wei, Feng[34] 2012 Case-control 471 234/237 59.7 ± 10.0 Serum Funduscopic examination 0.82 ± 0.28 0.85 ± 0.30 - - - Wei, Du[35] 2012 Case-control 96 52/44 62.7 ± 10.3 Serum Funduscopic examination 0.48 ± 0.10 0.68 ± 0.14 - - - Hui, Xu[36] 2011 Case-control 86 51/35 63.2 ± 5.8 Serum Funduscopic examination 0.50 ± 0.17 0.68 ± 0.26 - - - Yasuda, Miho[37] 2011 Cohort study 486 - - Serum Funduscopic examination - - - <0.60 1.00 0.60–0.69 1.41 (0.56, 3.54) 0.70–0.89 1.12 (0.41, 3.01) ≥0.90 0.39(0.12, 1.30) Cho, Ho Chan[13] 2011 Cross-sectional 102 54/48 59.6 ± 11.4 Serum Funduscopic examination 0.59 ± 0.24 0.88 ± 0.47 0.17 (0.00, 1.00)# - - Zhiyan, Su[39] 2010 Case-control 664 327/337 59.7 ± 9.73 Serum Funduscopic examination 0.74 ± 0.27 0.90 ± 0.33 0.20 (0.10, 0.37)# - - Yumei, Jia[38] 2010 Case-control 1062 571/491 58.2 ± 12.2 Serum Funduscopic examination 0.60 ± 0.22 0.78 ± 0.23 0.21 (0.10, 0.46)# - - Huang, E. J[18] 2006 Case-control 257 130/127 60.0±11.0 Serum Funduscopic examination 0.73 ± 0.33 0.69 ± 0.37 - - - a, As a continuous variable, the effect of bilirubin on DR; b, As a categorical variable, the effect of bilirubin on DR; TBL: total bilirubin level; n/r = not reported; #, The OR was adjusted. Quality analysis We assessed the quality of the included studies using seven items, which were selected from the STROBE statement [20]. Twenty-one studies were assessed as high quality and three studies were assessed as moderate quality. The detailed results on quality assessment of the included studies are shown in Table 2. 10.1371/journal.pone.0161649.t002Table 2 The quality analysis of the included studies in the meta-analysis. Study Objectives Study design Setting Participants Variables Statistical methods Main results Dan, Zhang(2015)[23] ● ● ● ● ● ● ● Wei, Wei(2015)[15] ● ● ● ● ● ● ● Fang, Chen(2015)[24] ● ● ◎ ● ● ● ● Junwei, Cai(2015)[19] ● ● ● ● ● ● ● Sekioka, Risa(2015)[16] ● ● ● ● ● ● ● Hamamoto, S(2015)[25] ● ● ● ● ● ● ● Dave, Apoorva(2015)[26] ● ● ● ● ● ● ● Xiaojing, Shang(2014)[17] ● ● ● ● ● ● ● Wei, Peng(2014)[27] ● ● ● ● ● ● ● Najam, Syeda Sadia(2014)[28] ● ● ● ● ● ● ● Kim, Eun Sook(2014)[29] ● ● ● ● ● ● ● Xiaozhong, Yu(2013)[14] ● ● ● ● ● ● ● Jie, Lai(2013)[30] ● ● ● ● ● ● ● Chan, K. H(2013)[31] ● ● ● ● ● ● ◎ Yajing, Luo(2013)[32] ● ● ● ● ● ● ● Yuan, He(2012)[33] ● ● ◎ ● ● ● ● Wei, Feng(2012)[34] ● ● ● ● ● ● ● Wei, Du(2012)[35] ● ● ● ● ● ● ● Hui, Xu(2011)[36] ● ● ● ● ● ● ● Yasuda, Miho(2011)[37] ● ● ● ● ● ● ● Cho, Ho Chan(2011)[13] ● ● ● ● ● ● ● Zhiyan, Su(2010)[39] ● ● ● ● ● ● ● Yumei, Jia(2010)[38] ● ● ● ● ● ● ● Huang, E. J(2006)[18] ● ● ● ● ● ● ● ●, the item was described in detail; ◎, the item was described partly. Overall and subgroup analysis Twenty of 24 studies [13–19, 23–27, 30, 32–36, 38, 39] had calculated the pooled SMD, and the results showed that TBL in the DR group was lower than that in the NDR group (SMD: –0.52, 95% CI: –0.67, –0.38), see Fig 2. As there was obvious heterogeneity (I2 = 86.8%), we performed a subgroup analysis, and the results showed that TBL in the DR group was lower than that in the NDR group. 10.1371/journal.pone.0161649.g002Fig 2 The SMD on the relationship between TBL and the risk of DR. Nine of 24 studies [13, 15, 16, 23–26, 38, 39] had calculated the pooled ORs, and the results showed that there was a significant negative relationship between TBL and the risk of DR (OR: 0.19, 95% CI: 0.14, 0.25), see Fig 3. We found that the ORs in 8 studies were adjusted, the OR in only one study [15] was not adjusted. The pooled adjusted OR also showed that there was a significant negative relationship between TBL and the risk of DR (OR: 0.16, 95% CI: 0.10, 0.22; I2 = 0). We also performed a subgroup analysis, and the results showed that there was a significant negative relationship between TBL and the risk of DR. 10.1371/journal.pone.0161649.g003Fig 3 The pooled ORs on the relationship between TBL and the risk of DR. The detailed results of the overall and subgroup analyses on the relationship between TBL and the risk of DR are shown in Table 3. 10.1371/journal.pone.0161649.t003Table 3 The SMD and pooled ORs on the relationship between TBL and the risk of DR. TBL in DR and NDR TBL and risk of DR No. of study SMD (95%CI) I2 (%) P for heterogeneity P for Egger’s test No. of study OR (95%CI) I2 (%) P for heterogeneity P for Egger’s test Overall 20 -0.52 (-0.67, -0.38) 86.8 0.000 0.547 9 0.19 (0.14,0.25) 0.0 0.608 0.017 Year of publication after 2013 12 -0.52 (-0.65, -0.38) 72.3 0.000 6 0.18 (0.12,0.26) 18.8 0.291 before 2012 8 -0.56 (-0.85, -0.27) 93.3 0.000 3 0.20 (0.12,0.33) 0.0 0.982 Study design case-control 16 -0.53 (-0.70, -0.35) 89.2 0.000 6 0.18 (0.12,0.26) 0.0 0.613 cross-sectional 4 -0.46 (-0.56, -0.36) 35.7 0.198 3 0.20 (0.12,0.34) 24.9 0.264 Number of subjects < 250 10 -0.70 (-0.94, -0.46) 76.5 0.000 5 0.20 (0.15,0.27) 0.0 0.614 ≥ 250 10 -0.39 (-0.57, -0.22) 90.4 0.000 4 0.07 (0.02,0.21) 0.0 0.934 Gender male/female ≥ 1 13 -0.55 (-0.75, -0.36) 88.6 0.000 7 0.18 (0.12,0.26) 4.1 0.395 male/female < 1 7 -0.49 (-0.70, -0.28) 84.5 0.000 2 0.20 (0.13,0.31) 0.0 1.000 Age < 60 13 -0.54 (-0.71, -0.37) 85.9 0.000 7 0.18 (0.12,0.26) 0.0 0.735 ≥ 60 7 -0.50 (-0.76, -0.25) 88.0 0.000 2 0.20 (0.12,0.34) 62.2 0.104 Quality of study high 18 -0.54 (-0.69, -0.39) 86.6 0.000 8 0.19 (0.14,0.26) 0.0 0.695 moderate 2 -0.40 (-0.88, -0.08) 88.8 0.003 1 0.07 (0.01,0.33) - - TBL: total bilirubin level. Sensitivity analysis We perform a sensitivity analysis to assess the stability and reliability of the results. After removing each single study from the pooled analysis, both the SMD and pooled ORs were unaffected (S1 and S2 Tables). Dose-response relationship between TBL and the risk of DR Six of 24 [15, 26, 28, 29, 31, 37] included information on dose-response which reflected the relationship between TBL and the risk of DR. We found a nonlinear relationship between TBL and the risk of DR (P = 0.017), see Fig 4. 10.1371/journal.pone.0161649.g004Fig 4 Dose—response analysis between TBL and the risk of DR. Publication bias As shown in Table 3, visual inspection of the Begg’s funnel plot or Egger’s test suggested no publication bias in SMD (S1 Fig), but publication bias was observed in the pooled ORs (S2 Fig). We performed the trim and fill method, and the results also showed a significant association between TBL and the risk of DR (OR: 1.76, 95% CI: 1.60–1.94). Discussion DR is a common microvascular complication in T2DM patients. The pathogenesis of DR is not yet fully understood and there is no effective cure [40, 41]. Early detection and treatment can significantly delay the development of DR and preserve vision in most patients. The prevention of diabetic blindness has become an important public health problem worldwide [42]. In recent years, research on the relationship between bilirubin and DR has gradually increased, but the results are inconsistent. Therefore, we performed a meta-analysis to investigate the relationship between TBL and DR. This meta-analysis not only compared TBL between a DR and NDR group, but investigated the dose-response relationship between TBL and the risk of DR. Briefly, we found that TBL in the DR group was lower than that in the NDR group, and there was a significant negative relationship between TBL and the risk of DR. The dose-response trend suggested that increased TBL may delay DR progression. Heterogeneity had an impact on the combined effect of the included studies. First, we used strict inclusion and exclusion criteria in the literature selection process. Second, we assessed the quality of included studies by selecting seven items from the STROBE statement, which provided a checklist of items on how to report observational research well. There were 21 high quality and three moderate quality studies. Third, because there were other potential confounding factors (year of publication, study design, number of subjects, gender, age), we carried out a subgroup analysis to determine the source of heterogeneity. In each subgroup analysis, we also found that TBL in the DR group was lower than that in the NDR group, and there was a significant negative relationship between TBL and the risk of DR. Fourth, we carried out a sensitivity analysis to assess the reliability of our meta-analysis. The results showed that no study influenced the combined effect. Therefore, our results were reliable. Although all traditional risk factors (hyperglycemia, dyslipidosis, hypertension and duration of diabetes) are associated with the development and progression of DR [43], the change in serum redox status in T2DM patients is an additional risk factor. Research has found that endothelial dysfunction and hyperglycemia are associated with lipid peroxidation [44, 45]. To date, the underlying protective mechanism of bilirubin on DR is not yet fully understood, and a hypothesis has been proposed. Bilirubin is recognized as an important endogenous antioxidant, and has strong antioxidant and anti-inflammatory effects on the microvasculature [7, 13, 46]. A series of biochemical changes take place in the retinal microvasculature, including oxidative stress, the polyol pathway, protein kinase C activation, and advanced glycation end product formation, which are mainly due to long-term high blood glucose [47]. Oxidative stress and inflammation play important roles in the pathogenesis of DR [47, 48]. Biochemical reactions induced by oxidative stress change the function and structure of the retinal microvasculature. These functional changes include basement membrane thickening, microvascular cell loss, capillary closure, and cellular capillary formation. The structural changes include altered blood flow, loss of intercellular junctions, and increased vessel permeability [49]. The results of an animal study indicated that inhibition of inflammation may inhibit the progression of early stage DR [47]. Thus, high TBL could inhibit oxidative stress and inflammation processes and delay or interrupt development of DR. To reduce the effects of the relevant factors on the results, we extracted and analyzed the adjusted OR to unadjusted OR. The ORs were adjusted to take into account common confounders (such as gender, age, BMI and biochemical indicators), this ensured the stability of the results of our meta-analysis. We found that the ORs in 8 studies were adjusted and the OR in only one study was not adjusted. We also carried out a subgroup analysis and the results were stable. Given the effect of liver disease on TBL, 22 of 24 studies excluded patients with liver disease, and only two studies did not clearly indicate whether patients with liver disease were excluded. These two studies were of high quality as evaluated by quality assessment, and we considered the results of these studies credible. During the sensitivity analysis, we excluded these two studies and the results were also stable. There are some limitations in our meta-analysis that should be taken into account. First, the combined effects were based on observational studies (SMD: 16 cross-sectional studies and four case-control studies; OR: six cross-sectional studiesand three case-control studies), and this may make the results unreliable. Therefore, we carried out a subgroup analysis and found that the combined effects in the cross-sectional and case—control studies were consistent with the overall results. Second, DR is a multi-factorial disease and influenced by both genetic and environmental risk factors, and most of the included studies were not adjusted by confounders, especially duration of diabetes. The risk of DR increased with longer duration of diabetes, and we were unable to extract the data from the included studies in our meta-analysis. Third, we found that only one cohort study analyzed the relationship between bilirubin level and the risk of DR, the other studies were cross-sectional and case-control studies. The cohort study did not find a significant association between bilirubin level and the risk of DR. Toyoshi Inoguchi et al [50] first reported a lower prevalence of DR in patients with diabetes and Gilbert syndrome, but this was also a cross-sectional study. Therefore, long-term cohort studies with large populations are needed to confirm the relationship between bilirubin level and the risk of DR. In summary, the results of our meta-analysis indicate that higher TBL may be protective against DR in subjects with diabetes, and TBL could be used as a biomarker to predict the risk of DR. In addition, due to the limitations in this meta-analysis, long-term, large-scale studies are needed to confirm our results. Supporting Information S1 Fig The funnel plot of the 20 studies on SMD. (TIF) Click here for additional data file. S2 Fig The funnel plot of the 20 studies on the pooled ORs. (TIF) Click here for additional data file. S1 PRISMA Checklist (DOC) Click here for additional data file. S1 Table Sensitivity analysis on SMD by removing each study in each model. (DOCX) Click here for additional data file. S2 Table Sensitivity analysis on the pooled ORs by removing each study in each model. (DOCX) Click here for additional data file. ==== Refs References 1 Zhu B , Wu X , Wang X , Zheng Q , Sun G . The association between passive smoking and type 2 diabetes: a meta-analysis . Asia-Pacific journal of public health / Asia-Pacific Academic Consortium for Public Health . 2014 ;26 (3 ):226 –37 . Epub 2014/05/16. 10.1177/1010539514531041 .24824522 2 Klein BE . Overview of epidemiologic studies of diabetic retinopathy . Ophthalmic epidemiology . 2007 ;14 (4 ):179 –83 . Epub 2007/09/27. 10.1080/09286580701396720 .17896294 3 van Dieren S , Beulens JW , van der Schouw YT , Grobbee DE , Neal B . The global burden of diabetes and its complications: an emerging pandemic . European journal of cardiovascular prevention and rehabilitation: official journal of the European Society of Cardiology, Working Groups on Epidemiology & Prevention and Cardiac Rehabilitation and Exercise Physiology . 2010 ;17 Suppl 1 :S3 –8 . Epub 2010/05/28. 10.1097/01.hjr.0000368191.86614.5a .20489418 4 Zhang X , Saaddine JB , Chou CF , Cotch MF , Cheng YJ , Geiss LS , et al Prevalence of diabetic retinopathy in the United States, 2005–2008 . Jama . 2010 ;304 (6 ):649 –56 . Epub 2010/08/12. 10.1001/jama.2010.1111 20699456 5 Pan HZ , Zhang H , Chang D , Li H , Sui H . The change of oxidative stress products in diabetes mellitus and diabetic retinopathy . The British journal of ophthalmology . 2008 ;92 (4 ):548 –51 . Epub 2008/03/29. 10.1136/bjo.2007.130542 .18369071 6 Ram M , Singh V , Kumar D , Kumawat S , Gopalakrishnan A , Lingaraju MC , et al Antioxidant potential of bilirubin-accelerated wound healing in streptozotocin-induced diabetic rats . Naunyn-Schmiedeberg's archives of pharmacology . 2014 ;387 (10 ):955 –61 . Epub 2014/06/28. 10.1007/s00210-014-1011-3 .24969350 7 Dekker D , Dorresteijn MJ , Pijnenburg M , Heemskerk S , Rasing-Hoogveld A , Burger DM , et al The bilirubin-increasing drug atazanavir improves endothelial function in patients with type 2 diabetes mellitus . Arteriosclerosis, thrombosis, and vascular biology . 2011 ;31 (2 ):458 –63 . Epub 2010/11/20. 10.1161/atvbaha.110.211789 .21088253 8 Djousse L , Levy D , Cupples LA , Evans JC , D'Agostino RB , Ellison RC . Total serum bilirubin and risk of cardiovascular disease in the Framingham offspring study . The American journal of cardiology . 2001 ;87 (10 ):1196 –200 ; A4 , 7 Epub 2001/05/18. .11356398 9 Troughton JA , Woodside JV , Young IS , Arveiler D , Amouyel P , Ferrieres J , et al Bilirubin and coronary heart disease risk in the Prospective Epidemiological Study of Myocardial Infarction (PRIME) . European journal of cardiovascular prevention and rehabilitation: official journal of the European Society of Cardiology, Working Groups on Epidemiology & Prevention and Cardiac Rehabilitation and Exercise Physiology . 2007 ;14 (1 ):79 –84 . Epub 2007/02/16. 10.1097/01.hjr.0000230097.81202.9f .17301631 10 Novotny L , Vitek L . Inverse relationship between serum bilirubin and atherosclerosis in men: a meta-analysis of published studies . Experimental biology and medicine (Maywood, NJ) . 2003 ;228 (5 ):568 –71 . Epub 2003/04/24. .12709588 11 Hosono S , Ohno T , Kimoto H , Shimizu M , Nozawa M , Genkawa R , et al No clinical correlation between bilirubin levels and severity of retinopathy of prematurity . Journal of pediatric ophthalmology and strabismus . 2002 ;39 (3 ):151 –6 . Epub 2002/06/08. .12051280 12 DeJonge MH , Khuntia A , Maisels MJ , Bandagi A . Bilirubin levels and severe retinopathy of prematurity in infants with estimated gestational ages of 23 to 26 weeks . The Journal of pediatrics . 1999 ;135 (1 ):102 –4 . Epub 1999/07/07. .10393613 13 Cho HC . The Relationship among Homocysteine, Bilirubin, and Diabetic Retinopathy . Diabetes & metabolism journal . 2011 ;35 (6 ):595 –601 . Epub 2012/01/17. 10.4093/dmj.2011.35.6.595 22247902 14 Yu X , Li J , Liu Z . The relationship between serum bilirubin level and type 2 diabetic retinopathy . Journal of Guiyang College of Traditional Chinese Medicine . 2013 ;(06 ):171 –2 . 15 Wei W , Lu Q . The relation of Bilirubin and malondialdehyde with diabetic retinopathy in Type 2 Diabetes Mellitus . Journal Of Inner Mongolia Medical University . 2015 ;(05 ):448 –52 . 16 Sekioka R , Tanaka M , Nishimura T , Itoh H . Serum total bilirubin concentration is negatively associated with increasing severity of retinopathy in patients with type 2 diabetes mellitus . Journal of Diabetes and Its Complications . 2015 ;29 (2 ):218 –21 . 10.1016/j.jdiacomp.2014.12.002 . WOS:000350531600011. 25536865 17 Shang X , Song C , Shao H , Xu D . Association of the homocysteine and bilirubin with the diabetic retinopathy . Chinese Journal of Prevention and Control of Chronic Diseases ,. 2014 ;(05 ):523 –5 . 18 Huang EJ , Kuo WW , Chen YJ , Chen TH , Chang MH , Lu MC , et al Homocysteine and other biochemical parameters in Type 2 diabetes mellitus with different diabetic duration or diabetic retinopathy . Clinica Chimica Acta . 2006 ;366 (1–2 ):293 –8 . 10.1016/j.cca.2005.10.025 . WOS:000236378200037. 19 Cai J , Li X , Xu Y , Zeng Y . Study on the correlation between serum bilirubin, high sensitive C reactive protein and diabetic retinopathy . Journal of Clinical Internal Medicine . 2015 ;(7 ):490 –1 . 20 von Elm E , Altman DG , Egger M , Pocock SJ , Gotzsche PC , Vandenbroucke JP . The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies . PLoS medicine . 2007 ;4 (10 ):e296 Epub 2007/10/19. 10.1371/journal.pmed.0040296 17941714 21 Greenland S , Longnecker MP . Methods for trend estimation from summarized dose-response data, with applications to meta-analysis . American journal of epidemiology . 1992 ;135 (11 ):1301 –9 . Epub 1992/06/01. .1626547 22 Orsini N , Li R , Wolk A , Khudyakov P , Spiegelman D . Meta-analysis for linear and nonlinear dose-response relations: examples, an evaluation of approximations, and software . American journal of epidemiology . 2012 ;175 (1 ):66 –73 . Epub 2011/12/03. 10.1093/aje/kwr265 22135359 23 Zhang D , Liu X , Liu B . Role of total bilirubin level in early diagnosis of diabetic retinopathy . Chin J Diabetes . 2015 ;23 (11 ):995 –7 . 24 Chen F , Huang W , Yan H , Ou X , Li M . Relationship between bilirubin and diabetic retinopathy . Chin J Clinicians(Electronic Edition) . 2015 ;(14 ):2665 –8 . 25 Hamamoto S , Kaneto H , Kamei S , Shimoda M , Tawaramoto K , Kanda-Kimura Y , et al Low bilirubin levels are an independent risk factor for diabetic retinopathy and nephropathy in Japanese patients with type 2 diabetes . Diabetes & metabolism . 2015 ;41 (5 ):429 –31 . 10.1016/j.diabet.2015.05.003 . MEDLINE:26054296.26054296 26 Dave A , Kalra P , Gowda BHR , Krishnaswamy M . Association of bilirubin and malondialdehyde levels with retinopathy in type 2 diabetes mellitus . Indian journal of endocrinology and metabolism . 2015 ;19 (3 ):373 –7 . 10.4103/2230-8210.152777 . MEDLINE:25932393. 25932393 27 Peng W , Wang A , Sun J , Mao Y , Liu G . The relationship between diabetic retinopathy and serum bilirubin level, blood lipid . China health care & nutrition . 2014 ;24 (4 ):1927 –8 . 28 Najam SS , Sun J , Zhang J , Xu M , Lu J , Sun K , et al Serum total bilirubin levels and prevalence of diabetic retinopathy in a Chinese population . Journal of Diabetes . 2014 ;6 (3 ):221 –7 . 10.1111/1753-0407.12085 . WOS:000334441600005. 24034422 29 Kim ES , Mo EY , Moon SD , Han JH . Inverse Association between Serum Bilirubin Levels and Arterial Stiffness in Korean Women with Type 2 Diabetes . PloS one . 2014 ;9 (10 ). 10.1371/journal.pone.0109251 . WOS:000343941200045. 30 Lai J . Study on the relationship between diabetic retinopathy and serum bilirubin . Chinese Journal of Postgraduates of Medicine . 2013 ;36 (1 ):55 –7 . 31 Chan KH , O'Connell RL , Sullivan DR , Hoffmann LS , Rajamani K , Whiting M , et al Plasma total bilirubin levels predict amputation events in type 2 diabetes mellitus: the Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) study . Diabetologia . 2013 ;56 (4 ):724 –36 . 10.1007/s00125-012-2818-4 . WOS:000316076300006. 23322233 32 Luo Y . Reiationship between serum bilirnbin in patients with type 2 diabetes and diabetes-related chronic complications [Master] : Dalian Medical University ; 2013 . 33 He Y , Yu D , Zhou S , Chen R , Lv L , Ma Z , et al The relative risk factors of diabetic retinopathy . Chin J Diabetes Mellitus . 2012 ;4 (10 ):601 –6 . 34 Feng W , Yang J . Relationship between serum bilirubin and diabetic retinopathy in patients with type 2 diabetes . Journal of Capital Medical University . 2012 ;(04 ):66 –9 . 35 Du W , Liu Z , Zhou Y , Luo M . Relationship between diabetic retinopathy and serum bilirubin level . Recent Advances in Ophthalmology . 2012 ;(05 ):484 –5 . 36 Xu H , Sun J , Xu D , Liu J . The study on the relationship between diabetic retinopathy and serum bilirubin level . Chinese journal of postgraduates of medline . 2011 ;34 (10 ):53 –4 . 37 Yasuda M , Kiyohara Y , Wang JJ , Arakawa S , Yonemoto K , Doi Y , et al High Serum Bilirubin Levels and Diabetic Retinopathy The Hisayama Study . Ophthalmology . 2011 ;118 (7 ):1423 –8 . 10.1016/j.ophtha.2010.12.009 . WOS:000292303000028. 21600659 38 Jia Y , Xu Y , Yao Z , Zhao Y . The study on the relationship between the severity of diabetic retinopathy and serum bilirubin . Chinese Journal of Postgraduates of Medicine . 2010 ;33 (34 ):9 –12 . 39 Su Z , Yang J , Zhang L . The relationship between diabetic retinopathy and serum bilirubin level . Journal of Capital Medical University . 2010 ;(01 ):117 –9 . 40 Zhu B , Wu X , Zhi X , Liu L , Zheng Q , Sun G . Methylenetetrahydrofolate reductase C677T polymorphism and type 2 diabetes mellitus in Chinese population: a meta-analysis of 29 case-control studies . PloS one . 2014 ;9 (7 ):e102443 Epub 2014/07/23. 10.1371/journal.pone.0102443 25047451 41 Rangasamy S , McGuire PG , Das A . Diabetic retinopathy and inflammation: novel therapeutic targets . Middle East African journal of ophthalmology . 2012 ;19 (1 ):52 –9 . Epub 2012/02/22. 10.4103/0974-9233.92116 22346115 42 Heng LZ , Comyn O , Peto T , Tadros C , Ng E , Sivaprasad S , et al Diabetic retinopathy: pathogenesis, clinical grading, management and future developments . Diabetic medicine: a journal of the British Diabetic Association . 2013 ;30 (6 ):640 –50 . Epub 2012/12/05. 10.1111/dme.12089 .23205608 43 Nguyen TT , Alibrahim E , Islam FM , Klein R , Klein BE , Cotch MF , et al Inflammatory, hemostatic, and other novel biomarkers for diabetic retinopathy: the multi-ethnic study of atherosclerosis . Diabetes care . 2009 ;32 (9 ):1704 –9 . Epub 2009/06/25. 10.2337/dc09-0102 19549733 44 Su Y , Liu XM , Sun YM , Jin HB , Fu R , Wang YY , et al The relationship between endothelial dysfunction and oxidative stress in diabetes and prediabetes . International journal of clinical practice . 2008 ;62 (6 ):877 –82 . Epub 2008/05/16. 10.1111/j.1742-1241.2008.01776.x .18479281 45 Wu Y , Tang L , Chen B . Oxidative stress: implications for the development of diabetic retinopathy and antioxidant therapeutic perspectives . Oxidative medicine and cellular longevity . 2014 ;2014 :752387 Epub 2014/09/03. 10.1155/2014/752387 25180070 46 Stocker R , Yamamoto Y , McDonagh AF , Glazer AN , Ames BN . Bilirubin is an antioxidant of possible physiological importance . Science (New York, NY) . 1987 ;235 (4792 ):1043 –6 . Epub 1987/02/27. .3029864 47 Madsen-Bouterse SA , Kowluru RA . Oxidative stress and diabetic retinopathy: pathophysiological mechanisms and treatment perspectives . Reviews in endocrine & metabolic disorders . 2008 ;9 (4 ):315 –27 . Epub 2008/07/26. 10.1007/s11154-008-9090-4 .18654858 48 Kern TS . Contributions of inflammatory processes to the development of the early stages of diabetic retinopathy . Experimental diabetes research . 2007 ;2007 :95103 Epub 2008/02/16. 10.1155/2007/95103 18274606 49 Frank RN . Diabetic retinopathy . The New England journal of medicine . 2004 ;350 (1 ):48 –58 . Epub 2004/01/02. 10.1056/NEJMra021678 .14702427 50 Inoguchi T , Sasaki S , Kobayashi K , Takayanagi R , Yamada T . Relationship between Gilbert syndrome and prevalence of vascular complications in patients with diabetes . Jama . 2007 ;298 (12 ):1398 –400 . Epub 2007/09/27. 10.1001/jama.298.12.1398-b .17895455

PMC005xxxxxx/PMC5003344.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757151310.1371/journal.pone.0161869PONE-D-16-23252Research ArticleBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesArtificial Gene Amplification and ExtensionPolymerase Chain ReactionResearch and Analysis MethodsMolecular Biology TechniquesArtificial Gene Amplification and ExtensionPolymerase Chain ReactionMedicine and Health SciencesParasitic DiseasesBiology and Life SciencesOrganismsAnimalsVertebratesAmniotesBirdsFowlGamefowlChickensBiology and Life SciencesOrganismsAnimalsVertebratesAmniotesBirdsPoultryChickensBiology and Life SciencesAgricultureLivestockPoultryChickensBiology and Life SciencesCell BiologyCellular TypesAnimal CellsGerm CellsGametocytesMedicine and Health SciencesParasitic DiseasesProtozoan InfectionsBiology and Life SciencesAnatomyBody FluidsBloodMedicine and Health SciencesAnatomyBody FluidsBloodBiology and Life SciencesPhysiologyBody FluidsBloodMedicine and Health SciencesPhysiologyBody FluidsBloodMedicine and Health SciencesHematologyBloodBiology and Life SciencesOrganismsProtozoansParasitic ProtozoansMalarial ParasitesMedicine and Health SciencesPharmacologyDrugsPrimaquineMonitoring the Prevalence of Leucocytozoon sabrazesi in Southern China and Testing Tricyclic Compounds against Gametocytes L. sabrazesi Transmission DynamicsZhao Wenting 1‡Pang Qin 1‡Xu Ruixue 1Liu Jianwen 1Liu Shengfa 1Li Jian 1*Su Xin-zhuan 12*1 State Key Laboratory of Cellular Stress Biology, Innovation Center for Cell Signaling Network, School of Life Sciences, Xiamen University, Xiamen, Fujian 361005, P. R. China2 Laboratory of Malaria and Vector Research, National Institute of Allergy and Infectious Diseases, National Institutes of Health (NIH), Bethesda, Maryland 20892, United States of AmericaCulleton Richard EditorInstitute of Tropical Medicine, JAPANCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: J. Li X-zS. Data curation: J. Li X-zS. Formal analysis: WZ J. Li X-zS. Funding acquisition: J. Li SL X-zS. Investigation: WZ QP RX J. Liu. Methodology: WZ J. Li. Project administration: J. Li SL X-zS. Supervision: J. Li SL X-zS. Writing – original draft: WZ J. Li X-zS. Writing – review & editing: J. Li X-zS. ‡ These authors are co-first authors on this work. * E-mail: xsu@niaid.nih.gov (X-zS); jianli_204@xmu.edu.cn (JL)29 8 2016 2016 11 8 e01618699 6 2016 13 8 2016 This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.Leucocytozoon parasites infect many species of avian hosts, including domestic chicken, and can inflict heavy economic loss on the poultry industry. Two major species of Leucocytozoon parasites have been reported in China, L. sabrazesi and L. caulleryi, although L. sabrazesi appears to be more widespread than L. caulleryi in southern China. The traditional method for detecting Leucocytozoon infection is microscopic examination of blood smears for the presence of mature gametocytes in circulation, which may miss infections with low parasitemia (gametocytemia) or immature gametocytes. Here we developed a PCR-based method to monitor L. sabrazesi infections at seven sites in four provinces of China after testing two PCR primer pairs based on parasite mitochondrial cytochrome b (cytb) and cytochrome c oxidase III (coxIII) genes. We compared the results of PCR detection with those of microscopic observation. As expected, the PCR assays were more sensitive than microscope examination in detecting L. sabrazesi infection and were able to detect parasite DNA after gametocytes disappeared in the blood stream. Using these methods, we investigated monthly dynamics of L. sabrazesi in chickens from a free-range farm in Xiamen, Fujian province of China, over one year. Our results showed that chickens were infected with L. sabrazesi year-round in southern China. Finally, we tested several compounds for potential treatment of Leucocytozoon infections, including primaquine, ketotifen, clomipramine hydrochloride, desipramine hydrochloride, sulfaquinoxaline, and pyrimethamine. Only primaquine had activity against L. sabrazesi gametocytes. Our results provide important information for controlling parasite transmission in southern China and disease management. http://dx.doi.org/10.13039/501100001809National Natural Science Foundation of China81220108019, 81271858, 81572017Li Jian Project 111 of the State Bureau of Foreign Experts and Ministry of Education of ChinaB06016Su Xin-zhuan http://dx.doi.org/10.13039/100000002National Institutes of HealthIntramural FundingSu Xin-zhuan This work was supported by the National Natural Science Foundation of China #81220108019, #81271858, and #81572017, by Fundamental Research Funds for the Central Universities (20720160057), by Project 111 of the State Bureau of Foreign Experts and Ministry of Education of China (B06016), and by the Division of Intramural Research, National Institute of Allergy and Infectious Diseases, National Institutes of Health. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Leucocytozoon parasites infect numerous species of avian hosts, including domestic chickens [1–7]. The parasites belong to a genus of parasitic protozoa in the phylum Apicomplexa that also contains parasites causing human malaria and Toxoplasmosis. The Leucocytozoon parasites have a complex life cycle involving two hosts, having merogony in fixed tissues (such as liver) and sexual stages (gametocytes) in blood cells of vertebrate hosts, and sporogony in simuliid flies (Simuliidae) or culicoides midges [1]. Infection with Leucocytozoon parasites could result in increased chicken mortality and decreased egg production, with symptoms of listlessness, green feces, anorexia, anemia, and even death [1, 8–10]. Detrimental effects on host reproductive success, host fitness, and economic losses by infections of Haemoproteus and/or Leucocytoazoon parasites have been reported [11–13]. Infected domestic chickens could have mortality rates as high as 56%, leading to significant economic loss [14]. However, it was also reported that Leucocytozoon simondi did not affect the growth rates of infected ducklings [15]. Treatment of Leucocytozoon infection is still difficult, although various antimalarial and anticoccidial compounds have been reported to have effects on bird blood protozoans such as Leucocytozoon parasites, including clopidol [16], primaquine [11, 15], and halofuginonepolystyrene sulfonate [17]. Additionally, sulfamonomethoxine and ormetoprim combinations were reported to prevent or reduce Leucocytozoon infection [18]. However, paludrine, atebrin, and sulphamerazine had no activities against L. simondi in ducklings [19]. Development of additional safe, effective, and affordable compounds for treating Leucocytozoon gametocytes to prevent transmission is necessary to control the disease. Previous surveys of Leucocytozoon infections in China were mostly based on species identification of the circulating gametocytes in blood smears [14, 20–23], although a polymerase chain reaction (PCR) method based on ribosomal DNA ITS2 sequence to detect L. caulleryi was described previously [24]. These studies have provided important information for the understanding of the disease prevalence and for disease control; however, the fact that most of L. sabrazesi infections have low parasitemia (Because only gametocytes are detected in the blood, parasitemia is equal to gametocytemia) can lead to false negatives in diagnosis based on microscopic examination. Detection of parasite DNA or RNA using PCR may inform potential infections even when blood smears are negative, and can improve the sensitivity of detecting parasites in a host. Indeed, there have been reports of detecting infections of Leucocytozoon and other blood protozoan parasites in avian and insect hosts using PCR methods employing primers derived from parasite genes such as the cytochrome b (cytb) gene [25–31]. For example, using primers from a segment of the cytb gene, Ortego and Cordero (2009) amplified DNA from blood samples of 206 nestling eagle owls (Bubo bubo) in the Toledo province of Central Spain and revealed a unique lineage of Leucocytozoon ziemanni in the eagle owls [26]. The cytb gene of Leucocytozoon lovati was also employed to study DNA from 490 insect samples of six black fly species [25]. Similarly, by amplifying a 478 bp fragment from the cytb gene, Ferraguti et al. (2013) identified six Haemoproteus and two Plasmodium lineages of blood parasites in 13 biting midges (out of 97 tested) and in 123 bird blood samples [28]. These studies suggest that PCR amplification of parasite mitochondrial DNA is a reliable approach for detecting blood protozoan parasites, including Leucocytozoon spp. In our previous study, we showed that domestic chickens in southern China were often infected with multiple strains of L. sabrazesi and that gametocytes could “relapse” or be persistent at low-level for 4–5 months without active transmission [23], suggesting potential year-round transmission if insect vectors are available. More recently, we showed that the gametocytes of L. sabrazesi infected chicken thrombocytes [32], but not other blood cells. Erythroblasts, red blood cells (RBCs), and mononuclear leukocytes have historically been reported to be the host cells of Leucocytozoon gametocytes [33–36]. To better understand the transmission dynamics in field settings in southern China, we investigated year-round L. sabrazesi infection found at a free-range farm by collecting chicken blood samples monthly from October 2014 to September 2015. We also designed and evaluated two pairs of PCR primers based on the parasite mitochondrial cytb and cytochrome c oxidase III (coxIII) genes and compared the results from the PCR methods with those of microscopic examination. We found that negative blood smear samples were often positive by PCR methods. Our results also showed the presence of parasites and/or parasite DNA in chicken blood samples collected year-round, with the highest prevalence in June. Finally, we tested several compounds, including some that were shown to block transmission of Plasmodium parasites [37], for potential treatment of Leucocytozoon infections. These results provide important information for better disease control and management. Materials and Methods Ethics Statement Chickens were purchased from the farm owners or with permission from the owners before blood drawing. The experiments were conducted under a protocol (protocol #XMULAC20130065) approved by the Animal Care and Ethics Committee at Xiamen University. Sampling sites and blood collection Chicken blood samples were collected from three sites in Fujian province (Haicang, geographic coordinates 24.4845° N, 118.0328° E; Zhangzhou, 24.5130° N, 117.6471°E; Fuzhou; 26.0745° N, 119.2965° E), two sites in Jiangxi province (Ganzhou, 25.8318° N, 114.9350° E; Ruijin, 25.8856° N, 116.0271° E), one site in Sichuan province (Dazhou, 31.2096° N, 107.4680° E), and one site in Shandong province (Binzhou, 37.3820° N, 117.9707° E) of China as previously described [23]. The chickens (Gallus gallus domesticus) were raised either in free-range farms or in the backyards of individual households. Two to three hundred microliters of blood was collected from the wing veins and was immediately mixed with anticoagulant (8 g/L citric acid and 22 g/L trisodium citrate, pH 7.2) in a 1.5 ml centrifuge tube. Thin blood smears were made, air-dried, fixed with 100% methanol, and stained with Giemsa stain. Parasitemia (or gametocytemia in this case) were estimated as the proportion of infected cells after counting 10,000 red blood cells (RBCs). DNA extraction Blood cells were lysed using saponin solution before DNA extraction. Briefly, 10–20 μl packed blood cells were suspended in 20 volumes of 0.1% saponin in PBS (pH 7.4) and kept on ice for 5 min. The parasites and nuclei of host cells were pelleted by centrifugation at 5000 rpm for 3 min and then washed with PBS (3X). The pellets were re-suspended in 400 μl lysis buffer (50 mM TrisCl, 100 mM EDTA, 2% SDS; pH 8.0, 100 μg/ml proteinase K), and incubated overnight at 50°C. DNA was extracted using the phenol-chloroform method and dissolved in ddH2O. DNA amplification and PCR assays Two pairs of primers were designed based on the mitochondrial sequence encoding cytb and coxIII genes: Ls-cytbF1(5’taa tca cat ggg ttt gtg ga3’) and Ls-cytbR1(5’gct ttg ggc taa gaa taa tac c3’), and Ls-coxIIIF2(5’taa cat tct aca tga tgt agt3’) and Ls-coxIIIR2(5’gta aaa gca cac tta tct ag3’) from cytb and coxIII gene, respectively. Oligonucleotide primers were designed to specifically amplify cytb and coxIII genes based on the mitochondrial DNA sequence of L. sabrazesi (NCBI accession No. AB299369.1). The L. sabrazesi mitochondrial cytb and coxIII gene sequences were aligned with those from representative avian blood parasites Leucotytozoon caulleryi (Accession No. AB302215.1), Haemoproteus columbae (NCBI accession No. FJ168562.1), Plasmodium gallinaceum (Accession No. AB250690.1), and the sequences of chicken (Gallus gallus domesticus) mitochondrial DNA (Accession No. KM096864.1). For each primer, there was a minimum of 3 base pair mismatches, including the 3’end nucleotide, to ensure amplification specificities (Fig 1A and 1B). The primers were synthesized by a commercial company (Sangon Biotech Co., Ltd., Shanghai). All PCR amplifications were performed in a 20 μl volume containing 4 μl of genomic DNA (~50 pg), 80 mM dNTPs, 0.8 mM MgCl2, 0.1 mM of each primer, and 1 unit of Taq DNA polymerase. The primers Ls-cytbF1 and Ls-cytbR1 amplified a fragment of 248 basepair (bp) in the cytb gene. The cycling conditions were: initial denaturation at 94°C for 2 min; cycling 94°C for 20s, 53°C for 20s, 68°C for 30s for 40 cycles; and a final step at 68°C for 2 min. The Ls-coxIIIF2 and Ls-coxIIIR2 primers amplified a DNA fragment of 294 bp; the cycling conditions were: initial denaturation at 94°C for 2 min; cycling 94°C for 20s, 50°C for 20s, and 68°C for 30s for 40 cycles; and a final step at 68°C for 2 min. The PCR products were separated on 2% agarose gels, visualized on a UV light box, and photographed using a Tanon-2500(R) Gel Imaging System (Tanon). 10.1371/journal.pone.0161869.g001Fig 1 Primer design and detection of diluted DNA samples from an infected chicken. A and B, Aligned primer sequences based on genes encoding mitochondrial cytochrome b (cytb, A) and cytochrome oxidase subunit III (coxIII, B) from L. sabrazesi (NCBI accession No. AB299369.1), Leucocytozoon caulleryi (Accession No. AB302215.1), Haemoproteus columbae (NCBI accession No. FJ168562.1), Plasmodium gallinaceum (Accession No. AB250690.1), and chicken (Gallus gallus domesticus, Accession No. KM096864.1). C, Amplifications of diluted DNA from an infected chicken with known gametocytemia. DNA sample from infected chicken (#2HC with parasitemia of 0.02%) blood obtained from Haicang, Fujian province, was diluted in water at ratios of 1:10; 1:50; 1:200; 1:1,000; 1:5,000; 1:20,000; 1:100,000; 1:500,000; 1:2×106; and 1:1×107 and was amplified using Ls-coxIIIF2/R2 and Ls-cytbF1/R1 primers, respectively. PCR products (4 μl each) were separated on a 2% agarose gel. A DNA band could be easily detected at 1:1,000 dilution using Ls-coxIIIF2/R2, and a band at 1:5,000 could be seen using the Ls-cytbF1/R1 primers. “+” indicates un-diluted DNA control. The figure is representative of two replicates with the same results. Drug treatment procedures Female chickens (Gallus gallus domesticus, six to 12 months old) infected with L. sabrazesi were confirmed by microscopic observation and PCR assays and then randomly divided into groups of two each for drug tests. The chickens were housed in wired cages within a room and were fed with commercial chicken food with balanced nutrition from Liaocheng Shunda Feedstuff Co., Ltd, China. The room was cleaned daily, and their droppings, appetite, and activity were monitored at the same time. Pyrimethamine were dissolved in DMSO; primaquine, ketotifen, sulfaquinoxaline, clomipramine hydrochloride, and desipramine hydrochloride were dissolved in water and stored at -20°C until use. The compounds were obtained from Sigma-Aldrich (St. Louis, MO). Chickens were given once-daily oral treatment for 14–20 consecutive days and were kept in a room with screened windows. Chickens receiving no drug treatment served as the control group. Blood samples were collected from chickens at 3- to 6-day intervals for one to two months during and post drug treatment for microscopic examination and for DNA extraction. Results Evaluation of PCR primer sets for the detection of parasite DNA To develop a reliable and sensitive method to detect parasite DNA in the blood, we synthesized two pairs of primers based on L. sabrazesi mitochondrial cytb and coxIII genes, respectively (Ls-cytb F1/R1 and Ls-coxIII F2/R2). The primer sequences were selected based on the aligned sequences of mitochondrial cytb and coxIII genes of L. sabrazesi and those of common avian blood parasites such as Haemoproteus and/or Leucocytozoon spp (Fig 1A and 1B). We first compared the sensitivity of the primer pairs in detecting diluted DNA samples from an infected chicken that had a known parasitemia (0.02%) and showed that both primer sets could detect parasite DNA diluted up to1,000–5,000 fold, with the cytb primers appearing to be slightly more sensitive (a faint band could be seen at 1:20,000 dilution) than the coxIII primers in detecting parasite DNA (Fig 1C). Amplification of water without DNA or uninfected chicken blood cells (see below) did not produce any DNA bands. The results suggest that the PCR methods could detect infection with a parasitemia as low as 2 gametocytes per 1×105 RBCs for coxIII primers and 2 gametocytes per 5×105 RBCs for the cytb primers. Comparison of microscopy and PCR detection We next used the primers to detect DNA samples from chickens that were microscopic positive or negative. Consistent with the results from diluted DNA samples, the Ls-cytbF1/R1 primers were more sensitive in detecting DNA in infected chickens, including some samples that were negative in blood smears (Fig 2A), whereas the coxIII primers appeared to be more specific, consistently producing a single band of expected size. 10.1371/journal.pone.0161869.g002Fig 2 PCR detection of parasite DNAs extracted from chicken blood samples with or without positive identification of gametocytes. A, PCR products amplified using the Ls-coxIIIF2/R2 and Ls-cytbF1/R1 primers, respectively. The numbers on top of the gels are gametocytemia per 1,000 red blood cells (RBCs) after counting 10,000 RBCs. DNAs were extracted from 20 μl infected blood, amplified using the Ls-coxIIIF2/R2 and Ls-cytbF1/R1 primers, respectively, and separated on a 2% agarose gel (4 μl PCR products loaded). DNA ladders of one hundred bp (100–500 bp) are on the left side of the gels. “+” indicates PCR positive; “?” positive with uncertainty; UD, undetectable gametocyte; The figure is representative of two replications of the same results. B, PCR products from microscopic negative samples amplified using the Ls-coxIIIF2/R2 and Ls-cytbF1/R1primers, respectively. C, Electropherograms from four microscopic negative samples by Ls-coxIIIF2/R2 primers showing double peaks in two samples and two nucleotide substitutions in one sample (0315#2). To further confirm the specificity of the PCR methods, we amplified 11 samples that were microscopic negative using the Ls-coxIIIF2/R2 and Ls-cytbF1/R1 primers. Bands of expected sizes were obtained from both primer sets, although additional bands were detected in some samples by the Ls-cytbF1/R1 primers (Fig 2B). We sequenced the PCR products from the Ls-coxIIIF2/R2 primers directly without cloning into a plasmid vector and obtained 11 sequences, including five sequences that appeared to have infections of more than one strain (double electropherogram peaks at one position) (Fig 2C and S1 Fig). The high frequency of mixed infections is consistent with our previous observations [23]. We next cloned the PCR products into Pmd18-T vector for sequencing and obtained 24 individual sequences; again, all the sequences matched the coxIII sequence of L. sabrazesi (S2 Fig). Clustering the amplified sequences with those of representative blood parasites showed that they closely matched the sequence of L. sabrazesi (S3 Fig). These results at least confirmed the specificity of the Ls-coxIIIF2/R2 primers for L. sabrazesi detection [23]. We next amplified additional DNA samples (a total of 102) from chickens previously collected from seven locations in four provinces in China [23] using the Ls-coxIIIF2/R2 primers (these primers were used to be conservative for PCR positives). As expected, more samples were positive by PCR amplification than microscopic observation (Table 1). Among the 102 samples, 45 were positive by microscopy, and all were positive by the Ls-coxIIIF2/R2 PCR method. Among the 57 negative blood smear samples after counting 10,000 RBCs, 13 were positive by PCR (Table 1), providing a positive rate of 56.9%. The results also supported that the PCR method was more sensitive than blood smear in detecting Leucocytozoon infection. 10.1371/journal.pone.0161869.t001Table 1 Comparison of microscopic examination of blood smear and the Ls-coxIIIF2/R2 PCR method in detecting Leucocytozoon sabrazesi infection. Date Sample No. Gametocytemia (%) PCR detection Date Sample No. Gametocytemia (%) PCR detection 11-Dec-12 ZZ-1#1 0.004 + 5-Jul-13 HC-3#4 0.001 + 11-Dec-12 ZZ-1#2 0 - 5-Jul-13 HC-3#5 0.0067 + 11-Dec-12 ZZ-1#3 0.011 + 5-Jul-13 HC-3#6 0.0047 + 11-Dec-12 ZZ-1#4 0 - 5-Jul-13 HC-3#7 0.0168 + 11-Dec-12 ZZ-1#5 0.0003 + 29-Mar-13 FZ-1#1 0 - 11-Dec-12 ZZ-1#6 0 - 29-Mar-13 FZ-1#2 0 + 11-Dec-12 ZZ-1#7 <0.0001 + 29-Mar-13 FZ-1#3 0 + 11-Dec-12 ZZ-1#8 0 - 29-Mar-13 FZ-1#4 0 + 11-Dec-12 ZZ-1#9 0.0007 + 29-Mar-13 FZ-1#5 0 - 11-Dec-12 ZZ-1#10 <0.0001 + 29-Mar-13 FZ-1#6 0 + 11-Dec-12 ZZ-1#11 <0.0001 + 29-Mar-13 FZ-1#7 0 + 11-Dec-12 ZZ-1#12 <0.0001 + 29-Mar-13 FZ-1#8 0 + 11-Dec-12 ZZ-1#13 0 - 29-Mar-13 FZ-1#9 0 + 11-Dec-12 ZZ-1#14 0.0003 + 29-Mar-13 FZ-1#10 0 - 11-Dec-12 ZZ-1#15 0.005 + 29-Mar-13 FZ-1#11 0 - 11-Dec-12 ZZ-1#16 0 - 9-Feb-13 GZ#1 0.0022 + 11-Dec-12 ZZ-1#17 0 - 9-Feb-13 GZ#2 0 - 11-Dec-12 ZZ-1#18 0.009 + 9-Feb-13 GZ#3 0.0046 + 11-Dec-12 ZZ-1#19 0.007 + 9-Feb-13 GZ#4 0 + 11-Dec-12 ZZ-1#20 0.009 + 9-Feb-13 GZ#5 0.0033 + 11-Dec-12 ZZ-1#21 0.003 + 9-Feb-13 GZ#6 0.0045 + 11-Dec-12 ZZ-1#22 <0.0001 + 9-Feb-13 GZ#7 0.0038 + 11-Dec-12 ZZ-1#23 0 - 9-Feb-13 GZ#8 0 + 11-Dec-12 ZZ-1#24 0.022 + 10-Mar-13 RJ#1 0 - 11-Dec-12 ZZ-1#25 <0.0001 + 10-Mar-13 RJ#2 0 - 25-Aug-13 ZZ-2#1 0 + 10-Mar-13 RJ#3 0 - 25-Aug-13 ZZ-2#2 0 + 10-Mar-13 RJ#4 0 - 25-Aug-13 ZZ-2#6 0.018 + 10-Mar-13 RJ#5 0 - 25-Aug-13 ZZ-2#7 0.03 + 10-Mar-13 RJ#6 0 - 25-Aug-13 ZZ-2#8 0.0043 + 10-Mar-13 RJ#7 0 - 25-Aug-13 ZZ-2#9 0.0057 + 10-Mar-13 RJ#8 0 - 25-Aug-13 ZZ-2#11 0.0057 + 10-Mar-13 RJ#9 0 - 25-Aug-13 ZZ-2#13 0 + 10-Mar-13 RJ#10 0 - 25-Aug-13 ZZ-2#14 0.0011 + 8-Feb-13 DZ#1 0 - 25-Aug-13 ZZ-2#19 0.0035 + 8-Feb-13 DZ#2 0 - 23-Mar-13 HC-1#1 0.0184 + 8-Feb-13 DZ#3 0 - 23-Mar-13 HC-1#2 0.0492 + 8-Feb-13 DZ#4 0 - 23-Mar-13 HC-1#3 0.013 + 8-Feb-13 DZ#5 0 - 23-Mar-13 HC-1#4 0.0408 + 8-Feb-13 DZ#6 0 - 23-Mar-13 HC-1#5 0 + 8-Feb-13 DZ#7 0 - 23-Mar-13 HC-1#6 0 - 8-Feb-13 DZ#8 0 - 23-Mar-13 HC-1#7 0.0252 + 8-Feb-13 DZ#9 0 - 23-Mar-13 HC-1#8 0.0108 + 8-Feb-13 DZ#10 0 - 23-Mar-13 HC-1#9 0 - 16-Feb-13 BZ#1 0 - 23-Mar-13 HC-1#10 0.0035 + 16-Feb-13 BZ#2 0 - 23-Mar-13 HC-1#11 0 - 16-Feb-13 BZ#3 0 - 23-Mar-13 HC-1#12 0.0097 + 16-Feb-13 BZ#4 0 - 23-Mar-13 HC-1#13 0.0204 + 16-Feb-13 BZ#5 0 - 5-Jul-13 HC-3#1 0.0343 + 16-Feb-13 BZ#6 0 - 5-Jul-13 HC-3#2 0.0487 + 16-Feb-13 BZ#7 0 - 5-Jul-13 HC-3#3 0.0103 + 16-Feb-13 BZ#8 0 - Gametocytemia (%) were obtained after counting a total 10,000 blood cells. “+” indicates detection of a PCR product of expected size; and “-” means no product of expected size was detected. Note: Sampling locations: ZZ, Zhangzhou, Fujian province; HC, Haicang, Xiamen, Fujian province; FZ, Fuzhou, Fujian province; GZ, Ganzhou, Jiangxi province; RJ, Ruijin, Jiangxi province; DZ, Dazhou, Sichuan Province; BZ, Binzhou of Shandong province. Note, some of the gametocytemia counts were reported previously [23]. Monitoring seasonal infection dynamics in a Haicang chicken farm We next used PCR and microscopy methods to monitor Leucocytozoon infection over a period of one year. We collected blood samples from chickens each month (missing February due to school winter break and national holidays), from October 2014 to September 2015. We then examined blood smears for gametocytes after counting at least 10,000 RBCs each and detected parasite DNA using the Ls-coxIIIF2/R2 PCR. We observed gametocytes in 59 out of 140 blood samples (42.1%) collected in 11 months (no positive smears were found in November, 2014), with infection rates ranging from 0% to 80% (Table 2 and S1 Table). When testing the samples using the Ls-coxIIIF2/R2 PCR primers, parasite DNAs were detected in 106 blood samples (75.7%) collected during all 12 months, with infection rates ranging from 33.3% to 100%. The data also showed that June had the highest infection rate by both methods. Three of the nine birds negative by blood smears in November 2014 were positive by the PCR method (Table 2). 10.1371/journal.pone.0161869.t002Table 2 Monthly surveys of Leucocytozoon infections using microscopy of blood smear and the Ls-coxIIIF2/R2 PCR assay. Month No. birds exam No. birds infected (by smear) % positive by smear No. birds infected (by PCR assay) % positive by PCR assay Oct., 2014 34 15 44.1 29 85.3 Nov., 2014 9 0 0 3 33.3 Dec., 2014 14 3 21.4 6 42.9 Jan., 2015 14 5 35.7 8 57.1 Mar., 2015 21 8 38.1 19 90.5 Apr., 2015 10 4 40 7 70 May., 2015 4 3 75 3 75 Jun., 2015 10 8 80 10 100 Jul., 2015 14 8 57.1 14 100 Aug., 2015 6 4 66.7 5 83.3 Sep., 2015 4 1 25 3 75 Sum 140 59 Ave 43.9 107 Ave 73.9 The animals were randomly selected from the farm at time of initial sampling. No sample was collected for February due to school winter break and national holidays. Monitoring infections in individual chickens over time We also monitored infections of three infected chickens from a free-range farm in Haicang district, Xiamen city, in Fujian province over several months using the same microscopy of blood smears and the PCR method. These chickens were kept in a room at the university campus where no simuliid vectors were found and were sampled weekly in April and May and biweekly from June to October. All three chickens were gametocyte positive in April and became negative in late August by microscopy (Table 3). However, they were all positive in every sampling time by the Ls-coxIIIF2/R2 PCR method, suggesting that they could still transmit the parasite after several months of no active transmission. 10.1371/journal.pone.0161869.t003Table 3 Weekly or biweekly infections of three chickens from a free-range farm in Xiamen, southern China, detected by blood smear or parasite-specific PCR.   HC#1 HC#2 HC#3 Date Gametocytemia (%) PCR detection Gametocytemia (%) PCR detection Gametocytemia (%) PCR detection 1-Apr-13 1 + 0 + 0.017 + 8-Apr-13 0.011 + <0.001 + 0.041 + 16-Apr-13 0.02 + 0.007 + 0.163 + 22-Apr-13 0.012 + 0.008 + 0.101 + 29-Apr-13 0.004 + 0.027 + 0.066 + 6-May-13 <0.001 + 0 + 0.084 + 12-May-13 <0.001 + 0.002 + 0.045 + 17-May-13 0 + <0.001 + <0.001 + 21-May-13 <0.001 + <0.001 + 0.009 + 28-May-13 <0.001 + <0.001 + 0.057 + 7-Jun-13 0 + <0.001 + 0.004 + 17-Jun-13 0 + 0 + 0.012 + 10-Jul-13 0 + 0 + 0.005 + 22-Jul-13 0 + <0.001 + 0 + 5-Aug-13 0 + 0.001 + 0.001 + 15-Aug-13 0 + 0 + 0 + 3-Sep-13 0 + 0 + 0 + 18-Sep-13 0 + 0 + 0 + 2-Oct-13 0 + 0 + 0 + 14-Oct-13 0 + 0 + 0 + Note: HC, Infected chickens from Haicang district, Xiamen city, Fujian province.”+”, indicates PCR products of expected size. Testing potential gametocidal drugs In our previous study of Plasmodium parasites, we identified a class of tricyclic compounds that could block malaria transmission, some of which also had anti-gametocyte activities [37]. Here we were also interested in investigating whether these compounds had activities against gametocytes of L. sabrazesi. We tested primaquine (2 mg/kg), a compound that is known to kill gametocytes of malaria parasites, ketotifen (2 mg/kg) that had some activity against Plasmodium gametocytes, clomipramine hydrochloride (5 mg/kg), desipramine hydrochloride (5 mg/kg), sulfaquinoxaline (20 mg/kg), and pyrimethamine (3.5 mg/kg) (Fig 3 and S2 Table). Primaquine clearly had activities against mature L. sabrazesi gametocytes. Mature gametocytes in chicken #5HC disappeared three days after initial primaquine administration and eight days in chicken #1HC, although a few gametocytes were found in #1HC 21 days after the treatment (Fig 3A). Gametocytes were present in the two chickens treated with ketotifen (#3HC and #4HC) and the two untreated controls (#6HC and #7HC) on most of the days during the periods of study (Fig 3B and 3C and S2 Table). For these chickens, the gametocytemia changed (up and down) with time, suggesting clearing of old gametocytes (due to aging or immunity) and releasing new ones into the blood stream. Similarly, treatment with clomipramine hydrochloride, desipramine hydrochloride, sulfaquinoxaline, or pyrimethamine did not affect gametocytemia, suggesting no or limited effects of these drugs on L. sabrazesi gametocytes (S2 Table). We also used coxIII PCR to detect DNA in the drug-treated chickens (Fig 3D and S2 Table). The results from PCR detection also suggested that only primaquine had some effects on gametocytes, although the parasite positive dates did not match those of microscopic examination. For the primaquine-treated chickens, #1HC became negative on day 12, and #5HC became PCR negative on day 8 after administration of the drug (Fig 3D). However, they both became positive on day 35, suggesting that parasites were being released from internal tissues. Additionally, #5HC also had weak PCR bands from day 15–day 38. In summary, results from both blood smear examination and PCR DNA detection suggest that only primaquine has some effect on L. sabrazesi gametocytes, but may not have much effect on tissue stages. The suppression of gametocytemia by primaquine also suggests that this drug may be used to kill gametocyte and/or interrupt transmission of L. sabrazesi. 10.1371/journal.pone.0161869.g003Fig 3 Detection of Leucocytozoon gametocytes or DNA after drug treatments. A-C, Curves of gametocytemia from two chickens after treatment (daily doses of 2 mg/kg for 14 days) with primaquine (A), ketotifen (B), or no-treatment controls (C). Each line in the graphs of A-C represents gametocytemia from an infected chicken. D, Agarose gels of PCR products from the Ls-coxIIIF2/R2 primers. #1 and #5 were treated with a dose of 2 mg/kg primaquine daily for 14 days; #3 and #4 were treated with a dose of 2 mg/kg ketotifen daily for 14 days; #6 and #7 were controls without treatment. All the chickens were from the Haicang (HC) farm. DNA sample preparation and PCR amplifications were as described in Materials and Methods. Discussion Accurate detection of Leucocytozoon infections is important for transmission control and disease management. Traditionally, diagnosis of Leucocytozoon infection is based on microscopic observation of mature gametocytes in blood smears. PCR amplification of parasite DNA, particularly when nested-PCR is performed [25, 27, 31], is expected to improve the sensitivity in detecting parasite infection. Indeed, PCR detections of blood protozoan infections, including Leucocytozoon infections, have been employed to survey prevalence in migrating birds and in insect vectors, to investigate parasite diversity, and to construct phylogenetic relationships [25, 27, 31]. In addition to amplification of parasite DNA from mature gametocytes, PCR can potentially detect DNA from any other stages of the parasite that are released into the blood stream, whereas microscopic examination is likely to miss immature gametocytes. Although various PCR methods have been reported to detect blood protozoan parasites in avian hosts, a method specifically to detect L. sabrazesi has not been described. To better monitor the transmission and infection of L. sabrazesi in domestic chickens in China, we tested two primer pairs derived from the parasite mitochondrial cytb and coxIII genes. Our results showed that both of the primer pairs were more sensitive in detecting L. sabrazesi infections than microscopic examination of blood smears, although we could not be sure that all the PCR positives (but smear negatives) were truly infected with the parasites. We decided to use the Ls-coxIII R2/R2 primers because they generally produced a single band of expected size, whereas the Ls-cytbF1/R1 primers sometimes had additional bands, including a band of ~350 bp in some samples (Fig 2A). The higher molecular weight band could be from some unknown organisms such as other blood parasites or from a L. sabrazesi strain that had a mutated allele, or it could simply be a non-specific product. To investigate whether the PCR products from microscopic negative samples were L. sabrazesi specific, we sequenced samples from 11 microscopic negative chickens and confirmed that all the sequences matched that of L. sabrazesi. Testing samples using the Ls-coxIIIF2/R2 primers showed that 254 of the 255 microscopy positives (45 individual chickens in Table 1,59 samples from the monthly survey in S1 Table, and 150 samples in the drug treatment in S2 Table) were also positive by the coxIII PCR primers, suggesting a low false negative rate (false negative rate = 1-254/255 = 0.04%). From the same samples, 178 were microscopic gametocyte negative, and among the 178 samples, 88 (55.8%) were also PCR negative. These results suggested that PCR primers did not amplify the host gene because host DNA was present in each sample. The reason for the differences could be explained by the higher sensitivity of the PCR method, as indicated by testing a diluted DNA sample with known parasitemia (Fig 1). The PCR method could detect parasitemia as low as 2 parasites per 100,000 RBCs, while our counting standard was 10,000 RBCs, which was already more than the number of RBCs counted in most routine microscopic examinations. It is difficult to accurately estimate the false positive rate for the coxIII PCR assay because we do not know the real infection rate. We sequenced PCR products from 11 chickens that were microscopic negative, but were PCR positives. There were at least 16 individual parasite strains considering five mixed infections (Cloned sequences are not reliable because we cannot rule out errors introduced by PCR in the individual cloned sequences); all the PCR products were confirmed to be from L. sabrazesi (S3 Fig). These results as well as the mismatches in the primer sequences comparing with those of other representative species of blood parasites, particularly L. caulleryi, suggest that the coxIII PCR primers we used are relative specific for L. sabrazesi. To improve the sensitivity in detecting low-level infections, a nested-PCR method can be considered; however, nested-PCR is also very sensitive to laboratory contaminations because of two-round amplifications. The PCR method we established will be useful for monitoring the transmission of L. sabranesi in endemic regions in the future. We investigated the infection rates each month for a full year and showed that chickens at a free-range farm in Haicang district, Xiamen city in Fujian province, were infected with L. sabrazesi year-round, with peak infection in June. The rainy season generally starts in late April and ends in August, and the average temperatures in the region range from 12°C in January/February to higher than 20°C from May to October. The higher infection rates in June–August were consistent with more rainfall and higher temperature that provide favorable conditions for vector reproduction and survival. The detection of year-round infection is also consistent with our previous observation of relapse of gametocytes; infected chickens could release gametocytes into circulation over a period of several months (Table 3) [23]. The previous observations of long-lasting gametocytes in the blood of infected chickens were further confirmed by our PCR survey monitoring DNA in the blood of three infected chickens. Our data suggest that transmission of the parasites to naïve chickens is likely to occur year-round on free-range farms in Xiamen or in southern China as long as insect vectors are available. We also tested several compounds for treating Leucocytozoon infection, including two known antimalarial drugs (primaquine and pyrimethamine). Primaquine is an antimalarial drug that is known to kill gametocytes and liver stages of the human malaria parasites P. falciparum and P. vivax, and pyrimethamine is mostly used to treat blood asexual stages of malaria parasites [38, 39]. Other compounds we tested (ketotifen, desipramine hydrochloride, and clomipramine hydrochloride) were tricyclic antihistamines/antidepressants (TCAs); some of which were recently shown to have transmission blocking activities against malaria parasites, e.g., preventing oocyst development in a mosquito vector [37]. Our results here showed that primaquine had activity against mature L. sabrazesi gametocytes, but not pyrimethamine. Primaquine is a member of the 8-aminoquinoline drugs and is effective in treating liver and sexual stages of some Apicomplexian parasites, including species of Plasmodium [40] and Leucocytozoon parasites [11, 15], so results of activities against L. sabrazesi are consistent the previous observations. No obvious activities against L. sabrazesi gametocytes were observed for ketotifen, desipramine hydrochloride, clomipramine hydrochloride, or sulfaquinoxaline at the dosages tested. In malaria infection, the transmission blocking activities of these TCAs were based on counting oocysts in mosquito midguts; mature gametocytes may not be the target stages of the compounds. Indeed, ketotifen had weak activity against mature malaria gametocytes, but was very potent in preventing oocyst development in mosquitoes [37]. Additionally, the presence of the parasites in the blood may not reflect the viability/transmissibility of the gametocytes, and bloods with undetectable gametocytes may still allow transmission from an avian host to insect vectors. It would be interesting to test the activities of these compounds in inhibiting the stages of fertilization or early oocyst development in the insect vectors, although they do not have activities against L. sabrazesi mature gametocytes. A simple procedure to evaluate gametocytocidal and transmission blocking activities of artesunate in a Plasmodium gallinaceum-avian model was recently described [41]. Similar methods using Simuliid flies can be developed and used to test the activities of potential transmission blocking compounds against Leucocytozoon parasites; unfortunately, we currently do not have an insectary to raise the insect vectors. This study develops a sensitive PCR method to detect L. sabrazesi infection in chicken. Using the PCR method and microscopic examination of blood smear, we also showed year-round infection of L. sabrazesi in Xiamen, southern China, although the infection rates are lower in the winter months. We also showed high prevalence of L. sabrazesi infection in four of the seven sampling locations in China. The data obtained in this study provide important information for formulating strategies for disease control and management. Supporting Information S1 Fig Aligned DNA sequences amplified from microscopic negative blood samples using Ls-coxIIIF2/R2 primers. PCR products were sequenced directly without cloning into a plasmid vector. (PDF) Click here for additional data file. S2 Fig Aligned DNA sequences of 24 individual clones originated from the 10 microscopic Leucocytozoon sabrazesi negative samples. The PCR products were first cloned into Pmd18-T vector, and DNAs from individual bacterial colonies were extracted and sequenced commercially. Name code: 0419#3p_C4, clone number 4 from sample 0419#3p. (PDF) Click here for additional data file. S3 Fig Clustering of PCR amplified sequence fragments using Ls-coxIIIF2/R2 primers with those of representative blood parasites. The sequences are: Leucocytozoon caulleryi (Accession No. AB302215.1); Haemoproteus columbae (NCBI accession No. FJ168562.1); Plasmodium gallinaceum (Accession No. AB250690.1), Gallus gallus domesticus (Accession No. KM096864.1), and the 11 amplified (uncloned) sequences. The sequences were aligned using Clustal W and clustered using the neighbor-joining method implemented in the program MEGA5 [42]. Bootstrap values higher than 60% are shown after 1,000 permutations. (PDF) Click here for additional data file. S1 Table Monthly surveys of Leucocytozoon infections in a free-range farm from Haicang district, Xiamen city. (XLSX) Click here for additional data file. S2 Table Drug screening against gametocytes of Leucocytozoon sabrazesi. (XLSX) Click here for additional data file. We thank Cindy Clark of the NIH Library for editing. ==== Refs References 1 Valkiunas G . Avin malaria parasites and other haemosporidia Boca Raton, Florida, USA : CRC Press ; 2005 . 2 Dezfoulian O , Zibaei M , Nayebzadeh H , Haghgoo M , Emami-Razavi A , Kiani K . Leucocytozoonosis in domestic birds in southwestern iran: an ultrastructural study . Iran J Parasitol . 2013 ;8 (1 ):171 –6 . Epub 2013/05/18. 23682276 3 Ishak HD , Dumbacher JP , Anderson NL , Keane JJ , Valkiunas G , Haig SM , et al Blood parasites in owls with conservation implications for the Spotted Owl (Strix occidentalis) . PLoS One . 2008 ;3 (5 ):e2304 Epub 2008/05/30. 10.1371/journal.pone.0002304 18509541 4 Bunbury N , Barton E , Jones CG , Greenwood AG , Tyler KM , Bell DJ . Avian blood parasites in an endangered columbid: Leucocytozoon marchouxi in the Mauritian Pink Pigeon Columba mayeri . Parasitology . 2007 ;134 (Pt 6 ):797 –804 . Epub 2007/01/05. 10.1017/S0031182006002149 .17201998 5 Argilla LS , Howe L , Gartrell BD , Alley MR . High prevalence of Leucocytozoon spp. in the endangered yellow-eyed penguin (Megadyptes antipodes) in the sub-Antarctic regions of New Zealand . Parasitology . 2013 ;140 (5 ):672 –82 . Epub 2013/01/31. 10.1017/S0031182012002089 .23361092 6 Imura T , Suzuki Y , Ejiri H , Sato Y , Ishida K , Sumiyama D , et al Prevalence of avian haematozoa in wild birds in a high-altitude forest in Japan . Vet Parasitol . 2012 ;183 (3–4 ):244 –8 . Epub 2011/08/13. 10.1016/j.vetpar.2011.07.027 .21831523 7 Peirce MA , Greenwood AG , Stidworthy MF . Leucocytozoon in captive penguins . Vet Rec . 2005 ;157 (25 ):819 –20 . Epub 2005/12/20. .16361479 8 Hunter DB , Rohner C , Currie DC . Mortality in fledgling great horned owls from black fly hematophaga and leucocytozoonosis . J Wildl Dis . 1997 ;33 (3 ):486 –91 . 10.7589/0090-3558-33.3.486 .9249694 9 Maley GJ , Desser SS . Anemia in Leucocytozoon simondi infections. I. Quantification of anemia, gametocytemia, and osmotic fragility of erythrocytes in naturally infected Pekin ducklings . Can J Zool . 1977 ;55 (2 ):355 –8 . .402177 10 Herman CM , Barrow JH Jr., Tarshis IB . Leucocytozoonosis in Canada geese at the Seney National Wildlife Refuge . J Wildl Dis . 1975 ;11 (3 ):404 –11 . .807752 11 Merino S , Moreno J , Sanz JJ , Arriero E . Are avian blood parasites pathogenic in the wild? A medication experiment in blue tits (Parus caeruleus) . Proc Biol Sci . 2000 ;267 (1461 ):2507 –10 . 10.1098/rspb.2000.1312 11197126 12 Atkinson CT , Forrester DJ , Greiner EC . Pathogenicity of Haemoproteus meleagridis (Haemosporina: Haemoproteidae) in experimentally infected domestic turkeys . J Parasitol . 1988 ;74 (2 ):228 –39 . .3128652 13 Morii T . A review of Leucocytozoon caulleryi infection in chickens . J Protozool Res . 1992 ;2 :128 –33 . 14 Liu W , Wang Z , Wei J , Chu K , Li J , Zhao G , et al Chicken Leucocytozoon disease and morphorlogical characterization of the pathogens (translated from Chiese) . Chinese Journal of veterinary science and technology 1995 ;25 (11 ):40 –1 . 15 Shutler D , Ankney CD , Mullie A . Effects of the blood parasite Leucocytozoon simondi on growth rates of anatid ducklings . Canadian Journal of Zoology . 1999 ;77 (10 ):1573 –8 . 16 Kohne HJ , Jones JE . Blood parameters in Leucocytozoon smithi infected poults fed clopidol . Poult Sci . 1975 ;54 (2 ):472 –4 . .241069 17 Tanabe A , Kondo Y , Toriumi T . [Effect of HPS on leucocytozoon and coccidial infection in chickens ]. Nihon Yakurigaku Zasshi . 1985 ;86 (3 ):175 –9 . .4085928 18 Tanabe A , Kondo Y , Toriumi T . [Effect of sulfamonomethoxine and ormetoprim on leucocytozoon infection in chickens ]. Nihon Yakurigaku Zasshi . 1986 ;87 (3 ):273 –9 . .3086195 19 Fallis AM . Observations on Leucocytozoon infections in birds receiving paludrine, atebrin, and sulphamerazine . Can J Res . 1948 ;26 (2 ):73 –6 . .18865552 20 Hong L , Chen X , Lin Y , Jiang H . Studies on the life cycle of Leucocytozoon sabrazesi Acta Zoologica Sinica . 1990 ;36 (2 ):123 –9 . 21 Chen X . Study on Leucocytozoon sabrazesi gametocyte development in chicken . China Animal Health Inspection . 1995 ;12 (5 ):8 –9 . 22 Weng Z , Zhang Y , Wang J , Zeng Y , Jiang X . Incidence, prevention and treatment of chicken Leucocytozoon . Fujian Journal of Animal Husbandry and Veterinary Medicine 1998 ;2 (10–11 ). 23 Zhao W , Cai B , Qi Y , Liu S , Hong L , Lu M , et al Multi-strain infections and 'relapse' of Leucocytozoon sabrazesi gametocytes in domestic chickens in southern China . PLoS One . 2014 ;9 (4 ):e94877 Epub 2014/04/15. 10.1371/journal.pone.0094877 24728499 24 Gq Li , Yl Hu , Sm Xiao . The establishment of a diagnostic PCR assay for Leucocytozoonosis caulleryi . Chinese Journal of Prevent ive Veterinary Medicine . 2003 ;25 (2 ):148 –51 . 25 Sato Y , Tamada A , Mochizuki Y , Nakamura S , Okano E , Yoshida C , et al Molecular detection of Leucocytozoon lovati from probable vectors, black flies (Simuliudae) collected in the alpine regions of Japan . Parasitol Res . 2009 ;104 (2 ):251 –5 . 10.1007/s00436-008-1183-1 .18791737 26 Ortego J , Cordero PJ . PCR-based detection and genotyping of haematozoa (Protozoa) parasitizing eagle owls, Bubo bubo . Parasitol Res . 2009 ;104 (2 ):467 –70 . 10.1007/s00436-008-1207-x .18818949 27 Ramey AM , Ely CR , Schmutz JA , Pearce JM , Heard DJ . Molecular detection of hematozoa infections in tundra swans relative to migration patterns and ecological conditions at breeding grounds . PLoS One . 2012 ;7 (9 ):e45789 10.1371/journal.pone.0045789 23049862 28 Ferraguti M , Martinez-de la Puente J , Ruiz S , Soriguer R , Figuerola J . On the study of the transmission networks of blood parasites from SW Spain: diversity of avian haemosporidians in the biting midge Culicoides circumscriptus and wild birds . Parasit Vectors . 2013 ;6 :208 10.1186/1756-3305-6-208 23856348 29 Garamszegi LZ . The sensitivity of microscopy and PCR-based detection methods affecting estimates of prevalence of blood parasites in birds . J Parasitol . 2010 ;96 (6 ):1197 –203 . Epub 2010/12/17. 10.1645/GE-2531.1 .21158636 30 Valkiunas G , Iezhova TA , Krizanauskiene A , Palinauskas V , Sehgal RN , Bensch S . A comparative analysis of microscopy and PCR-based detection methods for blood parasites . J Parasitol . 2008 ;94 (6 ):1395 –401 . Epub 2008/06/26. 10.1645/GE-1570.1 .18576856 31 Hellgren O , Waldenstrom J , Bensch S . A new PCR assay for simultaneous studies of Leucocytozoon, Plasmodium, and Haemoproteus from avian blood . J Parasitol . 2004 ;90 (4 ):797 –802 . 10.1645/GE-184R1 .15357072 32 Zhao W , Liu J , Xu R , Zhang C , Pang Q , Chen X , et al The gametocytes of Leucocytozoon sabrazesi infect chicken thrombocytes, not other blood cells . PLoS One . 2015 ;10 (7 ):e0133478 10.1371/journal.pone.0133478 26218846 33 Valkiunas G . Avian malaria parasites and other haemosporidia Boca Raton, Florida, USA : CRC Press ; 2005 . 34 Huff C . Schizogony and gametocyte development in Leucocytozoon simondi, and comparison with Plasmodium and Heamoproteus . J Infect Dis . 1942 ;71 :18 –32 . 35 Wingstrand KG . On some haematozoa of Swedish birds with remarks on the schizogony of Leucocytozoon sakharoffi Stockholm ,: Almqvist & Wiksells boktr ; 1947 . 31 p. p. 36 Steele EJ , Noblet GP . Gametocytogenesis of Leucocytozoon smithi . J Eukaryot Microbiol . 1993 ;40 (3 ):384 –91 . Epub 1993/05/01. .8508175 37 Eastman RT , Pattaradilokrat S , Raj DK , Dixit S , Deng B , Miura K , et al A class of tricyclic compounds blocking malaria parasite oocyst development and transmission . Antimicrob Agents Chemother . 2013 ;57 (1 ):425 –35 . 10.1128/AAC.00920-12 23129054 38 Baird JK , Hoffman SL . Primaquine therapy for malaria . Clin Infect Dis . 2004 ;39 (9 ):1336 –45 . 10.1086/424663 .15494911 39 Baird JK . Effectiveness of antimalarial drugs . N Engl J Med . 2005 ;352 (15 ):1565 –77 . .15829537 40 Ashley EA , Recht J , White NJ . Primaquine: the risks and the benefits . Malar J . 2014 ;13 :418 10.1186/1475-2875-13-418 25363455 41 Kumnuan R , Pattaradilokrat S , Chumpolbanchorn K , Pimnon S , Narkpinit S , Harnyuttanakorn P , et al In vivo transmission blocking activities of artesunate on the avian malaria parasite Plasmodium gallinaceum . Vet Parasitol . 2013 ;197 (3–4 ):447 –54 . 10.1016/j.vetpar.2013.07.024 .23937960 42 Tamura K , Peterson D , Peterson N , Stecher G , Nei M , Kumar S . MEGA5: molecular evolutionary genetics analysis using maximum likelihood, evolutionary distance, and maximum parsimony methods . Mol Biol Evol . 2011 ;28 (10 ):2731 –9 . 10.1093/molbev/msr121 21546353

PMC005xxxxxx/PMC5003345.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 10.1371/journal.pone.0160671PONE-D-15-56264Research ArticleBiology and Life SciencesNeuroscienceCognitive ScienceCognitive NeuroscienceCognitive NeurologyCognitive ImpairmentBiology and Life SciencesNeuroscienceCognitive NeuroscienceCognitive NeurologyCognitive ImpairmentMedicine and Health SciencesNeurologyCognitive NeurologyCognitive ImpairmentMedicine and Health SciencesMental Health and PsychiatryDementiaAlzheimer DiseaseMedicine and Health SciencesNeurologyDementiaAlzheimer DiseaseMedicine and Health SciencesNeurologyNeurodegenerative DiseasesAlzheimer DiseasePeople and PlacesDemographyDeath RatesBiology and Life SciencesPopulation BiologyPopulation MetricsDeath RatesMedicine and Health SciencesHealth CarePatientsMedicine and Health SciencesCardiovascular MedicineCardiovascular DiseasesBiology and Life SciencesPsychologyPsychometricsSocial SciencesPsychologyPsychometricsEngineering and TechnologyEquipmentBiology and Life SciencesNeuroscienceCognitive ScienceCognitionMemoryBiology and Life SciencesNeuroscienceLearning and MemoryMemoryCognitively-Related Basic Activities of Daily Living Impairment Greatly Increases the Risk of Death in Alzheimers Disease Cognitively-Related Impairment Greatly Increases the Risk of Death in Alzheimer's DiseaseLiang Fu-Wen 1Chan Wenyaw 2*Chen Ping-Jen 3Zimmerman Carissa 4Waring Stephen 5Doody Rachelle 61 Department of Public Health, National Cheng Kung University Hospital, College of Medicine, National Cheng Kung University, No.1, University Road, Tainan City 701, Taiwan2 Department of Biostatistics, University of Texas-Health Science Center at Houston, 1200 Pressler Street, E827, Houston, Texas 77030, United States of America3 Department of Geriatrics and Gerontology, Chi-Mei Medical Center, 901, Zhong-Hua Rd., Yong-Kang Dist., Tainan City 710, Taiwan4 Department of Psychology, Rice University, 6100 Main MS-27, Houston, Texas 77005, United States of America5 Essentia Institute of Rural Health, 502 East Second Street, Duluth, MN 55805, United States of America6 Alzheimer's Disease and Memory Disorders Center, Baylor College of Medicine,7200 Cambridge Street, A9.210, Houston, Texas 77030, United States of AmericaBrucki Sonia EditorUniversity Of São Paulo, BRAZILCompeting Interests: The authors have declared that no competing interests exist. Conceived and designed the experiments: FWL WC. Analyzed the data: FWL. Wrote the paper: FWL. Contributed to the interpretation of data for the work: PJC SCW. Contributed to the revision of the manuscript: CAZ RSD WC. * E-mail: wenyaw.chan@uth.tmc.edu29 8 2016 2016 11 8 e016067125 1 2016 22 7 2016 © 2016 Liang et al2016Liang et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Introduction Some Alzheimer’s disease (AD) patients die without ever developing cognitively impaired basic activities of daily living (basic ADL), which may reflect slower disease progression or better compensatory mechanisms. Although impaired basic ADL is related to disease severity, it may exert an independent risk for death. This study examined the association between impaired basic ADL and survival of AD patients, and proposed a multistate approach for modeling the time to death for patients who demonstrate different patterns of progression of AD that do or do not include basic ADL impairment. Methods 1029 patients with probable AD at the Baylor College of Medicine Alzheimer’s Disease and Memory Disorders Center met the criteria for this study. Two complementary definitions were used to define development of basic ADL impairment using the Physical Self-Maintenance Scale score. A weighted Cox regression model, including a time-dependent covariate (development of basic ADL impairment), and a multistate survival model were applied to examine the effect of basic ADL impairment on survival. Results As expected decreased ability to perform basic ADL at baseline, age at initial visit, years of education, and sex were all associated with significantly higher mortality risk. In those unimpaired at baseline, the development of basic ADL impairment was also associated with a much greater risk of death (hazard ratios 1.77–4.06) over and above the risk conferred by loss of MMSE points. A multi-state Cox model, controlling for those other variables quantified the substantive increase in hazard ratios for death conferred by the development of basic ADL impairment by two definitions and can be applied to calculate the short term risk of mortality in individual patients. Conclusions The current study demonstrates that the presence of basic ADL impairment or the development of such impairments are important predictors of death in AD patients, regardless of severity. The authors would like to thank the Cynthia and George Mitchell Foundation (http://cgmf.org/p/home.html) and the Cain Foundation (http://www.thecainfoundation.com/bor/site/) in the Baylor Alzheimer’s Disease and Memory Disorder Center for providing funds that contributed towards this analysis. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll data can be found in attached Supporting Information file (S1 File), minimal data set.Data Availability All data can be found in attached Supporting Information file (S1 File), minimal data set. ==== Body Introduction Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that impairs cognition and daily functioning. Affected patients all die with AD, although the factors that predict survival overlap with those that predict survival in non-demented individuals of similar age: especially age, education, and comorbid conditions [1–11]. Strong negative associations between survival and measures of AD symptom severity would suggest that simple disease progression may drive the duration of survival with AD [1,3,5–10]. Yet, the observation that some patients do and some patients do not pass through a stage of impaired basic ADL prior to death weakens this particular view. If the development of basic ADL impairment exerts an effect on survival that is separate from disease severity, estimating the time to death for patients with and without such basic ADL impairment could be helpful in planning effectively for future demands on medical and social resources [12]. Specifically, patients with impaired basic ADL may require higher care costs in the short term but lower projected medical costs compared to a less impaired subject if they die sooner. The purpose of our study was to examine the relationship between the development of basic ADL impairment, in models that include known predictors of survival, and risk of death among AD patients. Using multistate modeling, we provide a natural way to handle the transitions between disease onset, development of basic ADL impairment, and death while controlling for the expected effects of other baseline characteristics on survival. Materials and Methods Ethics statement Patients came to the Baylor College of Medicine Alzheimer’s Disease and Memory Disorders Center (ADMDC) between January 1989 and February 2008 were eligible to participate in the ADMDC longitudinal follow-up database [13]. This study was approved by the Baylor College of Medicine Institutional Review Board (H-9095). Written informed consent was obtained from all participants. Patient identification Of 1484 probable AD patients (NINCDS-ADRDA [14], 1029 (69%) met inclusion criteria for this study (i.e., had at least one follow-up visit and lived in the greater Houston area at the time of initial visit). Sociodemographic factors (age, sex, race, marital status, and education level), comorbid condition, and a standardized estimate of duration of cognitive symptoms were obtained at baseline along with a battery of psychometric tests as previously described [13,15]. Using the National Cholesterol Education Program–Adult Treatment Panel III guidelines, a cardiovascular disease equipment (CVDE) [16] was calculated based on the history of myocardial infarction, congestive heart failure, stent placement, diabetes mellitus, or high risk for congestive heart disease with any two of hypertension, hyperlipidemia, or current cigarette smoking. Patients were reassessed annually with the same battery of tests. If patients could not be contacted by phone after three attempts, they were considered lost to follow-up. Vital status was obtained from the National Death Index at six month intervals. Measures The duration of AD symptoms was estimated by an ADMDC physician using a standardized procedure [15]. The Mini Mental Status Examination (MMSE) was used to measure dementia severity [17]. Cognitively related basic ADL were assessed with the Physical Self-Maintenance Scale (PSMS) questionnaire developed by Lawton and Brody [18]. Six basic functional abilities were surveyed: bowel and bladder control, feeding, dressing, grooming, ambulation, and bathing. All six items were rated by discussion between a psychometrician and a caregiver on a 5-point rating scale from 1 (no impairment) to 5 (severe impairment) with descriptors for each level of severity. Our psychometricians are trained to rate the cognitive contribution to basic ADL impairment and to disregard impairment related to a physical comorbidity e.g. post-transurethral prostatectomy urinary incontinence. A total score of 6 represents completely intact basic ADL and scores greater than 6 at baseline were considered impaired. Two different definitions were utilized to define the development of basic ADL impairment after baseline: 1) a PSMS score greater than 6 at a post-baseline visit; or 2) an increase in PSMS score of ≥ 2 points relative to baseline. These two complementary definitions allowed for the assessment of both the new development of basic ADL difficulty and the relative worsening of basic ADL function compared to a patient's own baseline score. Statistical analysis First, descriptive statistics of the demographic and clinical characteristics were generated, and comparisons between the two groups of interest (patients with intact ability to perform basic ADL and those with impaired ability to perform basic ADL) were performed using either the χ2 test, the Student’s t test, or the log-rank test. Second, we used fixed-covariate Cox modeling to examine the simultaneous effects of potential independent variables on survival. The validity of the proportional hazards assumption was assessed using Schoenfeld residuals for each variable of interest. Third, time-dependent covariates, namely, development of basic ADL impairment (intact versus impaired), were added to the final Cox model to evaluate their corresponding effects on survival, adjusting for the effects of other selected variables. In addition, the inverse probability of censoring weights (IPCW) [19] was used as weighting in the Cox regression to adjust for possible bias due to loss to follow-up. The estimated weights are based on the predicted probability generated from a logistic regression in which the response variable tells whether the patient was lost to follow-up or not and the independent variables include age at first visit, sex, race, marital status, and education level, baseline PSMS and MMSE scores, duration of cognitive symptoms, and CVDE. Finally, the effect of change in basic ADL status on the prediction of patient outcome was examined using a novel application of multistate survival analysis methods [20–22]. In this approach, the development of Alzheimer’s disease, onset of basic ADL impairment, and death were treated as a three-state Markov chain (Fig 1), and the transitions between phases were modeled. We established a separate overall death rate for those without basic ADL impairment (any individual having impaired basic ADL prior to death was censored to determine this rate, Fig 1, hD1(t)), as well an overall death rate for those who had or developed basic ADL impairment (we used a left-truncated likelihood such that patients entered the risk set at the onset of basic ADL impairment to determine this rate, Fig 1, hD2(t)). A Cox model was then fitted to the data, using the occurrence of basic ADL impairment as the event. Cumulative hazard functions were calculated using the estimates of the risk coefficients and the cumulative baseline hazards using Breslow’s approach [23]. By applying this multistate approach, we obtained a complete picture of how the change in a patient's basic ADL impairment status influences the prediction of his or her survival in the presence of important covariates, that could differ for any path in the Markov chain. All statistical analyses were performed using SAS version 9.1 (SAS Inc., Cary, NC). All tests were considered significant at the 0.05 level. 10.1371/journal.pone.0160671.g001Fig 1 Three-state Markov chain used to derive predictive value of basic ADL impairment. hI(t) is the hazard rate for the time to basic ADL impairment. hD1(t) is the hazard rate for the time to death among basic ADL-intact patients. hD2(t) is the hazard rate for the time to death among basic ADL-impaired patients. The transitions between disease onset, development of basic ADL impairment, and death were modeled by quantifying the effect of baseline characteristics on basic ADL impairment and survival as well as the effect of developing basic ADL impairment on survival. Results Baseline characteristics of the 1029 patients in our sample are shown in Table 1. Among them, 552 were censored at the end of the study period or at the time of last known contact (N = 176). The majority of participants were white, female, and married. Compared to patients who had intact basic ADL, patients who had impaired basic ADL at baseline were significantly older, less educated, and had a longer duration of AD symptoms but the percentage of females did not differ in the two groups. As predicted, median survival time was significantly shorter (by almost 3 years) for patients who reported basic ADL impairment at the first visit (p<0.0001). Furthermore, 45.4% of patients with basic ADL impairment survived compared to 64.0% of patients who were not initially impaired (p<0.0001). 10.1371/journal.pone.0160671.t001Table 1 Baseline patient characteristics and corresponding comparisons by basic activities of daily living (basic ADL) impairment status. Characteristics All patients (n = 1029) Basic ADL-intact at baseline (n = 458) Basic ADL-impaired at baseline (n = 571) p-value Female, n (%) 708 (68.8) 306 (66.8) 402 (70.4) 0.224 White, n (%) 935 (90.9) 436 (95.2) 499 (87.4) <.0001¥ Married, n (%) 626 (60.8) 315 (68.8) 311 (54.5) <.0001¥ Death, n (%) 477 (46.4) 165 (36.0) 312 (54.6) <.0001¥ CVDE$, n (%) 632 (62.2) 271 (59.6) 361 (64.4) 0.118 Age at baseline* 75.1 ± 8.3 72.7 ± 8.1 (39, 92.8) 77.1 ± 8.0 (46.2, 96.5) <.0001€ Formal education* 13.6 ± 3.6 14.3 ± 3.5 (2, 29) 13.1 ± 3.6 (0, 26) <.0001€ Symptom duration* 3.9 ± 2.3 3.4 ± 2.1 (0.5, 18) 4.3 ± 2.4 (0.5, 13) <.0001€ MMSE score at baseline* 19.0 ± 7.0 22.4 ± 4.9 (0, 30) 16.3 ± 7.3 (0, 30) <.0001€ PSMS score at baseline* 8.5 ± 3.8 6.0 (6, 6) 10.5 ± 4.1 (7, 28) <.0001€ Median survival from initial visit (years) 6.6 8.1 5.3 <.0001£ * Results are reported as means ± SD (maximum, minimum) for continuous variables. € Student’s t test, p < 0.001; ¥χ2 test, p < 0.001; £log-rank test, p < 0.001. All statistical results refer to the comparison of the basic ADL-intact versus the basic ADL-impaired groups. $ CVDE: Cardiovascular disease equipment In the weighted Cox regression model (Table 2), age, education, sex, and baseline PSMS and MMSE scores were significant predictors of survival whereas race, marital status, and duration of dementia symptoms had no significant univariate effects on survival. Not unexpectedly, older age was associated with shorter mean survival times, as were less education and male sex when other variables in the multiple model were held constant. Also consistent with past literature, higher scores on the MMSE were associated with longer survival; for every one point reduction on the baseline MMSE, there was a 4% increase in adjusted risk of death. Higher baseline PSMS scores (reflecting basic ADL impairment) were associated with a higher risk of death, even when we controlled for the other demographic and clinical variables: For every one-point increase in baseline PSMS score, there was a 6% increase in risk of death, even when controlling for the other variables. This finding supports our hypothesis that basic ADL impairment may be an independent predictor of survival beyond disease severity alone. 10.1371/journal.pone.0160671.t002Table 2 Hazard ratios (95% confidence intervals) in the baseline Cox regression model and time-dependent Cox regression models for death. Baseline Model Time-dependent Model for Death Variable PSMS ≥ 7 ΔPSMS ≥ 2a Age at baseline 1.03 (1.02, 1.04) 1.03 (1.01, 1.04) 1.03 (1.02, 1.04) Sex (Ref: Female) 1.63 (1.41, 1.89) 2.02 (1.62, 2.51) 1.55 (1.34, 1.79) Race (Ref: White) 0.82 (0.65, 1.04) 0.31 (0.18, 0.51) 0.92 (0.73, 1.15) Education (years) 0.97 (0.96, 0.99) 0.97 (0.95, 0.99) 0.97 (0.97, 0.99) MMSE at baseline 0.96 (0.95, 0.97) 0.94 (0.92, 0.96) 0.97 (0.96, 0.98) PSMS at baseline 1.06 (1.04, 1.08) --- --- Symptom duration 1.00 (0.97, 1.03) 0.99 (0.94, 1.04) 1.01 (0.98, 1.04) Marital status(Ref: Married) 0.97 (0.84, 1.11) 0.72 (0.57, 0.90) 0.91 (0.79, 1.05) CVDE$ 1.10 (0.96, 1.25) 1.04 (0.85, 1.26) 1.09 (0.96, 1.24) ADL impairment during follow-up --- 1.77 (1.41, 2.22) 4.06 (3.30, 5.01) aAn increase in PSMS score of ≥ 2 points relative to baseline $ CVDE: Cardiovascular disease equipment Of the 458 patients whose basic ADL were not impaired at baseline 50.9% (N = 233) developed such impairment by the first definition (PSMS ≥ 7) and 41.5% (N = 190) by the second definition (gained 2 or more points). Because the baseline model does not quantify the effect of developing basic ADL impairment on survival over time, we looked at a time-dependent model. The results of the Cox regression analysis including time-dependent variables (the development of basic ADL impairment) are also displayed in Table 2. The development of basic ADL impairment, defined by development of a PSMS score greater than 6, was strongly and significantly associated with an increased risk of death relative to no basic ADL impairment (hazard ratio 1.77). Age, education and baseline MMSE score had modest but significant effects on survival in this model. There was also the expected effect of male sex reducing survival, but with a lower hazard ratio (1.54) compared to the baseline model predictive value. This argues for the need to include the time-dependent variable. Using our second definition of basic ADL impairment (increase of 2 or more points on the PSMS post-baseline), the effect of worsening basic ADL impairment was even strong and significantly associated with mortality (hazard ratio 4.06), while age, education, baseline MMSE, and sex had similar predictive values to the model using the first definition. The results in Table 2 suggest that the impact of ADL impairment on mortality is at least as strong as the impact of sex, and stronger than the impact of disease severity. Finally, we used a novel multistate modeling approach to estimate the appropriate coefficients or contributions of each variable to the risk of death. We included all potential predictors from the baseline model (age at first visit, sex, race, education level, baseline MMSE score, symptom duration, marital status, and CVDE). We set up a multi-state Cox model defining one state as pre-basic ADL impairment and one state as post ADL impairment. First we designated basic ADL impairment as the outcome event and, the best-fit model revealed that only age at first visit, baseline PSMS and MMSE scores were significant predictors of time to impairment, when controlling for other relevant factors (Table 3). We then moved to a two-stage model selecting survival as the outcome event and applied the estimates from the multistate modeling approach to simultaneously predict the probability of death for an average patient before and after basic ADL impairment (Table 3). The specific hazard ratios in Table 3 allow for individual prediction of mortality within a set interval of time. 10.1371/journal.pone.0160671.t003Table 3 Hazard ratios (95% confidence intervals) in the multistate model for predicting basic ADL impairment or death. Risk of Basic ADL impairment Risk of Death before basic ADL impairment Risk of Death after basic ADL impairment Variable PSMS ≥ 7 ΔPSMS ≥ 2a PSMS ≥ 7 ΔPSMS ≥ 2a PSMS ≥ 7 ΔPSMS ≥ 2a Age at baseline 1.01 (0.99, 1.02) 1.01 (1.00, 1.02) 1.05 (1.03, 1.08) 1.04 (1.03, 1.05) 1.01 (0.99, 1.03) 1.02 (1.01, 1.04) Sex (Male:Female) 0.97 (0.82, 1.15) 0.99 (0.85, 1.17) 2.08 (1.39, 3.12) 1.83 (1.50, 2.23) 1.92 (1.47, 2.52) 1.26 (1.01, 1.58) Race 1.04 (0.80, 1.34) 1.21 (0.93, 1.59) 1.36 (0.72, 2.56) 1.28 (0.94, 1.73) 12.5 (4.16, 37.8) 0.87 (0.60, 1.25) Education 1.02 (1.00, 1.04) 1.01 (0.99, 1.03) 0.95 (0.90, 1.01) 0.94 (0.92, 0.97) 0.99 (0.95, 1.02) 1.01 (0.98, 1.03) MMSE at baseline 0.93 (0.91, 0.95) 0.95 (0.94, 0.96) 0.91 (0.88, 0.96) 0.94 (0.92, 0.95) 0.94 (0.92, 0.97) 0.98 (0.96, 0.99) Symptom duration 0.99 (0.96, 1.03) 1.00 (0.97, 1.03) 0.97 (0.89, 1.05) 1.02 (0.98, 1.06) 1.02 (0.95, 1.09) 1.01 (0.96, 1.05) Marital status 0.90 (0.77, 1.06) 0.98 (0.84, 1.14) 0.70 (0.47, 1.06) 0.83 (0.69, 1.01) 1.85 (1.41, 2.43) 1.32 (1.07, 1.64) CVDE$ 1.07 (0.92, 1.25) 1.03 (0.90, 1.19) 0.59 (0.41, 0.84) 1.08 (0.90, 1.29) 1.36 (1.07, 1.72) 1.10 (0.91, 1.33) aΔPSMS = An increase in PSMS score of ≥ 2 points relative to baseline $ CVDE: Cardiovascular disease equipment In order to demonstrate the value of the multi-state model, we generated survival curves for male and female example patients. ADMDC population information was used to determine the characteristics of an average patient: 75.1 years old at the initial clinic visit with 13.6 years of education and a baseline MMSE score of 19.0. We used the estimates from Table 3 to graph the predicted conditional probability of death in the first 6 years (the overall median survival time of our population was 6.6 years) for a patient who did (the dashed lines) or did not (the solid lines) develop basic ADL impairment over the period of observation (Fig 2). Because sex was also a significant factor in predicting survival time, we calculated the conditional probability of death for each sex separately (men in blue and women in red). As illustrated in Fig 2, the development of basic ADL impairment predicts a much greater chance of death for both sexes. The effect of impaired ADL was most notable with the stricter definition of basic ADL impairment consisting of PSMS score greater than normal (>6). Male patients who developed basic ADL impairment had the greatest predicted probability of death, whereas female patients with no impairment showed the lowest predicted probability of death. The risk of death for impaired women was close to the risk for unimpaired men. The figure illustrates the relative risks of mortality for comparison purpose. 10.1371/journal.pone.0160671.g002Fig 2 Predicted conditional probability of death for patients with and without cognitive based basic ADL impairment. (A)In the left panel, impairment was defined by PSMS ≥ 7. (B)In the right panel, impairment was defined by PSMS ≥ 2 points of increase relative to baseline. The predicted conditional probability of death over the first 6 years was calculated for each sex (men in blue and women in red) assuming either intact (the solid lines) or impaired (the dashed lines) basic ADL at time of observation. Discussion The current study is the first to demonstrate a strong relationship between either the presence or development of basic ADL impairment and the risk of death among AD patients, and the first to quantify this risk over a relevant clinical interval (6 years). The development of basic ADL impairment was a strong signal of increased likelihood of death in both men and women, surpassing the importance of the known risk factors of age, sex, and baseline cognitive scores. This increase in risk was significant and substantial even when disease severity was controlled for in the analysis. Our data can be used to provide quantitative estimates of how changes in an individual patient’s functional abilities will affect their predicted survival over a 6-year period. Other results used in our model were consistent with previous studies showing that male sex, older age, less education, lower baseline MMSE score, and lower baseline basic ADL function significantly increase the risk of death in patients with AD [1–11,24–28]. Importantly, our analysis extends an understanding of the factors associated with mortality by showing that the development of basic ADL impairment is a harbinger of mortality, even when controlling for these other factors. For each point of basic ADL impairment at baseline, the risk of mortality increases 6%—which was comparable to the effects of age, education and baseline MMSE, and less than the effect of sex. However, the new development of basic ADL impairment, defined as either any increase in score or addition of two points of worsening was associated with a much greater risk of death: 1.77 times the risk by the definition of PSMS score greater than 6, and 4.06 times higher based upon a definition of gaining two or more points on this measure. This finding of an increased risk of mortality among patients who developed the basic ADL impairment during the follow-up further supports the importance of basic ADL impairment, regardless of timing of occurrence. Gambassi et al who investigated residents newly admitted to nursing homes with a diagnosis of probable Alzheimer’s disease found that the baseline basic ADL impairment was associated with increased mortality. However, Aguero-Torres et al found the impaired baseline functional status was associated with decreased survival in demented adults, but it became insignificant in patients with Alzheimer’s disease. Researchers have not reached a consensus regarding the best methods for modeling the progression of Alzheimer’s disease [29,30]. To best model the occurrence and timing of an intermediate event (in our case, onset of basic ADL impairment) that precedes the event of interest (in our case, death), Klein et al. [21] suggested using a Cox model with time-dependent covariates, whereas Andersen et al. [20] proposed modeling distinct baseline hazard rates for each possible transition in the multistate process. One common concern in such studies is the potential for biased hazard ratio estimates when between-group comparisons are done using baseline values and without considering the events that occur between baseline and death; such events may have the same effect in both groups of patients, have different effects in each group, or change the rate of death in one group while having no effect in the other. Even when differences in survival are found, the timing of intermediate events could mask differences in marginal probabilities of death between patient groups or make differences appear to exist when there are none. We addressed these issues in two ways: 1) by using time-dependent basic ADL impairment in a Cox regression model and 2) by using a multistate survival model in which a combination of proportional hazards regression and left-truncated proportional hazards regression were applied to synthesize estimates of predictive probabilities of death while taking into consideration the development of basic ADL impairment over time. We demonstrated that the second model is more accurate than the first model because it takes into consideration more than one path of progression and allows predictions to be made about individuals whose characteristics vary. One advantage of the current study is that it included a larger cohort of AD patients than many other studies. In fact, the participants in the Baylor ADMDC cohort represent one of the largest samples of individuals diagnosed with AD by standardized criteria and followed longitudinally in the United States. Another strength of our study was that only cases diagnosed as probable AD by NINCDS-ADRDA criteria [14] were used, which ensured a more homogeneous sample. Additionally, vital status information was available for 100% of the patients, which is another strength. Despite the benefits of our large sample size, this particular cohort does have some limitations. The Baylor College of Medicine ADMDC is a specialty center associated with an academic institution. The sample used in this non-population-based study may contain a disproportionate number of well-informed participants who have better access to health care and fewer comorbid conditions (and therefore higher rates of survival) than the general population [31]. However, this concern is mitigated by the fact that we accept all patients, whether self-referred or referred by others, and are one of the few memory disorders and dementia clinics in the southwestern United States. A limitation of our study is the fact that measurement of basic ADL impairment can be confounded by the presence of comorbid conditions affecting physical function but not cognition. To avoid this, our psychometricians are trained to rate the items on the PSMS based upon cognitive causes rather than physical causes after discussion with the caregiver. Some judgment is still involved in making this assessment, which is a limitation. It is also possible that some overestimation of true survival occurred because patients with rapidly progressing AD who died before they could be diagnosed were not included in our sample. We did not enter item-level data on the basic ADLs, so we cannot assess their significance for survival individually. Finally, comorbid conditions, such as cardiovascular disease, diabetes mellitus, and chronic obstructive pulmonary disease, were not simultaneously analyzed in the current study, because they did not play a significant role in patient survival in other analyses of our population [31,32]. Conclusions There is a relative lack of data in the literature regarding longitudinal aspects of AD. It is essential to determine which factors contribute to AD patients’ risk of death so that preventive and protective measures may be implemented and to improve planning for resources over the course of the disease. Such analyses have clear implications for health economic analysis and public policy. The current study demonstrates that both relative and absolute impairment in basic ADL are important predictors of death in patients diagnosed with AD even when controlling for other factors that influence survival. Our method can quantify this increased risk of death in individual patients when they develop such impairment. In the future, we hope to explore how AD treatment influences the likelihood of developing basic ADL impairment at any given time, as well as the survival of patients who have already developed such impairment. Supporting Information S1 File Data file. (XLSX) Click here for additional data file. We would like to thank the Cynthia and George Mitchell Fund and the Cain Foundation in the Baylor Alzheimer’s Disease and Memory Disorder Center for providing funds that contributed towards this analysis. ==== Refs References 1 Agüero-Torres H , Fratiglioni L , Winblad B . Natural history of Alzheimer's disease and other dementias: review of the literature in the light of the findings from the Kungsholmen Project . International journal of geriatric psychiatry . 1998 ; 13 : 755 –766 . 9850872 2 Belloni-Sonzogni A , Tissot A , Tettamanti M , Frattura L , Spagnoli A . Mortality of demented patients in a geriatric institution . Archives of gerontology and geriatrics . 1989 ; 9 : 193 –197 . 2589919 3 Bianchetti A , Scuratti A , Zanetti O , Binetti G , Frisoni GB , Magni E , et al Predictors of mortality and institutionalization in Alzheimer disease patients 1 year after discharge from an Alzheimer dementia unit . Dementia and Geriatric Cognitive Disorders . 1995 ; 6 : 108 –112 . 4 Bowen JD , Malter A , Sheppard L , Kukull W , McCormick W , Teri L , et al Predictors of mortality in patients diagnosed with probable Alzheimer's disease . Neurology . 1996 ; 47 : 433 –439 . 8757016 5 Claus JJ , van Gool WA , Teunisse S , Walstra GJ , Kwa VI , Hijdra A , et al Predicting survival in patients with early Alzheimer’s disease . Dementia and geriatric cognitive disorders . 1998 ; 9 : 284 –293 . 9701680 6 Hier DB , Warach JD , Gorelick PB , Thomas J . Predictors of survival in clinically diagnosed Alzheimer's disease and multi-infarct dementia . Archives of neurology . 1989 ; 46 : 1213 –1216 . 2818256 7 Knopman DS , Kitto J , Deinard S , Heiring J . Longitudinal study of death and institutionalization in patients with primary degenerative dementia . Journal of the American Geriatrics Society . 1988 ; 36 : 108 –112 . 3339214 8 Lapane KL , Gambassi G , Landi F , Sgadari A , Mor V , Bernabei R . Gender differences in predictors of mortality in nursing home residents with AD . Neurology . 2001 ; 56 : 650 –654 . 11245718 9 Mölsä P , Marttila R , Rinne U . Long-term survival and predictors of mortality in Alzheimer's disease and multi-infarct dementia . Acta neurologica scandinavica . 1995 ; 91 : 159 –164 . 7793228 10 Mölsä PK , Marttila R , Rinne U . Survival and cause of death in Alzheimer's disease and multi-infarct dementia . Acta Neurologica Scandinavica . 1986 ; 74 : 103 –107 . 3776457 11 Wolfson C , Wolfson DB , Asgharian M , M'Lan CE , Østbye T , Rockwood K , et al A reevaluation of the duration of survival after the onset of dementia . New England Journal of Medicine . 2001 ; 344 : 1111 –1116 . 11297701 12 Santerre R , Neun S . Helth economics: theory, insights and industry studies . Chicago : Irwin 1996 . 13 Doody R , Pavlik V , Massman P , Kenan M , Yeh S , Powell S , et al Changing patient characteristics and survival experience in an Alzheimer's center patient cohort . Dement Geriatr Cogn Disord . 2005 ; 20 : 198 –208 . 16088145 14 McKhann G , Drachman D , Folstein M , Katzman R , Price D , Stadlan E . Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease . Neurology . 1984 ; 34 : 939 –944 . 6610841 15 Doody RS , Dunn JK , Huang E , Azher S , Kataki M . A method for estimating duration of illness in Alzheimer's disease . Dement Geriatr Cogn Disord . 2004 ; 17 : 1 –4 .14560058 16 Doody R. S. , Demirovic J. , Ballantyne C. M. , Chan W. , Barber R. , Powell S. , & Pavlik V. Lipoprotein-associated phospholipase A2, homocysteine, and Alzheimer's disease . Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring . 2015 ; 1 (4 ), 464 –471 17 Folstein MF , Folstein SE , McHugh PR . "Mini-mental state". A practical method for grading the cognitive state of patients for the clinician . J Psychiatr Res . 1975 ; 12 : 189 –198 . 1202204 18 Lawton MP , Brody EM . Assessment of older people: self-maintaining and instrumental activities of daily living . Gerontologist . 1969 ; 9 : 179 –186 . 5349366 19 Weuve J. , Tchetgen Tchetgen E J , Glymour MM , Beck TL , Aggarwal NT , Wilson R S ,et al Accounting for bias due to selective attrition: the example of smoking and cognitive decline . Epidemiology (Cambridge, Mass.) . 2012 ; 23 (1 ) 20 Andersen PK , Hansen LS , Keiding N . Non-and semi-parametric estimation of transition probabilities from censored observation of a non-homogeneous Markov process . Scandinavian Journal of Statistics . 1991 ; 18 : 153 –167 . 21 Klein JP , Keiding N , Copelan EA . Plotting summary predictions in multistate survival models: probabilities of relapse and death in remission for bone marrow transplantation patients . Stat Med . 1993 ; 12 : 2315 –2332 . 8134735 22 Klein JP , Moeschberger ML Survival analysis: techniques for censored and truncated data : Springer Science & Business Media 2003 . 23 Agüero‐Torres H , Fratiglioni L , Guo Z , Viitanen M , Winblad B . Prognostic factors in very old demented adults: a seven-year follow-up from a population-based survey in Stockholm . Journal of the American Geriatrics Society . 1998 ; 46 : 444 –452 . 9560066 24 Heyman A , Peterson B , Fillenbaum G , Pieper C . The Consortium to Establish a Registry for Alzheimer's Disease (CERAD). Part XIV Demographic and clinical predictors of survival in patients with Alzheimer's disease . Neurology . 1996 ; 46 : 656 –660 . 8618662 25 Moritz DJ , Petitti DB . Association of education with reported age of onset and severity of Alzheimer's disease at presentation: implications for the use of clinical samples . Am J Epidemiol . 1993 ; 137 : 456 –462 . 8460626 26 Reisberg B , Ferris SH , Franssen EH , Shulman E , Monteiro I , Sclan SG , et al; Mortality and temporal course of probable Alzheimer's disease: a 5-year prospective study . Int Psychogeriatr . 1996 ; 8 : 291 –311 . 8994898 27 Stern Y , Tang MX , Albert MS , Brandt J , Jacobs DM , Bell K , et al Predicting time to nursing home care and death in individuals with Alzheimer disease . JAMA . 1997 ; 277 : 806 –812 . 9052710 28 van Dijk PT , van de Sande HJ , Dippel DW , Habbema JD . The nature of excess mortality in nursing home patients with dementia . J Gerontol . 1992 ; 47 : M28 –34 . 1538062 29 Doody RS , Pavlik V , Massman P , Rountree S , Darby E , Chan W . Predicting progression of Alzheimer’s disease . Alzheimers Res Ther . 2010 ; 2 . 30 Spiegel R , Doody RS , Hendrix S , Kahle-Wrobleski K . Modeling the course of Alzheimer's disease to improve clinical trials: symposium report . Alzheimers Res Ther . 2013 ; 5 : 29 doi: 10.1186/alzrt183 23767782 31 Waring SC , Doody RS , Pavlik VN , Massman PJ , Chan W . Survival among patients with dementia from a large multi-ethnic population . Alzheimer Dis Assoc Disord . 2005 ; 19 : 178 –183 . 16327343 32 Rountree SD , Chan W , Pavlik VN , Darby EJ , Siddiqui S , Doody RS .Persistent treatment with cholinesterase inhibitors and/or memantine slows clinical progression of Alzheimer disease . Alzheimers Res Ther . 2009 ; 1 : 7 doi: 10.1186/alzrt7 19845950

PMC005xxxxxx/PMC5003346.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757150710.1371/journal.pone.0161992PONE-D-16-13619Research ArticleMedicine and Health SciencesNeurologyEpilepsyEngineering and TechnologyEnvironmental EngineeringPollutionAir PollutionPhysical SciencesChemistryEnvironmental ChemistryPollutantsEcology and Environmental SciencesEnvironmental ChemistryPollutantsEarth SciencesAtmospheric ScienceMeteorologyHumidityPeople and PlacesGeographical LocationsAsiaChinaMedicine and Health SciencesHealth CarePatientsOutpatientsEarth SciencesAtmospheric ScienceMeteorologyWeatherEarth SciencesAtmospheric ScienceAtmospheric ChemistryGreenhouse GasesOzonePhysical SciencesChemistryEnvironmental ChemistryAtmospheric ChemistryGreenhouse GasesOzoneEcology and Environmental SciencesEnvironmental ChemistryAtmospheric ChemistryGreenhouse GasesOzoneThe Novel Relationship between Urban Air Pollution and Epilepsy: A Time Series Study Air Pollution and EpilepsyXu Chen 1Fan Yan-Ni 2Kan Hai-Dong 3Chen Ren-Jie 3Liu Jiang-Hong 4Li Ya-Fei 1Zhang Yao 1Ji Ai-Ling 5*http://orcid.org/0000-0003-3188-6265Cai Tong-Jian 1*1 Department of Epidemiology, College of Preventive Medicine, Third Military Medical University, Chongqing, China2 Information Department Medical Record Room, Second Affiliated Hospital, Fourth Military Medical University, Xi’an, China3 Department of Environmental Health, School of Public Health, Fudan University, Shanghai, China4 School of Nursing, University of Pennsylvania, Philadelphia, Pennsylvania, United States of America5 School of Public Health, Fourth Military Medical University, Xi'an, ChinaRusso Emilio EditorUniversity of Catanzaro, ITALYCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: TJC ALJ. Data curation: CX TJC. Formal analysis: CX HDK RJC JHL. Funding acquisition: TJC YZ YFL. Investigation: YNF TJC. Methodology: HDK RJC TJC. Project administration: TJC ALJ. Resources: YNF TJC. Supervision: TJC. Validation: YZ YFL. Visualization: TJC. Writing – original draft: CX TJC ALJ JHL YZ YFL. * E-mail: ctjcsl@netease.com; ctjcsl@hotmail.com (TJC); jailing2008@hotmail.com (ALJ)29 8 2016 2016 11 8 e01619925 4 2016 16 8 2016 © 2016 Xu et al2016Xu et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Background and purpose The data concerning the association between environmental pollution and epilepsy attacks are limited. The aim of this study was to explore the association between acute air pollution exposure and epilepsy attack. Methods A hospital record-based study was carried out in Xi’an, a heavily-polluted metropolis in China. Daily baseline data were obtained. Time-series Poisson regression models were applied to analyze the association between air pollution and epilepsy. Results A 10 μg/m3 increase of NO2, SO2, and O3 concentrations corresponded to 3.17% (95%Cl: 1.41%, 4.93%), 3.55% (95%Cl: 1.93%, 5.18%), and -0.84% (95%Cl: -1.58%, 0.09%) increase in outpatient-visits for epilepsy on the concurrent days, which were significantly influenced by sex and age. The effects of NO2 and SO2 would be stronger when adjusted for PM2.5. As for O3, a -1.14% (95%Cl: -1.90%, -0.39%) decrease was evidenced when adjusted for NO2. The lag models showed that the most significant effects were evidenced on concurrent days. Conclusions We discovered previously undocumented relationships between short-term air pollution exposure and epilepsy: while NO2 and SO2 were positively associated with outpatient-visits of epilepsy, O3 might be associated with reduced risk. http://dx.doi.org/10.13039/501100001809National Natural Science Foundation of China81372952Tongjian Caihttp://dx.doi.org/10.13039/501100001809National Natural Science Foundation of China81202220Zhang Yao http://dx.doi.org/10.13039/501100001809National Natural Science Foundation of China81472190Li Ya-Fei The authors received no specific funding for this work. Data AvailabilityAll relevant data are within the paper and its Supporting Information.Data Availability All relevant data are within the paper and its Supporting Information. ==== Body Introduction As one of the most prevalent neurological diseases, epilepsy affects more than 50 million people worldwide with a higher prevalence rate in low-income countries [1]. A meta-analysis reports that the median prevalence of epilepsy was 5.8‰ in developing countries [2]. However, the etiology and the mechanisms involved are not clear. Air pollution exerts great threats to human health, especially diseases of respiratory and cardiovascular systems [3,4]. In recent years, the relationship between air pollution and nervous system diseases has been recognized gradually. Calderon et al report that the sustained exposure to air pollutants can seriously affect pediatric central nervous system [5]. More importantly, the association between air pollution and neurodegenerative diseases is widely accepted [6]. To date, limited studies have been performed regarding the influence of air pollution on epilepsy. Although improved substantially, China has the worst air pollution in the world [7]. The industrialization, urbanization, and increased vehicle use result in the increased air pollution in major cities [8]. In this study, we investigated the effects of urban ambient air pollution on the attack of epilepsy in Xi’an, a heavily-polluted metropolis in China. We hope our data can provide clues for the associations between air pollution and epilepsy. Methods Data collection Xi’an, with an area of 10,108 km2 and a resident population of 8.262 million in 2014, is the largest city in northwestern China. It experiences some of the worst air pollution among China’s cities [9]. Here, we limited the area in urban districts of Xi’an, an area of 3,586 km2 with a resident population of 6.565 million. Daily outpatient data were obtained from Tangdu Hospital, one of the largest hospitals in western China. In the outpatient department, the physicians enter medical record data for each patient into the computer system, including individual characteristics (such as gender, age, and residence) and diagnoses. In this study, daily numbers of epilepsy outpatient-visits for epilepsy from urban areas of Xi’an between January 1, 2013 and December 31, 2014 were included. Patient records were de-identified and daily aggregated data were calculated for analysis. There was no individual interaction with patients. The protocol was approved by the Ethics Committee of Third Military Medical University. Daily (24h) air pollution data including particulate matter less than 10 μm in aerodynamic diameter (PM10), particulate matter less than 2.5 μm in aerodynamic diameter (PM2.5), carbon monoxide (CO), nitrogen dioxide (NO2), sulfur dioxide (SO2), and ozone (O3) from January 1, 2013 to December 31, 2014 were obtained from China National Environmental Monitoring Center. The daily concentrations for pollutants were averaged from the available data of thirteen fixed-site automated monitoring stations in urban areas of Xi’an. All the monitored results reflected the general urban levels according to their strategic location designs. In order to adjust for the potential confounding effects, daily weather data (mean temperature and relative humidity) were collected from China Meteorological Bureau. Statistical methods Risk of epilepsy and pollutants and weather data were based on a larger population, so we assumed a Poisson distribution. A generalized additive model (GAM) was applied. The epilepsy outpatient-visits, air pollutants, and weather data were linked by date, so we used time-series model to analyze. In order to examine the effect of air pollutants on outpatient-visits for epilepsy, we controlled the potential confounders such as long-term trend, the day of the week (DOW), and meteorological factors. We fitted non-parametric smoothing terms for the trend on days and a dummy variable for DOW. We applied natural smooth (ns) functions of calendar time. Akaka’s Information Criterion (AIC) was used to determine how well the models fitted the data and smaller AIC values of per year for time trends indicated the preferred model (S1 Table). The effects were stable when 7 or 8 degrees of freedom per year for time were used. Considering the results from AIC and other studies that had been published [10], 7 degrees of freedom (df) was found to be the best suitable for our current study. We incorporated ns functions of mean temperature (6 df) and relative humidity (3 df) to adjust for the potential nonlinear confounding effects of weather conditions based on published literature [11]. On the other hand, we compared the estimated effects by different degrees of freedoms per year for time, and the stable estimated effects suggested that the basic model was suitable (S2 Table). We also considered days of week and public holidays in the models as indicator variables. We defined lag effects of different days including both single-day lag from lag 0 to lag 7 and moving average of lag 07 (concurrent day and previous 7 days). As for air pollutants, we examined the delay days on the effects. In addition to single-day lag from lag 0 to lag 7, lag 07 was used to estimate the cumulative effects of pollutants. Residuals of each model were examined to check whether there were discernible patterns and autocorrelation by means of residual plots and the partial autocorrelation function (PACF) plots (S1 Fig). In addition, we also checked the lag effects for temperature, which could reflect the stability of models (S2 Fig). The independent model is described below: logE(Yβ) = βZt+ DOW + ns (time, df) + ns (temperature, 6) + ns (humidity, 3) + intercept Here E (Yβ) means the expected number of outpatient visits at day t; β represents the log-relative rate of outpatient visits associated with a unit increase of pollutant concentration; Zt indicates the pollutant concentrations at day t; DOW is day of the week effect; ns(time,df) is the natural spline function of calendar time; and ns(temperature/humidity,6/3) is the natural spline function for temperature and humidity with 6/3 df. After establishment of the basic models, we determined which pollutants had relationship with epilepsy, investigating the effects of lag days for single pollutants. Second, we investigated whether the associations between main pollutants and epilepsy were sensitive to the adjustment: other pollutants were included one by one as potential co-pollutants with major pollutants at all lag structures. Besides the above methods, we also performed sex- and age- specific analyses and plotted exposure–response relationship curves for main pollutants. We considered P<0.05 to be statistically significant. All analyses were conducted by R software (version 2.15.1) using the mgcv package. Unless specified otherwise, the pollutant effect was generally expressed as a percentage increase of outpatient epilepsy visits with a 10-μg/m3 increase of pollutants per day (95% confidence intervals (CIs)). Results Table 1 summarizes basic descriptive statistics. There were 20,368 epilepsy outpatient-visits during January 1, 2013 and December 31, 2014, including 12,041 males and 8,327 females. The average pollutant concentrations were 169 μg/m3 for PM10, 91 μg/m3 for PM2.5, 38.6 μg/m3 for SO2, 51.3 μg/m3 for NO2, 1.939 mg/m3 for CO, and 100 μg/m3 for O3. Meanwhile, the average daily temperature and humidity were 15.6°C and 61%. 10.1371/journal.pone.0161992.t001Table 1 Descriptive statistics for daily morbidity, concentrations of air pollutants, and weather conditions. Mean Min Max P25 P50 P75 SD Epilepsy 28 1 72 15 27 38 15 Sex Male 16.49 0 48 9 16 22 9.32 Female 11.4 0 34 6 11 16 6.77 Age Children (<18) 10.35 0 33 5 9 15 6.47 Adult (18–59) 16 0 51 9 15 22 9.18 Elderly (>59) 1.547 0 8 0 1 2 1.48 Air pollutant concentrations (24-h average) SO2 (μg/m3) 38.6 3.3 170.0 15.0 27.0 53.0 32.3 NO2 (μg/m3) 51.3 14.7 141.0 37.2 47.9 62.0 19.4 CO (mg/m3) 1.939 0.747 6.219 1.289 1.685 2.325 0.92 O3 (μg/m3) 100 7 311 50 87 144 63 PM2.5 (μg/m3) 91 10 606 44 66 105 79 PM10 (μg/m3) 169 18 1020 95 144 199 115 Meteorological measure (24-h average) Temperature (°C) 15.6 -5.0 33.9 6.5 16.6 24.2 9.9 Relative humidity (%) 61 16 97 49 61 74 17 Spearman correlation coefficients between air pollutants and meteorological factors were presented in Table 2. Except for humidity with PM2.5, CO and temperature, there were moderate positive correlation coefficients between pollutants and two meteorological factors. 10.1371/journal.pone.0161992.t002Table 2 Spearman correlation coefficients between daily air pollutant concentrations and weather conditions. PM2.5 SO2 NO2 O3 CO Temperature Humidity PM10 0.920** 0.719** 0.762** -0.206** 0.752** -0.469** -0.199** PM2.5 0.712** 0.723** -0.226** 0.791** -0.492** 0.024 SO2 0.769** -0.414** 0.785** -0.766** -0.259** NO2 -0.165** -0.725** -0.497** -0.184** O3 -0.454** 0.778** -0.130** CO -0.730** -0.056 Temperature -0.001 **P<0.01 Table 3 summarizes the all-, sex- and age- specific effects for each pollutant on the concurrent days. Positive and statistically significant relationships were observed with both SO2 (P<0.0001) and NO2 (P<0.001), the corresponding increases were 3.55% (95%Cl: 1.93%, 5.18%) and 3.17% (95%Cl: 1.41%, 4.93%) with per 10μg/m3 increase in the concentrations of pollutants. Unexpectedly, O3 had negative association with epilepsy (P<0.05): a 10 μg/m3 increase in it was associated with a 0.84% (95%Cl: -1.58%, -0.09%) decrease in outpatient-visits. In sex-specific analysis, males showed stronger associations for SO2 (P<0.0001) and O3 (P<0.01) than females. However, only NO2 (P<0.005) showed more significant association in females. In age-specific analysis, the effects of NO2 and SO2 on children (aged <18) tended to be stronger than those on adult (aged 18–59) and elderly (aged>59) groups. In contrast, the effect tended to be strongest in adult group (aged 18–59) for O3. Interestingly, there was no significant effect of air pollutants on elderly group. 10.1371/journal.pone.0161992.t003Table 3 Percent change (mean and 95% confidence interval) in daily outpatient-visits for epilepsy associated with a 10 μg/m3 (PM10, PM2.5, SO2, NO2, and O3) or 0.1 mg/m3 (CO) increase of air pollutants on the concurrent days in single-pollutant models. PM10 PM2.5 SO2 NO2 O3 CO All 0.14 (-0.13, 0.41) 0.20 (-0.23, 0.62) 3.55 (1.93, 5.18)*** 3.17 (1.41, 4.93)*** -0.84 (-1.58, -0.09)* 0.11 (-0.37, 0.59) Sex Male 0.22 (-0.08, 0.52) 0.27 (-0.21, 0.74) 3.81 (1.98, 5.64)*** 3.13 (1.13, 5.12)** -0.89 (-1.73, -0.04)* 0.16 (-0.38, 0.70) Female 0.01 (-0.33, 0.35) 0.09 (-0.44, 0.62) 3.17 (1.14,5.19)** 3.23 (1.05, 5.40)** -0.77 (-1.68, 0.15) 0.03 (-0.57, 0.62) Age <18 0.13 (-0.22, 0.48) 0.31 (-0.24, 0.85) 4.64 (2.54, 6.74)*** 4.55 (2.28, 6.82)*** -0.82 (-1.77, 0.14) 0.28 (-0.34, 0.90) 18–59 0.21 (-0.09, 0.51) 0.23 (-0.26, 0.71) 2.95 (1.09, 4.81)** 2.87 (0.86, 4.88)** -0.87 (-1.73, -0.02)* 0.03 (-0.51, 0.58) >59 -0.60 (-1.42, 0.22) -1.01 (-2.29, 0.27) 2.54 (-2.04, 7.11) -3.14 (-8.29, 2.01) -0.58 (-2.72, 1.56) -0.26 (-1.65, 1.13) *P<0.05 **P<0.01 ***P<0.001 Fig 1 shows the results from the single-lag day (lag0-lag7) and cumulative exposure models (lag 07) for the percent increase of epilepsy number with per 10μg/m3 increase in pollutants. The most significant effects were evidenced on the concurrent days (lag0). The obvious effects of NO2, SO2, and O3 concentration were limited on the concurrent and lag 1 days, with each 10μg/m3 increase of NO2, SO2 and O3 corresponded to 3.17% (95%Cl: 1.41%, 4.93%), 3.55% (95%Cl: 1.93%, 5.18%) and -0.84% (95%Cl: -1.58%, -0.09%) increase of outpatient-visits at lag 0 and 2.27% (95%Cl: 0.48%, 4.07%), 2.99% (95%Cl: 1.35%, 4.63%) increase at lag 1 for NO2 and SO2. In particular, negative association between epilepsy and NO2 was evidenced after lag 3, which might be a random and unexplainable result. 10.1371/journal.pone.0161992.g001Fig 1 Percent increase in number of daily outpatient-visits for epilepsy associated with a 10-μg/m3 increase of air pollutants using different lag days in single-pollutant models. Values are reported as means and 95% confidence intervals. Abbreviations: Lag 07 day-moving average concentrations. Fig 2 shows the association between number of epilepsy patients and increase of NO2, SO2 and O3 concentrations in two-pollutant models. Except for the adjustment of NO2 for SO2, all pollutants’ associations were fairly stable after adjusted with others. The associations between epilepsy visits and NO2 as well as SO2 were more significant after adjustment for PM2.5 (4.79% (95%Cl: 2.39%, 7.18%), 4.19% (95Cl: 2.31%, 6.07%)). As for O3, the estimated decrease effect was at -1.14% (95%Cl: -1.90%, -0.39%) when adjusted for NO2. 10.1371/journal.pone.0161992.g002Fig 2 Percent increase in number of daily outpatient-visits for epilepsy associated with a 10-μg/m3 increase of air pollutants (NO2, SO2, and O3) in two-pollutant models. Values are reported as means and 95% confidence intervals. Fig 3 shows that while the exposure–response relationships for NO2 and SO2 were generally linear and positive, O3 was generally linear and negative with outpatient-visits. From these curves, we did not observe any obvious threshold concentration below which there were no effects. 10.1371/journal.pone.0161992.g003Fig 3 Smoothing plots of air pollution. X-axis is the concentration of pollutants (μg/m3). The estimated mean percentage of change in daily outpatient-visits for epilepsy is shown by the solid line, and the dotted lines represent twice the point-wise standard error. Discussion In the present study, SO2 and NO2 had positive relationships with epilepsy visits. Unexpectedly, there was negative relationship between O3 and epilepsy outpatient-visits. The associations were significant for both males and females, and were more significant for children (aged <18) and adults (aged 18–59) than for the elder (aged>59). When adjusted for PM2.5, the effects of NO2 and SO2 were more obvious. As for O3, we found stronger “protective” effect after adjustment for NO2. The lag model showed that the most significant effects were evidenced on concurrent days. Specially, we found a negative association between epilepsy and NO2 after lag 3, which might be the random and unexplainable result. Considering NO2 always produces acute health effects, we inclined to focus on the effects on the concurrent days while the results from lag models were used as references. Before our study, only one study reported that epilepsy was influenced by air pollution. In Chile, Cakmak et al found that CO, O3, SO2, NO2, PM10, and PM2.5 were all positively associated with epilepsy hospitalization, and the results were not significantly influenced by age, sex, or season [12]. Different from that, only SO2 and NO2 were positively associated with epilepsy outpatient-visits here. More interestingly, we showed that O3 was negatively associated with epilepsy visits, suggesting that O3 might be a “protective” factor against epilepsy. The difference between our results and those of Cakmak et al might be due to the differences in races, climate, air pollution, or health care levels. NO2 is an important part of automobile exhaust. Heretofore, the direct evidence of epilepsy induced by NO2 is limited. In the south of Spain, even low level NO2 exposure and traffic-related air pollution had side-effects on neurodevelopment, especially in children [13]. NO2 exposure also induces the changes of triglyceride, free fatty acids, esterified fatty acid, ganglioside, lipase activity and lipid peroxidation, and eventually leads to impaired nerve function in central nervous system [14,15]. Our present work added new evidence concerning the neurotoxic effects of NO2. SO2 is a well-known irritant confirmed to induce respiratory responses. It has also been linked to nervous system disorders such as stroke and headache [16,17]. Cakmak et al have demonstrated the positive relationship between SO2 and epilepsy [12], which was consistent with our findings. The neurotoxic effect of SO2 has also been supported by animal studies. Sang et al reported the toxic effects of SO2 on hippocampus, including protein oxidation, DNA-protein crosslinks, and apoptosis [18]. Further work showed that SO2 may cause synaptic injuries in hippocampus via its derivatives [19]. More recently, chronic SO2 inhalation above environmental standard has been found to induce neurotoxicities via neuroinflammation [20]. In summary, SO2 exposure might be an environmental risk factor for epilepsy attacks. However, the mechanisms involved needed to be further clarified. Though the stratospheric O3 layer protects human health against excess solar ultraviolet radiation [21], the O3 in earth's surface air can be dangerous. However, in the present study, negative association was observed between ambient O3 and epilepsy. This can be supported by a series of previous studies. León et al provided the first experimental evidence that repeated ozone administration at atoxic doses induces adaptation to oxidative stress, enabling animals to maintain hepatocellular integrity in response to CCl4 poisoning [22]. Controlled ozone administration also promotes oxidative preconditioning, preventing the damage induced by reactive oxygen species (ROS) and protecting against liver ischaemia-reperfusion (I/R) injury [23]. For epilepsy, the activated A1 adenosine receptors (A1Rs) play a pivotal role in antiepileptic and neuroprotective functions [24] while antioxidant/prooxidant imbalance is an important factor for epilepsy [25]. Ozone can exert its protective effects against pentylenetetrazole (PTZ)-induced epilepsy by re-establishment of cellular redox balance and A1Rs [25]. With the development of clinical applications, reasonability and effectiveness of O3 therapy are gradually being understood [26]. Medical O3, a gas mixture of O3 and O2, has been used as a curative agent for the treatment of different diseases over 150 years [27]. For mechanisms, O3 has therapeutic properties such as antimicrobial, anti-inflammatory, modulation of antioxidant defense system and apoptosis [27]. By using O2/O3 mixture to re-infusion of autologous blood slowly, ozonated autohemotherapy (OA) has favorable effects on nervous systems diseases, such as neurologic recovery in spontaneous spinal epidural hematoma [28] and antidepressant effect [29]. OA can also improve the level of brain-derived neurotrophic factor in patients with cognitive impairment [29]. Moreover, various cerebrovascular diseases are important causes in epilepsy, especially ischemic stroke [30], while ozone therapy has a potential role in the treatment of ischemic disorders [31,32]. However, the neurotoxic effects of excess O3 exposure are also well-known. Oxidative stress is involved in many neurodegenerative diseases, in which the over-production of O3 leads to both pathological and functional brain injuries [33]. Moreover, the environmental O3 exposure is associated with a series of adverse effects within the central nervous system, such as decreased cognitive response, decrease in motor activity and headaches, disturbances in the sleep-wake cycle, cell degeneration, and neurochemical alterations [34]. Taken together, the paradoxical effects of O3 may depend on its concentration: while high concentration of O3 may impair nervous system, low level of O3 may act as a protector. However, more studies are needed to further confirm that. Nevertheless, limitations should also be mentioned. First, the study design is ecological in nature, which limits its ability for causal inference. The precise personal exposure level of air pollution may differ based on the location, time spent out side, and other factors. Just like most previous time-series studies, we used the average levels to represent the general exposure levels, which may not accurately reflect individual level associations. Though such ecological study design alone cannot demonstrate causality, it provides results worthy of further research [35]. Second, the data were only from urban areas of one city, which made it difficult to generalize the results to rural areas or to other cities. Third, more individual information such as BMI, smoking habits, drug history were unavailable, so we only performed stratified analyses by the data we could get, including sex and age. Fourth, our health data were collected based on two years of the data from only a single hospital, which may not represent the complete situation. Future work with data based on wider areas and longer periods may be helpful to further clarify the association between air pollution and epilepsy attacks. To establish the causal relationship between air pollutants and epilepsy, further epidemiological designs such as cohort studies can be performed in which personal exposure levels can be monitored with more accurate measures (e.g. personal monitors). Conclusively, we demonstrated that short-term exposure to NO2 and SO2 were associated with increased outpatients for epilepsy. On the contrary, O3 was associated with decreased risk of epilepsy, suggesting that O3 might be a “protective” factor. Our data may contribute to the limited information concerning the effects of air pollution on epilepsy attacks. Supporting Information S1 Fig Partial autocorrelation functions for the residuals from the time series analyses with 7 degrees of freedom per year. Dashed horizontal lines show the test that the apparent autocorrelation is non-zero, suggesting that the basic model may be appropriate. (TIF) Click here for additional data file. S2 Fig Percent change in daily outpatient-visits for epilepsy associated with a 10 μg/m3 (PM10, PM2.5, SO2, NO2, and O3) or 0.1 mg/m3 (CO) increase of air pollutants using temperature lag model. The associations between pollutants and epilepsy were still existing, suggesting that the basic model was steady. (TIF) Click here for additional data file. S1 Table The results of checking different degree of freedoms per year for time trends by Akaka’s Information Criterion (AIC). (DOC) Click here for additional data file. S2 Table Percent change (mean and 95% confidence interval) in daily outpatient-visits for epilepsy associated with a 10 μg/m3 (NO2, SO2, and O3) increase of air pollutants on the concurrent days with different degree of freedoms per year. (DOC) Click here for additional data file. The authors thank Yuming Guo from University of Queensland and Xiaochuan Pan from Peking University for their advice in the preparation of the manuscript. The authors are also grateful to the individuals who provided help in the data collection. ==== Refs References 1 Goodarzi P , Aghayan HR , Soleimani M , Norouzi-Javidan A , Mohamadi-Jahani F , Jahangiri S , et al Stem cell therapy for treatment of epilepsy . Acta Med Iran . 2014 ; 52 : 651 –655 . 25325201 2 Ngugi AK , Bottomley C , Kleinschmidt I , Sander JW , Newton CR . Estimation of the burden of active and life-time epilepsy: a meta-analytic approach . Epilepsia . 2010 ; 51 : 883 –890 . 10.1111/j.1528-1167.2009.02481.x 20067507 3 Ye X , Peng L , Kan H , Wang W , Geng F , Mu Z , et al Acute Effects of Particulate Air Pollution on the Incidence of Coronary Heart Disease in Shanghai, China . PLoS One . 2016 ; 11 : e0151119 10.1371/journal.pone.0151119 26942767 4 Xu Q , Li X , Wang S , Wang C , Huang F , Gao Q , et al Fine Particulate Air Pollution and Hospital Emergency Room Visits for Respiratory Disease in Urban Areas in Beijing, China, in 2013 . PLoS One . 2016 ; 11 : e0153099 10.1371/journal.pone.0153099 27054582 5 Calderon-Garciduenas L , Kulesza RJ , Doty RL , D'Angiulli A , Torres-Jardon R . Megacities air pollution problems: Mexico City Metropolitan Area critical issues on the central nervous system pediatric impact . Environ Res . 2015 ; 137 : 157 –169 . 10.1016/j.envres.2014.12.012 25543546 6 Migliore L , Coppede F . Environmental-induced oxidative stress in neurodegenerative disorders and aging . Mutat Res . 2009 ; 674 : 73 –84 . 10.1016/j.mrgentox.2008.09.013 18952194 7 Kan H , Chen B , Hong C . Health impact of outdoor air pollution in China: current knowledge and future research needs . Environ Health Perspect . 2009 ; 117 : A187 10.1289/ehp.12737 19478975 8 Chen B , Kan H , Chen R , Jiang S , Hong C . Air pollution and health studies in China—policy implications . J Air Waste Manag Assoc . 2011 ; 61 : 1292 –1299 . 22168112 9 Cao J , Xu H , Xu Q , Chen B , Kan H . Fine particulate matter constituents and cardiopulmonary mortality in a heavily polluted Chinese city . Environ Health Perspect . 2012 ; 120 : 373 –378 . 10.1289/ehp.1103671 22389181 10 Lu F , Zhou L , Xu Y , Zheng T , Guo Y , Wellenius GA , et al Short-term effects of air pollution on daily mortality and years of life lost in Nanjing, China . Sci Total Environ . 2015 ; 536 : 123 –129 . 10.1016/j.scitotenv.2015.07.048 26204048 11 Cai J , Zhao A , Zhao J , Chen R , Wang W , Ha S , et al Acute effects of air pollution on asthma hospitalization in Shanghai, China . Environ Pollut . 2014 ; 191 : 139 –144 . 10.1016/j.envpol.2014.04.028 24836410 12 Cakmak S , Dales RE , Vidal CB . Air pollution and hospitalization for epilepsy in Chile . Environ Int . 2010 ; 36 : 501 –505 . 10.1016/j.envint.2010.03.008 20452673 13 Freire C , Ramos R , Puertas R , Lopez-Espinosa MJ , Julvez J , Aguilera I , et al Association of traffic-related air pollution with cognitive development in children . J Epidemiol Community Health . 2010 ; 64 : 223 –228 . 10.1136/jech.2008.084574 19679705 14 Farahani H , Hasan M . Nitrogen dioxide induced changes in level of free fatty acids, triglyceride, esterified fatty acid, ganglioside and lipase activity in the guinea pig brain . J Environ Sci Health B . 1992 ; 27 : 53 –71 . 1556389 15 Farahani H , Hasan M . Effect of NO2 on lipids and lipid peroxidation in the CNS of the guinea-pig . Pharmacol Toxicol . 1990 ; 66 : 146 –149 . 2315266 16 Szyszkowicz M , Rowe BH , Kaplan GG . Ambient sulphur dioxide exposure and emergency department visits for migraine in Vancouver, Canada . Int J Occup Med Environ Health . 2009 ; 22 : 7 –12 . 10.2478/v10001-009-0006-7 19329386 17 Shah AS , Lee KK , McAllister DA , Hunter A , Nair H , Whiteley W , et al Short term exposure to air pollution and stroke: systematic review and meta-analysis . BMJ . 2015 ; 350 : h1295 10.1136/bmj.h1295 25810496 18 Sang N , Hou L , Yun Y , Li G . SO(2) inhalation induces protein oxidation, DNA-protein crosslinks and apoptosis in rat hippocampus . Ecotoxicol Environ Saf . 2009 ; 72 : 879 –884 . 10.1016/j.ecoenv.2008.07.007 18722661 19 Yun Y , Yao G , Yue H , Guo L , Qin G , Li G , et al SO(2) inhalation causes synaptic injury in rat hippocampus via its derivatives in vivo . Chemosphere . 2013 ; 93 : 2426 –2432 . 10.1016/j.chemosphere.2013.08.063 24099899 20 Yao G , Yue H , Yun Y , Sang N . Chronic SO2 inhalation above environmental standard impairs neuronal behavior and represses glutamate receptor gene expression and memory-related kinase activation via neuroinflammation in rats . Environ Res . 2015 ; 137 : 85 –93 . 10.1016/j.envres.2014.11.012 25498917 21 Diffey B . Climate change, ozone depletion and the impact on ultraviolet exposure of human skin . Phys Med Biol . 2004 ; 49 : R1 –11 . 14971768 22 Leon OS , Menendez S , Merino N , Castillo R , Sam S , Perez L , et al Ozone oxidative preconditioning: a protection against cellular damage by free radicals . Mediators Inflamm . 1998 ; 7 : 289 –294 . 9792340 23 Ajamieh H , Merino N , Candelario-Jalil E , Menendez S , Martinez-Sanchez G , Re L , et al Similar protective effect of ischaemic and ozone oxidative preconditionings in liver ischaemia/reperfusion injury . Pharmacol Res . 2002 ; 45 : 333 –339 . 12030798 24 Swiader MJ , Kotowski J , Luszczki JJ . Modulation of adenosinergic system and its application for the treatment of epilepsy . Pharmacol Rep . 2014 ; 66 : 335 –342 . 10.1016/j.pharep.2013.10.005 24905507 25 Mallok A , Vaillant JD , Soto MT , Viebahn-Hansler R , Viart Mde L , Perez AF , et al Ozone protective effects against PTZ-induced generalized seizures are mediated by reestablishment of cellular redox balance and A1 adenosine receptors . Neurol Res . 2015 ; 37 : 204 –210 . 10.1179/1743132814Y.0000000445 25258110 26 Bocci VA . Scientific and medical aspects of ozone therapy. State of the art . Arch Med Res . 2006 ; 37 : 425 –435 . 16624639 27 Sancak EB , Turkon H , Cukur S , Erimsah S , Akbas A , Gulpinar MT , et al Major Ozonated Autohemotherapy Preconditioning Ameliorates Kidney Ischemia-Reperfusion Injury . Inflammation . 2016 ; 39 : 209 –217 . 10.1007/s10753-015-0240-z 26282390 28 Yu L , Lu X , Shi H , Wang Q . Does ozone autohemotherapy have positive effect on neurologic recovery in spontaneous spinal epidural hematoma? Am J Emerg Med . 2014 ; 32 : 949 e941 –942 . 29 Coppola L , Luongo C , Pastore A , Masciello C , Parascandola RR , Mastrolorenzo L , et al Ozonized autohaemotransfusion could be a potential rapid-acting antidepressant medication in elderly patients . Int J Geriatr Psychiatry . 2010 ; 25 : 208 –213 . 10.1002/gps.2322 19521954 30 Berg AT . Seizures and Epilepsy after Ischemic Stroke . Epilepsy Curr . 2003 ; 3 : 129 –130 . 15309055 31 Clavo B , Catala L , Perez JL , Rodriguez V , Robaina F . Ozone Therapy on Cerebral Blood Flow: A Preliminary Report . Evid Based Complement Alternat Med . 2004 ; 1 : 315 –319 . 15841265 32 Clavo B , Suarez G , Aguilar Y , Gutierrez D , Ponce P , Cubero A , et al Brain ischemia and hypometabolism treated by ozone therapy . Forsch Komplementmed . 2011 ; 18 : 283 –287 . 10.1159/000333795 22105041 33 Rivas-Arancibia S , Guevara-Guzman R , Lopez-Vidal Y , Rodriguez-Martinez E , Zanardo-Gomes M , Angoa-Perez M , et al Oxidative stress caused by ozone exposure induces loss of brain repair in the hippocampus of adult rats . Toxicol Sci . 2010 ; 113 : 187 –197 . 10.1093/toxsci/kfp252 19833740 34 Martinez-Lazcano JC , Gonzalez-Guevara E , del Carmen Rubio M , Franco-Perez J , Custodio V , Hernandez-Ceron M , et al The effects of ozone exposure and associated injury mechanisms on the central nervous system . Rev Neurosci . 2013 ; 24 : 337 –352 . 10.1515/revneuro-2012-0084 23585211 35 Morgenstern H . Ecologic studies in epidemiology: concepts, principles, and methods . Annu Rev Public Health . 1995 ; 16 : 61 –81 . 7639884

PMC005xxxxxx/PMC5003347.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 27571518PONE-D-15-5277810.1371/journal.pone.0162080Research ArticleResearch and Analysis MethodsMathematical and Statistical TechniquesStatistical MethodsForecastingPhysical SciencesMathematicsStatistics (Mathematics)Statistical MethodsForecastingEngineering and TechnologyCivil EngineeringTransportation InfrastructureRoadsEngineering and TechnologyTransportationTransportation InfrastructureRoadsPhysical SciencesMathematicsStatistics (Mathematics)Confidence IntervalsComputer and Information SciencesComputer NetworksInternetSocial SciencesSociologySocial ResearchBiology and Life SciencesChronobiologyCircadian RhythmsEngineering and TechnologyEquipmentCommunication EquipmentCell PhonesEngineering and TechnologyCivil EngineeringTransportation InfrastructureRoadsHighwaysEngineering and TechnologyTransportationTransportation InfrastructureRoadsHighwaysPredicting Road Conditions with Internet Search Predicting Road Conditionshttp://orcid.org/0000-0002-3134-6616Askitas Nikolaos * IZA - Institute for the Study of Labor, Bonn, Germany Dovrolis Constantine Editor Georgia Institute of Technology, UNITED STATES Competing Interests: The author has declared that no competing interests exist. Conceptualization: NA. Data curation: NA. Formal analysis: NA. Funding acquisition: NA. Investigation: NA. Methodology: NA. Project administration: NA. Resources: NA. Software: NA. Supervision: NA. Validation: NA. Visualization: NA. Writing - original draft: NA. Writing - review & editing: NA. * E-mail: askitas@iza.org2016 29 8 2016 11 8 e01620804 12 2015 17 8 2016 © 2016 Nikolaos Askitas2016Nikolaos AskitasThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Traffic congestion is an important problem both on an individual and on a societal level and much research has been done to explain and prevent their emergence. There are currently many systems which provide a reasonably good picture of actual road traffic by employing either fixed measurement points on highways or so called “floating car data” i.e. by using velocity and location data from roaming, networked, GPS enabled members of traffic. Some of these systems also offer forecasting of road conditions based on such historical data. To my knowledge there is as yet no system which offers advance notice on road conditions based on a signal which is guaranteed to occur in advance of these conditions and this is the novelty of this paper. Google Search intensity for the German word stau (i.e. traffic jam) peaks 2 hours ahead of the number of traffic jam reports as reported by the ADAC, a well known German automobile club and the largest of its kind in Europe. This is true both in the morning (7 am to 9 am) and in the evening (4 pm to 6 pm). The main result of this paper is then that after controlling for time-of-day and day-of-week effects we can still explain a significant additional portion of the variation of the number of traffic jam reports with Google Trends and we can thus explain well over 80% of the variation of road conditions using Google search activity. A one percent increase in Google stau searches implies a .4 percent increase of traffic jams. Our paper is a proof of concept that aggregate, timely delivered behavioural data can help fine tune modern societies and prompts for more research with better, more disaggregated data in order to also achieve practical solutions. The author received no specific funding for this work. Data AvailabilityThe data is now published and publicly available at http://dx.doi.org/10.15185/izadp.9503.1.Data Availability The data is now published and publicly available at http://dx.doi.org/10.15185/izadp.9503.1. ==== Body Introduction According to the German automobile club ADAC, in 2014 there have been 475,000 traffic jams on German highways which amounted to 960,000 kilometres of jammed traffic. These numbers represent an increase of 14.4% and 15.6% respectively compared to the year before. This is just an instance of a multiyear trend which is only expected to get worse even though the report of the ADAC estimates that most of the current increases are due to progress in the method of documentation of traffic jams. According to the ADAC report (https://www.adac.de/infotestrat/adac-im-einsatz/motorwelt/Staubilanz2014.aspx) these traffic jams amounted to 285,000 lost hours which is 32 years. Adverse road conditions such as traffic congestion are known to cause a host of undesired side effects (see for example [1]). These vary from increased pollution (e.g. gas emissions), energy waste, additional transportation and production costs to waste of labor and delays in product deliveries. Traffic congestion is also known to impact public health negatively ([2]), to cause stress and road rage ([3]) and to even affect the unborn ([4]). If we thought of a city, a region, a country or any other social unit as a large living organism road traffic would be one of its circadian rhythms and traffic jams would be an obstruction to its entrainment. It is hence not surprising that obstructing the smooth flow of traffic sends ripples of negative effects deep into many aspects of socioeconomic life. Understanding, forecasting and preventing adverse road conditions is therefore important for the benefit of the drivers but also obviously for economic reasons. The emergence of traffic jams is a complex matter and may have many causes. Some are simple and have to do with fluctuating degradation of the available infrastructure (e.g. accidents, road constructions, weather and the like) and some are more complex and depend on behavioural, topological and dynamical complexities. Two well known such complex phenomena are discussed in [5], where it is shown that expanding the road network may paradoxically worsen road performance and in [6] where the spontaneous emergence of the so called “phantom jams” are experimentally investigated or in [7] where a theoretical model is discussed. There is an extensive literature both empirical and theoretical which develops “fundamental diagrams of traffic” i.e. studies the relationships between traffic density, velocity and flow (see e.g. [8] and the literature therein). The basic intuition in this type of research is that the higher the traffic density is the more likely it becomes that driver actions are interlinked: high density implies more or less that if one driver breaks then all drivers do so as well with an increasing abruptness. Once a critical density is reached traffic will almost certainly stall. Forecasting and preventing traffic jams is hence a hard problem. If we had a global ledger where future traffic participants would volunteer their itineraries in advance things might be different. We could then get an advance estimate of future traffic and even try to prevent traffic density from reaching critical mass by deploying a kind of leaky bucket algorithm (see [9]) by which we queue departures and still get drivers to their destinations faster than otherwise. Of course such a ledger of itineraries is highly unrealistic but Google search data might be a good proxy. What an individuals searches for in Google reveals something about their state or their intent. If they search for a recipe we know what they might cook, if they search for the bus departure timetable they might take the bus and if they search for information on traffic congestion along a certain highway they are likely to drive on it at a later point in time. The core observation which sparked this paper is that every morning stau searches in Germany peak at 7 am and then they start to dissipate as drivers are being injected into the traffic. Two hours later, at 9:00 am, we have the morning peak of ADAC traffic reports. Similarly at 16:00 hrs every afternoon we observe the search peak and two hours later, at 18:00 hrs, we see a peak of the ADAC traffic reports. Clearly performing a Google search and driving at the same time are mutually exclusive and this is what creates this advance. A future proliferation of mobile applications which allow one to speak searches into a cell phone as well as to graphically get current crowdsourced traffic conditions may diminish the effectiveness of this method but currently such search intensity is a good proxy for the number of people who intent to drive soon. By searching for traffic jam information ahead of driving, drivers provide us with advance warning of their intent to drive. The number of people who will drive every morning has some regularities, like 24-hour or 7-day periodicities, but it also has a certain stochastic component. The latter cannot be forecasted with historical data as it is impossible to predict when some people will be redeployed by their work or have a private emergency and hence change their driving plans. In this paper I propose a formula for estimating the country wide, aggregate number of hourly ADAC traffic jam reports which has time of the day, day of the week and hourly stau search intensity from two hours ago as its explanatory variables. I hence can forecast this crude indicator of road conditions two hours in advance. There are currently several systems which provide live road conditions. They are so called “probe-based” systems or based on fixed measurement points on highways and may use crowdsourcing of volunteers, leveraging the fact that they are roaming members of a digital network and are equipped with Global Positioning Systems able to compute and transmit position, direction and velocity of motion. Such systems include Google Traffic, a feature of Google Maps and the crowdsourcing system of INRIX. More providers can be found here: https://en.wikipedia.org/wiki/Traffic_reporting. Lastly the literature cited in [10] covers a good portion of the literature which uses geo-located data to study mobility questions. To my knowledge there is no system available which can have as much as two hours advance on the emergence of such road conditions and does so not by using an algorithm trained on past values of road conditions but by using behavioural measurements guaranteed to occur before the emergence of road congestion. Such current “floating car data” systems (e.g. [11]) can predict at most 15-30 minutes ahead. My target variable is a crude proxy for road conditions but this paper is a proof of concept that with better target data Google Search can be a powerful predictive tool of traffic conditions. In fact this paper is among very few in the literature of Google Search data which uses hourly data, has a clear behavioural foundation and detects two phenomena with a causal phase difference. Moreover it suggests that the Google Traffic team ought to work closer with the Google Trends team to better forecast traffic. Two hours advance notice is an enormous amount of time when it comes to traffic jams. Materials and Methods ADAC traffic jam data On September 28 at 14:59 hrs I started collecting ADAC traffic jam data, programmatically every five minutes, from the website (https://www.adac.de/reise_freizeit/verkehr/aktuelle_verkehrslage/default.aspx) of the German automobile club ADAC. Prior to resorting to data scraping I had contacted ADAC about access to their past data but unfortunately I could not elicit a response. The data collection for the purpose of this paper ends on November 14 2015 at 13:59 hrs. Each observation consists of a timestamp, region name and current count. I use the countrywide aggregate, average hourly count as my target time series. Fig 1 depicts some diagnostics for this data. 10.1371/journal.pone.0162080.g001Fig 1 ADAC traffic jams. Natural logs of hourly ADAC traffic jam reports (top left), its autocorellations (top right), spectral density (bottom left) and histogram of percentage changes (bottom right). Data Source: ADAC (adac.de) and own calculations. As expected the series exhibits a significant amount of autocorellation, a strong circadian rhythm with day-of-week and hour-of-day fixed effects and, due to non-linearities and endogeneities, a leptokurtic and fat tailed, unimodal distribution of hourly percentage changes. The mean hourly number of traffic jams in that period is 177 with a standard deviation of 80, a minimum of 80 and a maximum of 486. The data on the numbers of traffic jams on the ADAC website is based on reports from volunteers who call in to report the traffic congestion they are currently in. In that respect it is the “low tech” version of, so called, probe-based traffic reporting which use floating car data, whose more modern high tech variant is done with roaming, networked GPS capable smartphones. This means that our target variable could very well have been the data from systems such as Google Traffic, INRIX or any other system which uses floating car data to reflect current road conditions. The fact that we use the ADAC data is simply a result of data availability. Google Trends Data The Google Trends data are hourly data of searches containing the word stau which is German for traffic jam. They are obtained from the Google Trends data provisioning system which can be found by pointing one’s browser at the Google Trends website (https://www.google.com/trends/). I am discussing some aspects of Google Trends data in [12]. The reader who wants to use this data may also want to consult a more practical data description in [13]. Before moving on to discussing the hourly data I dissect such searches in order to provide support for my identification strategy. In this way I demonstrate that assuming that the searches are made by drivers is not entirely unfounded. Fig 2 shows weekly searches which contain the word “stau” since 2004 together with a reduced series. The enormous spike on December 2010 is due to an extraordinary traffic jam in Germany due to snowfall. The trend in this figure is in agreement with the annual ADAC traffic jam reports. 10.1371/journal.pone.0162080.g002Fig 2 Google Searches for “stau.” The reduced series are searches containing the word stau without containing any of: nrw, a7, a2, a3, a1, wdr, a8, a5, aktuell, autobahn, a9, info, swr3, bayern, hamburg, a4, staumelder, adac, berlin, a6, verkehr, ffh, hessen, köln, münchen, swr, a81, a61, deutschland. A 60% of the total stau search volume is accounted for by these words. Data Source: Google Trends (www.google.com/trends) and own calculations. The reduced series is constructed by subtracting those searches which contain any of the top 30 additional words. We see that these 30 words account for 60% of the total volume on the average. These keywords are city or region names (NRW, Bayern, Hamburg, Berlin Hessen, Köln, München, Deutschland) or names of highways (A7, A2, A3, A1, A8, A5, A9, A4, A6, A81, A61). Such searches even reveal which highway the driver will be driving on. We also have radio stations (WDR, FFH, SWR) and other websites such as the one of ADAC. Although we cannot subtract them from the volume due to limitations from the Google Trends data provisioning system we know a further set of keywords which are contained in the reduced series. These keywords further support the thesis that these are prospective drivers. They are: a45, stuttgart, a40, elbtunnel, frankfurt, a44 stau, meldungen, hr3, a14, app, hannover, online, a24, niedersachsen, ndr, bremen, a46, a10, staumeldungen, aktuelle, karlsruhe, wdr2, a31, a57, a43, baden württemberg, nürnberg, bw, nachrichten, antenne, dresden, b10 stau, a7, a96, österreich, sachsen, bonn,a5. The highway numbers contained as well as the city and region names imply that our method can predict not just the nationwide aggregate road conditions but also regional conditions. The largest portion of stau searches are those that also contain the string “NRW” which stands for Nord-Rhein-Westfalen the region with the densest highway network in Germany as can be seen by consulting www.autobahnatlas-online.de. Table 1 depicts the towns from where searches for four major highways originate. 10.1371/journal.pone.0162080.t001Table 1 Origins of four highway-specific Google “stau” searches. stau A1 stau A2 stau A3 stau A7 Glauchau Glauchau Glauchau Glauchau Bremen Gütersloh Oberhausen Flensburg Oldenburg Bielefeld Düsseldorf Drochtersen Vechta Hanover Cologne Norderstedt Osnabrück Brunswick Koblenz Hamburg Münster Wolfenbüttel Aschaffenburg Hanover Dortmund Magdeburg Würzburg Göttingen Erlangen Kassel For the highways A1, A2, A3, A7 searches for stau A1, stau A2, stau A3 and stau A7 come from towns along the respective highways. The reader may easily verify that the towns where these searches originate from are situated along the respective highways thus providing support for the claim that such data can be used for regional traffic forecasting as they can be thought of as early itinerary announcements of sorts. An interesting accidental observation is striking from the table. The town Glauchau appears in the top position for all highways. In other words this town in an exception to the rule of searches for a highway coming from along the highway. This leads to the conjecture that something beyond and above end user behaviour happens there. Very close to the town is Volkswagenwerk Zwickau the well known automaker. It would appear as though this automaker may already be looking into this type of data. To recap, by looking at Google stau searches not only do we hope to parsimoniously identify soon-to-be members of traffic but furthermore to locate the highway they will soon be on. The characteristics of the hourly series of Google searches for stau can be seen in Fig 3. 10.1371/journal.pone.0162080.g003Fig 3 Hourly Google Searches for “stau.” The natural logs of the hourly stau search intensity (top left), its autocorellations (top right), spectral density (bottom left) and histogram of percentage changes (bottom right). Data Source: Google Trends (www.google.com/trends) and own calculations. Results Informed by the observation in Fig 4 I estimate five “Granger causality” ([14]) models: St=αSt−2+β+ϵt(M1) St=αGt−2+β+ϵt(M2) St=αD+βH+γ+ϵt(M3) St=αD+βH+γSt−2+δSt−24+ϵ+ϵt(M4) St=αD+βH+γGt−2+δ+ϵt(M5) where St is the natural log of the number of ADAC traffic jam reports at time t (hour), D is the day of the week (with values 0, 1, 2, 3, 4, 5 where 0 = Sunday), H is the hour of the day (with values 0−24) and Gt−2 is the natural log of the Google search intensity for “stau” at time t − 2. We estimate the parameters α, β, γ, δ, ϵ; ϵt is the error term. I use partly overlapping letters for the intercepts in the models as there is no danger of confusion. The regressions are restricted to the fixed effects from days and hours that are statistically significant for at least one of the models. 10.1371/journal.pone.0162080.g004Fig 4 Google Searches for “stau” vs ADAC reports. A cross correlogram between the hourly number of ADAC traffic jam reports and the hourly Google search intensity for stau, based on the entire observation time interval of 51 days, establishes that Google search has a two hour advance on road conditions. Data Source: Google Trends, ADAC and own calculations. Table 2 summarises the results of the five OLS regressions with each column being one model depicting its estimated intercepts and their statistical significance together with the model’s adjusted R2, Akaike Information Criterion and Root Mean Squared Error. The first two models M1 and M2 are the simplest possible models where we respectively regress the natural log of the number of ADAC reports on the (natural log of) its second lag and the (natural log of) the second lag of Google search intensity for stau, the model M3 explores the extend to which day-of-week and hour-of-day fixed effects suffice to explain the variation of the number of ADAC traffic jam reports. Model M4 is the full blown autoregressive model using fixed effects together with two hour and twenty-four hour lags. This model is the one which is used as the benchmark to beat and is meant to demonstrate the extend to which “historical” values of the series can predict it. Finally model M5 includes day-of-week and hour-of-day fixed effects together with the second lag of the Google stau searches. Clearly Eq M5 is the best model on all counts: lowest AIC, smallest RMSE and explains significantly more of the variation (85.1%) of our target variable than the second best model (67.3%). Since clearly Eq M2 beats Eqs M1 and M2 is the best model we have established our claim that the two hour lagged Google searches capture much of the stochastic behaviour of daily mobility which would otherwise remain elusive. Notice that the “historical” baseline model M4 ought to be similar to the prediction methods used by current systems using floating car data. 10.1371/journal.pone.0162080.t002Table 2 Forecasting the number of ADAC traffic reports. M1 M2 M3 M4 M5 St St St St St coef./p-value coef./p-value coef./p-value coef./p-value coef./p-value St−2 .659*** .326*** (.000) (.000) St−24 .244*** (.000) Gt−2 .368*** .415*** (.000) (.000) D = 0 .000 .000 .000 (.) (.) (.) D = 1 .433*** .337*** .268*** (.000) (.000) (.000) D = 2 .465*** .240*** .337*** (.000) (.000) (.000) D = 3 .527*** .287*** .382*** (.000) (.000) (.000) D = 4 .559*** .301*** .399*** (.000) (.000) (.000) D = 5 .480*** .232*** .154*** (.000) (.000) (.000) D = 6 .076* –.056 –.051* (.033) (.110) (.023) H = 7 .000 .000 .000 (.) (.) (.) H = 8 .437*** .320*** .253*** (.000) (.000) (.000) H = 9 .565*** .345*** .342*** (.000) (.000) (.000) H = 10 .402*** .082 .304*** (.000) (.099) (.000) H = 11 .294*** –.037 .344*** (.000) (.479) (.000) H = 12 .293*** .016 .424*** (.000) (.739) (.000) H = 13 .323*** .075 .435*** (.000) (.114) (.000) H = 14 .383*** .121* .423*** (.000) (.011) (.000) H = 15 .469*** .177*** .466*** (.000) (.000) (.000) H = 16 .562*** .230*** .492*** (.000) (.000) (.000) H = 17 .706*** .309*** .531*** (.000) (.000) (.000) H = 18 .778*** .332*** .552*** (.000) (.000) (.000) H = 19 .651*** .192*** .448*** (.000) (.001) (.000) const. 1.734*** 4.631*** 4.509*** 1.974*** 3.985*** (.000) (.000) (.000) (.000) (.000) Adj. R2 .434*** .631*** .612*** .693*** .851*** AIC 531.027 49.119 –57.311 –185.875 –639.008 RMSE .306 .247 .227 .204 .141 No. of cases 1126.000 1126.000 611.000 598.000 609.000 *** p < 0.01, ** p < 0.05, * p < 0.1 Robustness Tests My analysis has demonstrated that Google Search intensity for stau contains advance information on the number of traffic jams two hours before they occur. One can clearly not do much better than that since I only have a crude aggregate measure of road conditions. Better data would most likely allow better models to come to fruition. In this chapter I would like to perform some robustness tests to better support the validity of my results. I do so by offering scatter plots for data and their fit for the Google data model vs its nearest competitor models, by performing a rolling forecasting exercise and by performing a bootstrap regression to demonstrate that confidence intervals are robust. Fig 5 contains scatter plots and regression lines for the naive fixed effects model M3, the historical model M4 and the mode M5 which is informed by the Google Trends data. Clearly Eq M5 fits the data best. It becomes apparent that higher degree terms are present in model M5. This is not surprising since the probability of pairwise interaction increases with the number of active drivers but reaches a saturation point after a certain critical value is reached. This is depicted in the bottom right of Fig 5 which shows a third degree polynomial fit. 10.1371/journal.pone.0162080.g005Fig 5 Comparing Eqs M3, M4 and M5. Clearly the regression line of Eq M5 captures the data better than that of Eqs M4 and M3. A higher degree term is apparent (bottom right). Data Source: Google Trends (www.google.com/trends), ADAC and own calculations. A practitioner whose job is to analyse and predict road conditions can and will update his or her model as data becomes available. Since Google Trends data are updated hourly so can the prediction model using these data be continuously re-estimated. I performed a rolling forecast robustness test for models M4 and M5 by estimating the models up to 7 am for all but the last few days (for a total of 10 repetitions) and predicting the 9 am peak of the ADAC reports and similarly for the afternoon estimating up to 16:00hrs and predicting the 18:00hrs peak. I found that the RMSE in the morning trials is .36 for model M4 and .23 for model M5 whereas in the afternoon it is .22 and .17 respectively. In other words Eq M5 reduces the error by 36% in the morning and by 23% in the afternoon compared to the historical model. Finally I used bootstrap regression to demonstrate the robustness of the confidence intervals when we forecast traffic jams with Google data. Since there is not enough data to do bootstrapping with day-of-week and hour-of-day effects I opted for a reduction of the data and two simpler models which on the one hand allow Google data to demonstrate their predictive power and on the other hand the bootstrapping technique to work. The first model regresses the daily number of ADAC report at 9:00 am on the Google search intensity for stau at 7:00 am. The second model similarly regresses the ADAC reports at 18:00 hrs on the Google search intensity for stau at 7:00 am and at 16:00 hrs. In both models we have a total of 51 data points. The models can be written as follows: Sd9=αSd7+β+ϵd(Mm) Sd18=αSd7+βSD16+γ+ϵd,(Ma) where α, β, γ are the intercepts to be estimated, ϵd is the error, d = 1, …, 51 counts the days in the dataset and Sdx for x = 7, 9, 16, 18 is the number of ADAC traffic jams at x hrs on day d. For the morning model Mm we get α = .66 and β = 4. Their 95% confidence intervals are [.56, .76] and [3.82, 4.25] respectively and they become [.59, .72] and [3.89, 4.17] respectively after bootstrapping. The model explains 77% of the observed variation. For the afternoon model Ma we estimate the intercepts to be α = .43, β = .28 and γ = 4.1. Their respective confidence intervals are [.31, .56], [.19, .38] and [3.8, 4.36] and they become [.32, .55], [.22, .35] and [3.85, 4.32] under bootstrapping. The model explains 72% of the observed variation. Discussion Google search intensity for the word stau is an elegant and parsimonious way to capture driving intent in the near future in Germany and hence to predict road conditions two hours in advance by a method more informed than one which simply uses “historical data” or calendar fixed effects. Even after taking calendar fixed effects into account there is still a significant amount of stochasticity in the number of cars that drive off every day. Whether redeployed employees, private persons responding to unforeseen events or other stochastic changes there is a large amount of variation historical data cannot possibly capture. Google searches for stau are a good proxy for drivers telling us they will drive and are hence as close as we can currently come to having a public ledger of future traffic participants. I have demonstrated this by predicting the countrywide number of ADAC traffic jam reports, an admittedly crude proxy for road conditions. Better target data is needed and more research is necessary to operationalise the results of this paper which is hence only a proof of concept. Since Google searches for stau are often accompanied by city, region or highway information access to better data such as the floating car data of Google Traffic or other traffic jam information providers (coupled with geolocating the IP address from which the search is initiated) could allow one to build predictive systems and in fact even preventive systems. As a consequence Google Trends may predict Google Traffic and their combination might evolve into a prediction and prevention system. This paper is another indication that social science in the upcoming future will indeed look and feel more and more like “doing physics with particles that have feelings” (“Imagine how much harder physics would be if electrons had feelings!”—Richard Feynman, Caltech graduation ceremony). I would like to thank the editor and three anonymous referees for remarks and suggestions which improved the exposition in this paper. ==== Refs References 1 Duranton G , Turner MA . The Fundamental Law of Road Congestion: Evidence from US Cities . American Economic Review 2011 ;101 (6 ):2616 –2652 . 10.1257/aer.101.6.2616 2 Levy JI , Buonocore JJ , Von Stackelberg K . Evaluation of the public health impacts of traffic congestion: a health risk assessment . Environmental health 2010 ;9 (1 ):1 –12 . 10.1186/1476-069X-9-65 20064246 3 Hennessy DA , Wiesenthal DL . Traffic congestion, driver stress, and driver aggression . Aggressive behavior 1999 ;25 (6 ):409 –423 . 10.1002/(SICI)1098-2337(1999)25:6<409::AID-AB2>3.0.CO;2-0 4 Currie J , Walker R . Traffic congestion and infant health: Evidence from E-ZPass . American Economic Journal: Applied Economics 2011 ;3 (1 ):65 –90 . 5 Braess D , Nagurney A , Wakolbinger T . On a paradox of traffic planning . Transportation science 2005 ;39 (4 ):446 –450 . 10.1287/trsc.1050.0127 6 Sugiyama Y , Fukui M , Kikuchi M , Hasebe K , Nakayama A , Nishinari K , Tadaki SI , Yukawa S . Traffic jams without bottlenecks—experimental evidence for the physical mechanism of the formation of a jam . New Journal of Physics 2008 ;10 (3 ):033001 10.1088/1367-2630/10/3/033001 7 Flynn MR, Kasimov AR, Nave JC, Rosales RR, Seibold B. On “jamitons”, self-sustained nonlinear traffic waves. arXiv preprint arXiv:0809.2828. 2008 Sep 17. 8 Helbing D . Derivation of a fundamental diagram for urban traffic flow . The European Physical Journal B 2009 ;70 (2 ):229 –241 . 10.1140/epjb/e2009-00093-7 9 Kulkarni VG . Fluid models for single buffer systems. Frontiers in queueing: Models and applications in science and engineering . CRC Press : Boca Raton 1997 ;321 –338 . 10 Lenormand M , Tugores A , Colet P , Ramasco JJ . Tweets on the Road . PloS One . 2014 8 20 ;9 (8 ):e105407 10.1371/journal.pone.0105407 25141161 11 De Fabritiis C, Ragona R, Valenti G. Traffic estimation and prediction based on real time floating car data. 2008 11th International IEEE Conference on Intelligent Transportation Systems 2008 Oct 12:197–203. 12 Askitas N . Google search activity data and breaking trends . IZA World of Labor 2015 : 206 . 13 Askitas N . Trend-Spotting in the Housing Market . Cityscape: A Journal of Policy Development and Research—U.S. Department of Housing and Urban Development—Office of Policy Development and Research 2016 ;18 (2 ):185 –198 . 14 Granger CW . Investigating causal relations by econometric models and cross-spectral methods . Econometrica: Journal of the Econometric Society 1969 8 ;37 (3 ):424 –438 . 10.2307/1912791

PMC005xxxxxx/PMC5003348.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757097110.1371/journal.pone.0162102PONE-D-16-10146Research ArticleMedicine and Health SciencesVascular MedicineBlood PressureBiology and Life SciencesBiochemistryLipidsCholesterolMedicine and Health SciencesEndocrinologyEndocrine DisordersDiabetes MellitusMedicine and Health SciencesMetabolic DisordersDiabetes MellitusMedicine and Health SciencesVascular MedicineBlood PressureHypertensionBiology and Life SciencesAnatomyBody FluidsBloodBlood PlasmaMedicine and Health SciencesAnatomyBody FluidsBloodBlood PlasmaBiology and Life SciencesPhysiologyBody FluidsBloodBlood PlasmaMedicine and Health SciencesPhysiologyBody FluidsBloodBlood PlasmaMedicine and Health SciencesHematologyBloodBlood PlasmaMedicine and health sciencesDiagnostic medicineDiabetes diagnosis and managementHbA1cBiology and life sciencesBiochemistryProteinsHemoglobinHbA1cPeople and PlacesGeographical LocationsAsiaSingaporeBiology and Life SciencesBiochemistryLipidsLipid ProfilesAge-Related Changes in the Cardiometabolic Profiles in Singapore Resident Adult Population: Findings from the National Health Survey 2010 Cardiometabolic Profiles of an Asian PopulationLoh Tze Ping 12Ma Stefan 2Heng Derrick 3Khoo Chin Meng 456*1 Department of Laboratory Medicine, National University Hospital, 5 Lower Kent Ridge Road, 119074 Singapore2 Epidemiology & Disease Control Division, Ministry of Health, 16 College Road, 169854 Singapore3 Public Health Group, Ministry of Health, 16 College Road, 169854 Singapore4 Department of Medicine, National University Hospital, 5 Lower Kent Ridge Road, 119074 Singapore5 Department of Medicine, Yong Loo Lin School of Medicine, National University of Singapore, 10 Medical Drive, 117597 Singapore6 Cardiovascular and Metabolic Disorders Program, Duke-NUS Graduate Medical School, 8 College Road, 169857 SingaporeNorata Giuseppe Danilo EditorUniversita degli Studi di Milano, ITALYCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: TPL SM DH CMK. Data curation: TPL SM DH CMK. Formal analysis: TPL CMK. Funding acquisition: SM DH. Investigation: SM DH. Methodology: SM DH. Project administration: TPL SM DH CMK. Resources: TPL SM DH CMK. Software: SM DH. Supervision: TPL SM DH CMK. Validation: SM DH. Visualization: TPL SM DH CMK. Writing – original draft: TPL SM DH CMK. Writing – review & editing: TPL SM DH CMK. * E-mail: cmkhoo11@gmail.com29 8 2016 2016 11 8 e016210210 3 2016 17 8 2016 © 2016 Loh et al2016Loh et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.We describe the centile trends of the blood pressure, glycemia and lipid profiles as well as renal function of a representative population who participated in the Singapore National Health Survey in 2010. Representative survey population was sampled in two phases, first using geographical/ residential dwelling type stratification, followed up ethnicity. 2,407 survey participants without any self-reported medical or medication history for diabetes mellitus, hypertension and dyslipidemia were included in this analysis. All biochemistry analyses were performed on Roche platforms. After excluding outliers using Tukey's criteria, the results of the remaining participants were subjected to lambda-mu-sigma (LMS) analysis. In men, systolic blood pressure increased linearly with age. By contrast, an upward inflection around late 40s was seen in women. The diastolic blood pressure was highest in men in the late 30s-50s age group, and in women in the late 50s-60s age group. All glycemia-related parameters, i.e. fasting and 2-hour plasma glucose and HbA1c concentrations increased with age, although the rate of increase differed between the tests. Total cholesterol and LDL-cholesterol concentrations increased with age, which became attenuated between the early 30s and late 50s in men, and declined thereafter. In women, total cholesterol and LDL-cholesterol concentrations gradually increased with age until late 30s, when there is an upward inflection, plateauing after late 50s. Our findings indicate that diagnostic performance of laboratory tests for diabetes may be age-sensitive. Unfavourable age-related cardiovascular risk profiles suggest that the burden of cardiovascular disease in this population will increase with aging population. The authors received no specific funding for this work. Data AvailabilityThe data contained in this submission belongs to the Ministry of Health and is not available for public access under local regulations. Interested party can contact Dr. Stefan Ma (Stefan_MA@moh.gov.sg).Data Availability The data contained in this submission belongs to the Ministry of Health and is not available for public access under local regulations. Interested party can contact Dr. Stefan Ma (Stefan_MA@moh.gov.sg). ==== Body Introduction Cardiovascular disease is one of the major causes of morbidity and mortality globally. It is important to understand the trends and prevalence of its risk factors in the general population for current and future public healthcare planning. A National Health Survey (NHS) was conducted by the Ministry of Health in 2010 to obtain a representative view of the general health of Singapore resident adults [1]. In this study, we described the age-related changes of the blood pressure, glycemia and lipid profiles as well as renal function of a representative population who participated in NHS 2010 and had no self-reported history of these chronic diseases. Materials and Methods Study subjects The data included this study were derived from the cross-sectional NHS conducted between 17 March 2010 and 13 June 2010. Detailed description of the survey design can be found in the official report [1], and received institutional ethics review board approval (Medical & Dental Board, Health Promotion Board, ref: 005/2009). The participants provided written informed consent for further analysis of the collected data. Briefly, the sampling of the survey participants was performed in two phases. In phase 1, geographical zones and residential dwelling units were stratified and selected to yield a representative dwelling type distribution. In phase 2, a random sample of 7,696 individuals was selected from households identified in phase 1. Disproportionate stratified sampling was used to ensure sufficient sample size for reliable prevalence estimates of the minority ethnic groups such that the sample composed of 30% Chinese, 30% Malays, 30% Indians and 10% others. After excluding ineligible individuals for reasons such as pregnancy, recent childbirth or institutionalisation, death and overseas sojourn, 4,337 out of 7,512 eligible individuals aged 18 to 79 years participated in the survey (representing a participation rate of 57.7%). Of the survey participants, 2,407 individuals aged 18 to 79 years without any self-reported medical or medication history for diabetes mellitus, hypertension and dyslipidemia were included in our study (S1 Fig). Blood pressure measurement Blood pressure (BP) was measured using mercury sphygmomanometer after adequate resting. Two BP readings were taken 30 seconds apart and averaged. Laboratory analysis All blood samples were collected after an overnight fasting of at least ten hours, using standard phlebotomy procedure. The oral glucose tolerance test (OGTT) was performed by administering 75g of glucose (Trutol), and measurement of the plasma glucose concentration was repeated two hours later. Other laboratory parameters measured included glycated haemoglobin (HbA1c), fasting total cholesterol, high-density lipoprotein cholesterol (HDL), direct low-density lipoprotein cholesterol concentrations (LDL) and creatinine. The estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation [2]. Statistical analysis Tukey's criteria were applied at 10-year intervals to identify outlying values, defined as any value lying below the first quartile value minus 3 times the interquartile range, or above the third quartile value plus 3 times the inter-quartile range [3]. Outlying values identified in this manner were removed from subsequent analysis. The remaining data were analysed using the LMS Chartmaker Light software (Medical Research Council, Cambridge, United Kingdom) to fit smoothed centile curves to reference data. Smoothed centile lines were generated for the 3rd, 10th, 25th, 50th, 75th, 90th and 97th percentiles. Details of this method have been described previously [4]. Results The average and standard deviation of the blood pressure, glycemia and lipid profiles as well as renal function of the male and female subjects are summarised in Table 1. The age distribution of the remaining subjects is shown in Table 2. The distribution of the laboratory parameters and BP measurements is shown in Fig 1. The participants included in this study did not have past medical or medication history for the diabetes mellitus, hypertension, dyslipidaemia or chronic kidney disease. Hence, those participants with parameters above the disease thresholds are considered undiagnosed subjects. 10.1371/journal.pone.0162102.t001Table 1 The average and standard deviation of the blood pressure, glycemia and lipid profiles as well as renal function of the male and female subjects. Males Females t-test Risk factors Mean SD Mean SD p-values Systolic blood pressure, mmHg 116 13 109 14 <0.0001 Diastolic blood pressure, mmHg 73 10 69 10 <0.0001 Pulse pressure, mmHg 43 41 12 11 <0.0001 Fasting plasma glucose, mmol/L 5.2 0.5 5.1 0.5 <0.0001 2-hour post-oral glucose tolerance test, mmol/L 6.2 1.9 6.3 1.8 0.12 HbA1c, % 5.7 0.4 5.7 0.3 0.02 Total cholesterol, mmol/L 5.30 0.96 5.18 0.97 0.003 Low density lipoprotein cholesterol, mmol/L 3.37 0.86 3.12 0.85 <0.0001 High density lipoprotein cholesterol, mmol/L 1.27 0.32 1.52 0.37 <0.0001 Triglycerides, mmol/L 1.32 0.63 1.01 0.50 <0.0001 Estimated glomerular filtration rate, mL/min/1.73 m2 98 15 107 15 <0.0001 10.1371/journal.pone.0162102.t002Table 2 The age distribution of the subjects included in the analysis. Age (in years) Male Female Total <30 273 363 636 30–40 316 370 686 40–50 288 341 629 50–60 146 163 309 >60 74 73 147 Total 1097 1310 2407 10.1371/journal.pone.0162102.g001Fig 1 Centile charts of cardiometabolic parameters of subjects who participated in the Singapore National Health Survey 2010 and had no history of chronic illnesses. The broken lines represent (from bottom to top), the 3rd, 10th, 25th, 50th, 75th, 90th, 97th percentile lines of the survey population. The solid horizontal lines represent limits of normality or desired targets. Older men and women tend to have higher BP than their younger counterparts. Systolic BP increased linearly with age in men. In women, a similar relationship was observed but with an upward inflection around late 40s. As such, systolic hypertension defined as systolic BP of 140 mmHg and above was present in 3% of men at the age of 38 years, and 10% at the age of 58 years. In comparison, similar diagnosis was present in 3% of women at older age of 48 years, and 10% at same age of 58 years. The diastolic BP was highest in men in their late 30s-50s, and in women in their late 50s-60s. However, the diastolic BP declined in older individuals. All glycemia-related parameters, i.e. fasting and 2-hour plasma glucose and HbA1c concentrations, increased with age. Approximately 3% of the population had impaired fasting glucose (defined as having fasting plasma glucose of 6.0–6.9 mmol/L) at the age of 28 years, and more than 10% at the age of 40 years and above. Diabetes mellitus, defined as having fasting plasma glucose of 7.0 mmol/L and above, was present in 3% to 10% of the population around the age of 48 years. More than 10% of the population had impaired glucose tolerance, defined as having 2-hour plasma glucose of 7.8–11.0 mmol/L, by the age of 30 years. Diabetes mellitus, defined as having 2-hour plasma glucose of 11.1 mmol/L and above, was present in 3% of the population by the age of 30 years, and more than 10% around the age of 48 years. Using HbA1c of 6.5% and above as the diagnostic threshold, diabetes mellitus was present in 3% of the population by age of 40 years, and between 3–10% by the age of 50 years. Total cholesterol and LDL-cholesterol concentrations increased with age, which became attenuated between the early 30s and late 50s in men, and declined thereafter. In women, total cholesterol and LDL-cholesterol concentrations gradually increased with age until late 30s, when there was an upward inflection, plateauing after late 50s. Hypercholesterolemia defined as having fasting total cholesterol of 5.2 mmol/L and above was present in more than 50% of our study subjects around age of 40 years. In both men and women, HDL-cholesterol showed a U-shaped trend with age, where higher HDL-concentrations were observed in both younger and older subjects. The eGFR declined continuously with age with a steeper gradient observed in men. As such, more than 10% of the men had stage 3 chronic kidney disease (i.e. eGFR <60 mL/min/1.73 m2) by late 70s. Discussion Based on the results from NHS 2010, the cardiometabolic health profile of apparently healthy adults living in Singapore deteriorated with age. The systolic blood pressure, fasting and 2-hour plasma glucose, and LDL-cholesterol concentrations were higher in older men and women. In addition, there were gender differences in the age-related metabolic profiles, where systolic blood pressure, total and LDL-cholesterol concentrations in women were stable during the reproductive age, and only increased rapidly when they reached menopausal age. Systolic BP increased with age but was attenuated in women during reproductive age, when female sex hormones such as estrogen and progesterone exert favourable effect on the vascular tone and renin-angiotensin system [5]. In contrast to systolic BP, we found that the diastolic BP was lower in older subjects, which could be due to increased large artery stiffness with increasing age [6,7]. Age-related stiffening of large arteries is associated with a decreased capacity of the elastic reservoir. During systole, greater proportion of stroke volume is being delivered to the periphery, reducing retention within the large arteries. At the beginning of diastole, the reduced residual blood volume and elastic recoil of the large arteries cause the diastolic pressure to decline [6,7]. The result of higher systolic but lower diastolic blood pressure is higher pulse-pressure with older age. In epidemiological studies, age-related reduction in diastolic blood pressure or higher pulse pressure has been shown to predict cardiovascular disease [6–8]. The reason for the increase in the fasting and 2-hour plasma glucose with age is likely due to an increase in the peripheral insulin resistance [9–11], age-related deterioration in β-cell function, physical inactivity, sarcopenia and obesity [12–14]. HbA1c also increases at a rate of 0.05–0.1% (0.55–1.09 mmol/mol) per decade, which is independent of the fasting and oral glucose tolerance [11]. The diagnostic specificity of HbA1c in detecting hyperglycaemia reduces with age. As a result, there have been calls to consider age-dependent hyperglycaemia and HbA1c diagnostic cutoff to maintain the diagnostic performance of this test [9,11]. In our study, we observed that total cholesterol concentration was lower in older men but higher in older women, and HDL-cholesterol concentration was higher in older subjects, which are consistent with previous cross-sectional and longitudinal studies [15–19]. These changes may be hormonally driven [20]. Total cholesterol rises steeply around menopause in women [21,22], and is primarily determined by the increase in LDL-cholesterol concentration. Lower total cholesterol concentration in older individuals could be contributed by an age-related reduction in the food intake, cholesterol absorption, in cholesterol synthesis, and low LDL apo-B transport [17,19]. Interestingly, it has been suggested that lower total cholesterol concentration indicates frailty and can predict functional decline in older women [23]. Some of the observed changes in lipid profiles may be related to the change in sex hormone throughout adult life, where free testosterone decreases HDL in men while estradiol decreases LDL in women. We observed that the eGFR was negatively associated with age in a near linear manner, with a steeper decline in men than women. This could be contributed by longer exposure to higher blood pressure in men throughout adulthood. The change in eGFR has been shown to be negatively associated with systolic BP, smoking, fibrinogen, and albumin/creatinine ratio [24,25]. On the other hand, high alcohol consumption in men and high physical activity in women were positively associated with eGFR.24 Whether the decline in the eGFR represents “normal” aging and its implication for diagnosis, progression and prognosis of chronic kidney and related diseases is still a subject of great debate [26,27]. The lower eGFR in the older population also indicate that they are more prone to acute kidney injury during an acute illness or dehydration. It is important to bear in mind the cross-sectional nature of this study when interpreting any observed changes in the risk factors relates to the static population. These may give rise to potential confounding factors such as survival bias, where subjects with favourable longevity/survival factors (e.g. genotype or lifestyle) are over-represented in the study population, which may skew the data. A longitudinal study design would negate this limitation. Moreover, the exclusion of subjects with self-reported medical or medication history for diabetes mellitus, hypertension and dyslipidemia may result in underestimation of the trends observed. Another interesting future direction for research would be to examine race-related differences in these cardiovascular risk profiles, which was not assessed here due to the relatively small sample size. In summary, our findings of unfavourable age-related cardiovascular risk profiles suggest that the burden of cardiovascular disease will increase as the Singapore resident population ages. The present data supports advocating health screening beginning at 40 years of age. However, further analysis on cost-effectiveness should be undertaken. Healthcare policy that emphasizes on primary prevention and early detection and optimal treatment of hypertension, diabetes and dyslipidemia will help to mitigate the increase of cardiovascular disease in the local population. Supporting Information S1 Fig The number and characteristics of the subjects excluded from the analysis. (TIFF) Click here for additional data file. ==== Refs References 1 Ministry of Health (Singapore) . National Health Survey 2010 report . Singapore : Epidemiology & Disease Control Division, Ministry of Health , 2011 Available at https://www.moh.gov.sg/content/dam/moh_web/Publications/Reports/2011/NHS2010%20-%20low%20res.pdf (accessed 5 December 2014). 2 Levey AS , Stevens LA , Schmid CH , Zhang YL , Castro AF 3rd, Feldman HI , et al A new equation to estimate glomerular filtration rate . Ann Intern Med 2009 ;150 :604 –12 . 19414839 3 Tukey JW . Mathematics and the picturing of data . Proc Int Congress Math 1974 ;2 :523 –31 . 4 Loh TP , Antoniou G , Baghurst P , Metz MP . Development of paediatric biochemistry centile charts as a complement to laboratory reference intervals . Pathology 2014 ;46 :336 –43 . 10.1097/PAT.0000000000000118 24798150 5 Coylewright M , Reckelhoff JF , Ouyang P . Menopause and hypertension: an age-old debate . Hypertension 2008 ;51 :952 –9 . 10.1161/HYPERTENSIONAHA.107.105742 18259027 6 Franklin SS , Gustin W 4th, Wong ND , Larson MG , Weber MA , Kannel WB , et al Hemodynamic patterns of age-related changes in blood pressure. The Framingham Heart Study . Circulation 1997 ;96 :308 –15 . 9236450 7 Franklin SS . Hypertension in older people: part 1 . J Clin Hypertens (Greenwich) 2006 ;8 :444 –9 .16760685 8 Chrysant SG , Chrysant GS . The age-related hemodynamic changes of blood pressure and their impact on the incidence of cardiovascular disease and stroke: new evidence . J Clin Hypertens (Greenwich) 2014 ;16 :87 –90 .24373633 9 Kim JH , Shin JH , Lee HJ , Kim SY , Bae HY . Discordance between HbA1c and fasting plasma glucose criteria for diabetes screening is associated with obesity and old age in Korean individuals . Diabetes Res Clin Pract 2011 ;94 :e27 –29 . 10.1016/j.diabres.2011.07.013 21835487 10 Menke A , Rust KF , Savage PJ , Cowie CC . Hemoglobin A1c, fasting plasma glucose, and 2-hour plasma glucose distributions in U.S. population subgroups: NHANES 2005–2010 . Ann Epidemiol 2014 ;24 :83 –9 . 10.1016/j.annepidem.2013.10.008 24246264 11 Dubowitz N , Xue W , Long Q , Ownby JG , Olson DE , Barb D , et al Aging is associated with increased HbA1c levels, independently of glucose levels and insulin resistance, and also with decreased HbA1c diagnostic specificity . Diabet Med 2014 ;31 :927 –35 . 10.1111/dme.12459 24698119 12 Szoke E , Shrayyef MZ , Messing S , Woerle HJ , van Haeften TW , Meyer C , et al Effect of aging on glucose homeostasis: accelerated deterioration of beta-cell function in individuals with impaired glucose tolerance . Diabetes Care 2008 ;31 :539 –43 . 18083793 13 Amati F , Dubé JJ , Coen PM , Stefanovic-Racic M , Toledo FG , Goodpaster BH . Physical inactivity and obesity underlie the insulin resistance of aging . Diabetes Care 2009 ;32 :1547 –9 . 10.2337/dc09-0267 19401446 14 Liu J , Wu YY , Huang XM , Yang M , Zha BB , Wang F , et al Ageing and type 2 diabetes in an elderly Chinese population: the role of insulin resistance and beta cell dysfunction . Eur Rev Med Pharmacol Sci 2014 ;18 :1790 –7 . 24992623 15 Newschaffer CJ , Bush TL , Hale WE . Aging and total cholesterol levels: cohort, period, and survivorship effects . Am J Epidemiol 1992 ;136 :23 –34 . 1415129 16 Yamada M , Wong FL , Kodama K , Sasaki H , Shimaoka K , Yamakido M . Longitudinal trends in total serum cholesterol levels in a Japanese cohort, 1958–1986 . J Clin Epidemiol 1997 ;50 :425 –34 . 9179101 17 Upmeier E , Lavonius S , Heinonen P , Viitanen M , Isoaho H , Arve S , et al Longitudinal changes in serum lipids in older people the Turku elderly study 1991–2006 . Age Ageing 2011 ;40 :280 –3 . 10.1093/ageing/afq180 21252037 18 Weijenberg MP , Feskens EJ , Kromhout D . Age-related changes in total and high-density-lipoprotein cholesterol in elderly Dutch men . Am J Public Health 1996 ;86 :798 –803 . 8659652 19 Abbott RD , Yano K , Hakim AA , Burchfiel CM , Sharp DS , Rodriguez BL , et al Changes in total and high-density lipoprotein cholesterol over 10- and 20-year periods (the Honolulu Heart Program) . Am J Cardiol 1998 ;82 :172 –8 . 9678287 20 Walter M . Interrelationships among HDL metabolism, aging, and atherosclerosis . Arterioscler Thromb Vasc Biol 2009 ;29 :1244 –1250 . 10.1161/ATVBAHA.108.181438 19667114 21 Akahoshi M , Soda M , Nakashima E , Shimaoka K , Seto S , Yano K . Effects of menopause on trends of serum cholesterol, blood pressure, and body mass index . Circulation 1996 ;94 :61 –6 . 8964119 22 Matthews KA , Crawford SL , Chae CU , Everson-Rose SA , Sowers MF , Sternfeld B , et al Are changes in cardiovascular disease risk factors in midlife women due to chronological aging or to the menopausal transition? J Am Coll Cardiol 2009 ;54 :2366 –73 . 10.1016/j.jacc.2009.10.009 20082925 23 Schalk BW , Visser M , Deeg DJ , Bouter LM . Lower levels of serum albumin and total cholesterol and future decline in functional performance in older persons: the Longitudinal Aging Study Amsterdam . Age Ageing 2004 ;33 :266 –72 . 15082432 24 Kronborg J , Solbu M , Njølstad I , Toft I , Eriksen BO , Jenssen T . Predictors of change in estimated GFR: a population-based 7-year follow-up from the Tromsø study . Nephrol Dial Transplant 2008 ;23 :2818 –26 . 10.1093/ndt/gfn148 18400822 25 Jiang S , Sun X , Gu H , Chen Y , Xi C , Qiao X , et al Age-related change in kidney function, its influencing factors, and association with asymptomatic carotid atherosclerosis in healthy individuals—a 5-year follow-up study . Maturitas 2012 ;73 :230 –8 . 10.1016/j.maturitas.2012.07.014 22951150 26 Conte G , Minutolo R , De Nicola L . Pro: Thresholds to define chronic kidney disease should not be age-dependent . Nephrol Dial Transplant 2014 ;29 :770 –774 ; discussion 780–2. 10.1093/ndt/gft324 24449100 27 Glassock RJ . Con: Thresholds to define chronic kidney disease should not be age dependent . Nephrol Dial Transplant 2014 ;29 :774 –9 ; discussion 779–82. 10.1093/ndt/gft306 24449099

PMC005xxxxxx/PMC5003349.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757135910.1371/journal.pone.0161412PONE-D-16-21705Research ArticlePhysical SciencesChemistryChemical CompoundsOrganic CompoundsAminesCatecholaminesDopaminePhysical SciencesChemistryOrganic ChemistryOrganic CompoundsAminesCatecholaminesDopamineBiology and Life SciencesBiochemistryNeurochemistryNeurotransmittersBiogenic AminesCatecholaminesDopamineBiology and Life SciencesNeuroscienceNeurochemistryNeurotransmittersBiogenic AminesCatecholaminesDopamineBiology and Life SciencesBiochemistryHormonesCatecholaminesDopamineBiology and life sciencesGeneticsEpigeneticsRNA interferenceBiology and life sciencesGeneticsGene expressionRNA interferenceBiology and life sciencesGeneticsGenetic interferenceRNA interferenceBiology and life sciencesBiochemistryNucleic acidsRNARNA interferenceBiology and Life SciencesNeuroscienceSensory PerceptionVisionBiology and Life SciencesPsychologySensory PerceptionVisionSocial SciencesPsychologySensory PerceptionVisionResearch and Analysis MethodsModel OrganismsAnimal ModelsDrosophila MelanogasterBiology and Life SciencesOrganismsAnimalsInvertebratesArthropodaInsectsDrosophilaDrosophila MelanogasterBiology and Life SciencesBiochemistryProteinsDopamine TransportersBiology and Life SciencesBiochemistryNeurochemistryNeurochemicalsDopaminergicsBiology and Life SciencesNeuroscienceNeurochemistryNeurochemicalsDopaminergicsBiology and Life SciencesAnatomyNervous SystemSynapsesMedicine and Health SciencesAnatomyNervous SystemSynapsesBiology and Life SciencesPhysiologyElectrophysiologyNeurophysiologySynapsesMedicine and Health SciencesPhysiologyElectrophysiologyNeurophysiologySynapsesBiology and Life SciencesNeuroscienceNeurophysiologySynapsesBiology and Life SciencesBehaviorVisual Attention in Flies—Dopamine in the Mushroom Bodies Mediates the After-Effect of Cueing Cueing Selective Visual AttentionKoenig Sebastian Wolf Reinhard Heisenberg Martin *Rudolf Virchow Center for Experimental Biomedicine, University of Würzburg, Joseph-Schneider-Straße 2, 97080, Würzburg, GermanySkoulakis Efthimios M. C. EditorBiomedical Sciences Research Center Alexander Fleming, GREECECompeting Interests: The authors have declared that no competing interests exist. Conceptualization: SK RW MH. Data curation: SK. Formal analysis: SK RW. Funding acquisition: MH. Investigation: SK. Methodology: SK RW. Project administration: MH. Resources: RW. Software: SK RW. Supervision: MH. Validation: SK RW MH. Visualization: SK RW MH. Writing – original draft: SK RW MH. Writing – review & editing: SK RW MH. * E-mail: heisenberg@biozentrum.uni-wuerzburg.de29 8 2016 2016 11 8 e016141230 5 2016 4 8 2016 © 2016 Koenig et al2016Koenig et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Visual environments may simultaneously comprise stimuli of different significance. Often such stimuli require incompatible responses. Selective visual attention allows an animal to respond exclusively to the stimuli at a certain location in the visual field. In the process of establishing its focus of attention the animal can be influenced by external cues. Here we characterize the behavioral properties and neural mechanism of cueing in the fly Drosophila melanogaster. A cue can be attractive, repulsive or ineffective depending upon (e.g.) its visual properties and location in the visual field. Dopamine signaling in the brain is required to maintain the effect of cueing once the cue has disappeared. Raising or lowering dopamine at the synapse abolishes this after-effect. Specifically, dopamine is necessary and sufficient in the αβ-lobes of the mushroom bodies. Evidence is provided for an involvement of the αβposterior Kenyon cells. http://dx.doi.org/10.13039/501100001659Deutsche ForschungsgemeinschaftHE 986/20-1 (Reinhart Koselleck)Heisenberg Martin This work was supported by the Rudolf-Virchow-Center and German Science Foundation, Reinhart Koselleck HE 986/20-1. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Flies (Drosophila melanogaster) like other animals and humans can restrict their visual responses to parts of the visual field [1–6]. This property of vision is called selective visual attention (SVA). More appropriately it should be called 'spatially selective visual attention' to distinguish it from attention to novelty or other visual features such as colors, patterns or motion (for a review see [7]). SVA refers exclusively to spatial selectivity. The fly is assumed to have a focus of attention (FoA) which it can spontaneously shift to particular areas of its visual field and to be more likely to respond to visual stimuli occurring in this region than elsewhere. In flies SVA has been studied mainly in stationary flight of tethered animals. Under these highly restricted experimental conditions visual stimuli can be presented in defined regions of the fly's visual field and the fly shows by its intended turning (yaw-torque) to which of the stimuli it responds. For instance, one can present two black stripes at symmetrical positions in the fronto-lateral visual field of the fly and move them suddenly at the same time from front to back as seen by the fly [5]. The fly responds to each of the stripes in about 1/3 of the trials, presumably to the one closest to its FoA. In 1/3 of the trials the fly does not respond at all. Whether this implies that neither stripe happens to be close enough to the FoA is not known. Using this test we have recently shown that the fly keeps the FoA for several seconds at the location to where it has endogenously shifted it [6]. In humans this endogenously driven process is called a top-down modulation of covert attention [8]. In the present study we investigate shifts of the FoA elicited by external visual cues [1, 2, 5]. For instance, in the two-stripes test above one can direct the fly's attention to one side by briefly oscillating one of the stripes prior to the displacement. This increases the frequency of responses to the displacement of the stripe on the side of the oscillation and reduces responses to the stripe on the other side. The cue and the test can be separated in time as well as spatially [5]. In the first part of the present study we reinvestigate external visual cueing. It turns out to be behaviorally more complex than had been expected. In the second part we focus on the after-effect of cueing and show that it is mediated by dopamine signaling and that for this behavioral effect dopamine is required in the mushroom bodies (MBs). The MBs are prominent symmetrical structures in the two brain hemispheres (e.g. [9]), each consisting of about 2000 intrinsic neurons, the so-called Kenyon cells. They have their dendrites dorso-posteriorly in the so-called calyx, their axons constitute the MB peduncle which anteriorly divides into the medial and vertical lobes. The mass of parallel Kenyon cells consists of 3 compartments, α/β, α’/β’ and γ. The α/β compartment is further subdivided into α/βs (surface), α/βc (core) and α/βp (posterior). The α/βp compartment can be distinguished from the main part of the MB by its morphology. Its about 90 Kenyon cells are connected to a special dendritic region, the ventral accessory calyx, separated from the main calyx, and they mix with fibers of other compartments in the β lobe[10]. The MBs are involved in a variety of complex brain functions such as learning and memory, sleep regulation, decision-making, context generalization and higher order motor control [11–15]. With the exception of learning and memory it is still obscure how the MBs contribute to these functions. We show here that interference with dopamine signaling in the α/β compartment and specifically the α/βp Kenyon cells suppresses the after-effect of cueing. Materials and Methods Flies Flies were cultured at 25°C on standard medium with 60% relative humidity on a 12h light/dark cycle. Wild type CantonS (CS) and GAL4 lines MB247, OK107 and c305 were of the Wuerzburg stock collection (Biozentrum, Department of Neurobiology and Genetics); c739 was obtained from Bloomington stock collection, NP1131 from Max-Planck-Institute of Neurobiology, Martinsried, Germany, c708 from Scott Waddell (Oxford University, UK). The mutant radish1 was provided by Josh Dubnau (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY), the mutant DopEcR by Bertram Gerber (Leibniz Institute for Neurobiology, Magdeburg, Germany), the RNAi stocks were from VDRC stock center (#106961 and #12082; Vienna, Austria) and all fumin lines from Kazuhiko Kume (Nagoya University, Nagoya, Japan). For tethering, we anesthetized 2 to 4 days old female flies by cooling and used a micromanipulator to glue (ESPE Sinfony™ DO3, 3M, Neuss, Germany) them to a triangular-shaped wire-hook made of copper (Ø = 0.05mm). This procedure also fixed the fly’s head to its thorax and prevented independent motion of the two body parts. After polymerization of the glue (10s pulse of blue LED light, < 0.5cm distance) flies were kept in single vials with access to water for a minimum of 2h. Pharmacological experiments To influence dopamine signaling, 2 days old radish1 or CS flies were put for 14h on 5mg of Methylphenidate hydrochloride (Sigma) mixed under 10ml of regular food. To inhibit dDAT or DopR1 function, we kept 2 days old CS flies for 14h on 10ml of regular food with 30mg Desipramine hydrochloride or 1mg (+)-Butaclamol hydrochloride (Sigma), respectively. For inhibition of dopamine synthesis, freshly hatched CS flies were put on 20ml of regular food with 8mg α-Methyl-DL-tyrosine (Sigma) for 120h. We verified uptake of food by the addition of a non-hazardous blue dye, which after 14h stained the abdomen of the flies. To ablate the mushroom bodies, CS flies were treated with Hydroxyurea (Sigma) [16]. Dopamine was measured by HPLC (M. Krischke, Botany Department, University of Wuerzburg). Setup A fly was attached to the torque-meter and centered in a cylindrical arena (Ø = 90mm, h = 90mm) covering 360° x ±45° of the fly's visual sphere (Fig 1A). Visual stimuli were projected (BenQ W770ST, 120Hz) onto a rectangular plate, which held the ends of 180 (horizontal) x 32 (vertical) single light-guides. The other ends of the light-guides penetrated the wall of the arena, thereby transmitting the visual stimuli from the plate to the arena. The arena’s floor was covered with black cardboard and experiments were performed in a dark chamber. Position, timing and geometrical properties of the visual stimuli were controlled and updated at 300Hz using custom-made software (VB.NET). A torque-meter was used to measure the generated yaw-torque and the values were stored on the controlling computer’s hard disk at 100Hz. Experiments were performed under open-loop conditions, i.e. giving the fly no visual feedback of its generated yaw-torque. 10.1371/journal.pone.0161412.g001Fig 1 Cueing in selective visual attention. (A) Apparatus: Light-guide arena and torque meter. (B) Experimental procedure. The fly is attached to a torque-meter and positioned in the center of a light-guide arena. Stimuli are generated by a computer and projected onto one end of the light-guides. These display the visual stimuli at the inner surface of the arena. Two black stripes (width w = 18°; at ψ0 = + 45° and ψ0 = – 45°) are displaced from front to back on white background (Δψ = 30° at 150°/s) and then slowly reset to their initial position (v = 20°/s). They remain there for the inter-trial interval (ITI). The displacement may be preceded by an oscillation of one of the stripes (cue; e.g. Δψcue = 15°, fcue = 10Hz and durcue = 1s) and a post-cue pause. (C) Immediate cueing and cueing after-effect. If the cue is immediately followed by the test, a strong bias in the response frequencies can be seen (ICE). It is slightly reduced, if a post-cue pause of 1, 2 or 3s is introduced (CAE). If the stripes remain stationary after the cue for 4 or 5s, the cue does no longer bias the responses towards the not cued stripe. Stripe width w = 18°; oscillation amplitude Δψcue = 15° (N = 52). (D) Cueing does not reduce the overall response rate. The same flies are tested either without cueing, with cueing of one stripe or cueing of both stripes (N = 71). (E) ICE and CAE are observed with positive and negative cueing (stripe width w = 6°; Δψcue = 2° for attractive cueing; Δψcue = 4° for repulsive cueing; N = 22). Error bars are SEMs (*P < 0.05, **P < 0.01, ***P < 0.001). Stimulus conditions Unless stated otherwise, two black 18° wide and ±45° high stripes were presented on a white background, centered at ψ0 = ±45° in the fronto-lateral visual field of the fly. The stripes were displaced from front to back by Δψ = 30° at 150°/s and then slowly reset to their initial position at 20°/s. The inter-trial interval was set to 2s and prior to each displacement a cue, followed by a post-cue pause was added. To cue, we oscillated one of the stripes (Δψcue = 15°, fcue = 10Hz and durcue = 1s) and during the post-cue pause the stripes remained stationary. Usage of different cueing amplitudes and stripe width are stated in the corresponding text and figure legends. Six different sets of post-cue pauses (0, 1, 2, 3, 4 and 5s) were used in a typical experiment. A single set included six displacements, three times with cueing of the left stripe and three times of the right stripe. The order of sets as well as the order of the cued sides within each set was randomized. In experiments with a single stripe, both stripes were controlled by an identical protocol, but only one was shown/displaced. A response was scored when yaw-torque modulation exceeded the range between maximum and minimum yaw-torque values recorded within 0.5s prior to displacement by more than 60% within 0.5s after onset of the displacement. Left (counterclockwise; ccw) and right (clockwise; cw) responses as seen from the position of the fly were scored separately. If no sufficiently large yaw-torque modulation could be detected, a no response (nr) was scored. Data evaluation Flies had to respond to at least 22 of the 36 test trials to be included in the evaluation. To quantify the effect of cueing, a response index (RI) was used to show the distribution of responses between the cued and the not cued stripe. It was calculated as RI = (rfcued—rfuncued) / (rfcued + rfuncued) so that an equal number of responses towards and away from the cued stripe yielded a RI of 0. For every fly a separate RI was computed for each of the 6 different sets of post-cue pauses. All single fly RIs were then averaged for each post-cue pause. The wild type data consistently showed a tripartite pattern. The highest RI was found for post-cue pause 0s, whereas for pauses 4 and 5s the RI was not different from zero. Post-cue pauses 1-3s showed plateau values, which were significantly different from zero (Fig 1C). Thus, for averaging the data were grouped as post-cue pause 0s (immediate cueing effect, ICE) and post-cue pauses 1, 2 and 3s (cueing after-effect, CAE). In the experiments investigating the consequences of a reduced cueing duration or the efficacy of cueing in the upper or lower visual field, the CAE included only post-cue pauses of 1s. Statistical analysis All data were tested for normal distribution using a Kolmogorov-Smirnov test. If data were normally distributed, a one-sample t test was used to compare values with zero and a two-sample t test was used to compare values with each other. Bonferroni corrections were used for multiple comparisons. If no normal distribution could be assumed, either a Wilcoxon Matched Pairs test for dependent pairwise comparisons or a Mann-Whitney test was used to test two groups against each other. A Wilcoxon-Signed-Rank test was used to compare values with zero. Comparison of more than two values was achieved by a one-way ANOVA with Holm-Sidak’s multiple comparisons test, if the data were normally distributed and otherwise with a Kruskal-Wallis test with Dunn’s test for multiple comparisons (* = p < 0.05, ** = p < 0.01, or *** = p < 0.001). Results A) Properties of visual cueing A cue can be attractive or repellent To study cueing in selective visual attention (SVA) we used the paradigm of [5]but with different stimulus parameters (see Material and Methods). In the two-stripes test stripes (w = 18°; ψ0 = + and– 45°) were simultaneously displaced front-to-back and reset slowly to their initial positions. Each of the two motion stimuli alone would have elicited a prevalently syn-directional response, incompatible with a simultaneous response to the other stripe. The simultaneous displacement forced the fly to choose one stripe and the fly tried to turn towards that stripe. To specifically address the effects of cueing on SVA, one of the two stripes was oscillated before the displacement (Fig 1B). A typical stimulus sequence consisted of an inter-trial interval of 2s during which the stripes remained stationary, followed by an oscillation of one of the stripes (Δψcue = 15°) for 1s and by a post-cue pause of 0-5s before the displacement and subsequent resetting of both stripes. As in the experiments of [5]the cue had a strong influence on the response frequencies, however, in the opposite direction. Responses towards the not-cued stripe became more, responses towards the cued stripe less frequent (Fig 1D). The overall response rate, as in the earlier experiments, remained constant. Even if both stripes were cued at the same time, the no-response rate stayed the same (Fig 1D). Thus, with the present experimental conditions in the light-guide arena the cue was repulsive (Fig 1D), while in the LED arena [5] it had been attractive. Apparently, a cueing event on one side prior to the test can make one of the targets more attractive or repellent than the other. Moreover, the fly may just ignore the cue. [5]had observed that the flickering of a bright stripe of LEDs at the position of one of the test stripes had no cueing effect. As in previous studies [5, 6], in the present study the same cueing event was presented over and over again with one or the other stripe, up to 36 times. With none of the cueing conditions it was always attractive or always repellent. It merely influenced the ratio of the response frequencies towards and away from the cued side. As in a single series the cueing event was always the same, we assumed that also the evaluation of it by the fly was the same (attractive or repellent), but that the cueing was sometimes effective and sometimes ineffective. Alternatively, however, the result of the evaluation of the cueing event by the fly might have changed from trial to trial and the ratio of response frequencies just reflected the time ratio of assessments of the cueing effect as attractive and repellent by the fly. This alternative could not be ruled out, but we consider it unlikely. We successfully reproduced the positive cueing in the LED arena with the old stimulus parameters. In those experiments the cue had been followed immediately by the test (post-cue pause = 0s). As mentioned above, with this time course (ICE: Immediate Cueing Effect) the cue in the new arena had a slightly larger effect than with the post-cue pause of 1-3s (cueing after-effect; CAE; Fig 1C). Next, we checked various stimulus conditions in the new setup to see under which conditions cueing was attractive and under which repulsive. Reducing the luminance of the light-guide arena (127.0μW/cm2) to a value (4.0μW/cm2) lower than that of the LED arena (17.5μW/cm2) did not restore attractive cueing. With narrow test stripes (w = 6°) and an oscillation amplitude of Δψcue = 4°) as was used by [5]cueing again was repulsive. However, with a smaller oscillation amplitude (w = 6°; Δψcue = 2°) cueing was attractive. With the same flies cueing was attractive when the stripe was oscillated with a 2° amplitude and repulsive when the amplitude was increased to 4° (Fig 1E). With the 10Hz oscillation it was just the oscillation amplitude that mattered. Selective visual attention in the light-guide arena To gather further insights into SVA and cueing in the light-guide arena, we used broad stripes (w = 18°) and a large oscillation amplitude (Δψcue = 15°). Repulsive cueing was observed under various conditions, such as inverted contrast, with male flies instead of females, black stripes on green background, flicker instead of oscillation or just showing a different background color on one side for 1s as the cue. In two cases (flickering grey stripe and changing background color) only an ICE was observed but no CAE (S1 Fig). We went back to analyzing the effect of cueing in the response to the displacement of a single vertical black stripe (Fig 2). In half of the trials the stripe oscillated (cueing event) prior to the sudden shift. Most often the fly responded with a strong phasic yaw-torque modulation in the same direction as the motion stimulus (Fig 2A, Single, syn-directional) To show the time course of the yaw torque responses, these were synchronized for averaging at the beginning of the rising phase. Response latencies are shown in Fig 2B. If the fly would have been in closed-loop with its visual surround this response would have served to counter-balance the stimulus motion. Remarkably, in rare cases the fly generated yaw-torque spikes with opposite polarity to that of the displacement (Fig 2A, Single, anti-directional to test). Their latency was only about half the latency of syn-directional responses (Fig 2B, Single). They had a smaller amplitude and during the resetting of the stripe the fly generated a weak syn-directional response before its yaw-torque reached the base line again (Fig 2A, Single, anti-directional). The fast anti-directional responses might be interpreted as some kind of escape behavior. In ~25% of the tests the fly showed no response at all. The cue prior to the displacement did not influence the frequency of occurrence of the response types (Fig 2C, Single, no cue and stripe cued). 10.1371/journal.pone.0161412.g002Fig 2 Characterization of yaw-torque responses. (A) Average yaw-torque traces of CS flies under two different experimental conditions. Flies respond to displacements of a single stripe syn- or anti-directionally. No differences can be seen in the responses with regard to the absence (‘Single’, N = 25) or presence (‘One of two’, N = 35) of a second stationary stripe. (B) Response latencies. Anti-directional responses to a single stripe are faster than syn-directional ones. No such fast responses can be observed, when two stripes are displaced. (C) Response frequencies. If only one stripe is displaced, the majority of responses is towards the direction of motion of this stripe. Only a small fraction of anti-directional responses can be observed. The simultaneous displacement of two stripes causes an increase in no responses and the cue biases the responses towards the not cued stripe (‘Both’, N = 52). (D) Slope and height of yaw-torque responses (red and blue) and spontaneous body saccades (green). Responses are grouped as syn- and anti-directional. No such distinction can be made in the ‘Both’ condition. Syn-directional responses differ in the analyzed parameters from anti-directional ones, which in turn are comparable to body saccades. (E) Average yaw-torque traces of CS flies in response to the simultaneous displacement of two stripes. Responses, syn- and anti-directional to cue, both resemble the syn-directional responses to a single stripe. (F) Response latencies have a similar value of about 100ms (N = 22), irrespective of whether the cue biases the response frequency towards the cued (‘Positive cueing’) or the not cued stripe (‘Negative cueing’), Error bars are SEMs (*P < 0.05, **P < 0.01, ***P < 0.001). Similar yaw-torque patterns could be found, if the fly faced two stripes at ψ0 = + and– 45° and only one of them was displaced. The overall response frequencies remained the same and it again did not matter whether the displaced stripe was cued or not (Fig 2C, One-of-two, displaced stripe cued and stationary stripe cued). Also the response latencies of syn- and anti-directional responses were about the same in the ‘Single’ and ‘One-of-two’ conditions (Fig 2B). In between trials the fly occasionally generated spontaneous yaw-torque spikes (body saccades, [2, 17]). Their response pattern resembled that of anti-directional responses and they could thus clearly be separated from syn-directional ones. As their dynamics nevertheless differed between the ‘Single’ and the ‘One of two’ experiments, it is likely that they were influenced by stimulus parameters other than the motion of the stripe (Fig 2D). Interestingly, if both stripes were displaced, the fly increased the frequency of no responses (Fig 2C, Both) and suppressed the fast escape attempts (Fig 2B, Both and see below). Syn- and anti-directional responses with regard to the cued stripe looked much like the syn-directional (attractive) responses when only one stripe was displaced (Fig 2E). All these properties of SVA did not differ for repellent and attractive cueing if these were measured side by side in the same animals (w = 6°; Δψcue = 2 vs 4°; see for instance response latencies in Fig 2F). Response polarity is independent of yaw-torque Using a visual learning task in the flight simulator, [3] provided evidence that yaw-torque and the location in the visual field to which a fly attends might be connected. Additionally, [4] found side-specific higher levels of brain activity depending on the direction in which a fly tried to turn. We looked for such a connection by analyzing the yaw-torque during 1s before the displacements and categorized it according to the cued side (cwcue; ccwcue) as well as syn- and anti-directional responses as ‘cwcue-syn’, ‘cwcue-anti’, ‘ccwcue-syn’ and ‘ccwcue-anti’. A putative influence of the preceding yaw-torque level on response polarity as well as an influence of the cue on the yaw-torque-level should thus become visible as a difference of the yaw-torque histograms of the 4 groups. No such difference was observed (Fig 3). The result again was the same for repulsive and attractive cueing (data for attractive cueing not shown). We conclude that the flies were able to shift their FoA independent of yaw-torque in our paradigm. 10.1371/journal.pone.0161412.g003Fig 3 Yaw-torque level does not influence response polarity. Responses are grouped as (syn,cwcue), (anti,cwcue), (syn,ccwcue) and (anti,ccwcue). Histograms of CS yaw-torque values generated within 1s before the displacements show no difference (N = 138). Visual field properties [5] had discovered that the FoA could be cued in the lower but not in the upper visual field (LVF; UVF). In the light-guide arena we confirmed this effect for positive and negative cueing measuring the ICE. For the CAE, however, we found negative cueing in both the UVF and the LVF but positive cueing in neither (Fig 4A and 4B). For negative cueing in the UVF a CAE was observed but no ICE, implying that ICE and CAE might be independent of each other. 10.1371/journal.pone.0161412.g004Fig 4 Visual field properties of cueing. (A) For positive cueing conditions (‘6°|2°’) no cueing is observed (N = 32) if the stripes are restricted to the upper visual field (UVF). If stripes are restricted to lower visual field (LVF) cueing elicits the ICE but not the CAE. (B) Negative cueing ('6°|4°’) elicits the CAE irrespective of whether the stripes are restricted to the UVF or LVF with the exception of the ICE which is elicited only in the LVF (N = 23). Is a repellent cue initially attractive? We considered the possibility that repellent cueing might be a two-step process. The first appearance of the cue might attract the FoA and a short inspection of the cue might trigger the second step, a shift of the FoA away from this side. We therefore measured the dynamics of the cueing effect to see whether the repulsion was immediate or preceded by a short attractive phase. After a shortened cueing duration the FoA might still be on the cued side at the onset of the displacement. We shortened the cueing and measured both the ICE and the CAE. (The CAE in this experiment was measured with a post-cue pause of only 1s instead of 1-3s.) In short, no indication of an initially positive cueing effect was found (Fig 5A and 5B). At the earliest moment cueing could be observed it was negative. Cueing occurred already with two cycles of the oscillation (cueing duration = 0.2s), while 0.1s were not long enough to result in an ICE or CAE. This does not exclude the possibility that cueing might attract the FoA already with a short cueing duration, but this attraction would not show, since the response would not be activated before the FoA was shifted from the cued to the not cued side. For the ICE a similar picture emerged for positive cueing. The cue had to last for at least 0.2s to cause an ICE. Interestingly, no CAE was observed with positive cueing, if cueing was shorter than 1s. Again, what happened to the FoA before remains open. 10.1371/journal.pone.0161412.g005Fig 5 (A) Reducing cueing duration. If the cue duration is shorter than 0.2s no cueing is observed. Shortening of a negative cue ('6°|4°’, N = 23) does not make it less repellent. (B) For positive (‘6°|2°’, N = 22) but not negative cueing the oscillation has to last at least for 1s to elicit the CAE. All error bars are SEMs (*P < 0.05, **P < 0.01, ***P < 0.001). B) Neuronal mechanisms Reducing dopamine via αMT abolishes the CAE In many animal species attentional mechanisms have been shown to involve the dopaminergic system [18–23]. We wondered whether this also applies to SVA in Drosophila. We used negative cueing and recorded it immediately after the cue disappeared (ICE) and 1-3s later (cueing after-effect, CAE; Materials & Methods). Wild type flies were fed with α-methyl-DL-tyrosine (αMT), an inhibitor of tyrosin hydroxylase (TH), which in turn is a rate-limiting enzyme in the synthesis of dopamine (Fig 6A). The reduction in dopamine led to a loss of both the ICE and CAE (Fig 6B). Thus, sufficient amounts of dopamine are crucial for the cueing effect. 10.1371/journal.pone.0161412.g006Fig 6 Reduced dopamine synthesis. (A) Feeding the drug α-methyl-DL-tyrosine (αMT) decreases dopamine levels in the fly [N (samples) = 19, 18; 20 heads per sample; Tukey whisker box plot]and (B) suppresses ICE and CAE [N (flies) = 23, 44; means and SEMs]. (*P < 0.05, **P < 0.01, ***P < 0.001). Dopamine concentrations in (A) are shown as boxplots with Tukey whiskers. If not otherwise stated, all N's are flies. Interfering with the dopamine transporter Mutant fumin (dDATfmn or fmn) flies have a defect in the gene for the dopamine transporter (dDAT), leading to a defective dopamine re-uptake and, in turn, to increased dopamine signaling [24] at dopamine synapses. dDATfmn flies show hyperactivity as well as alterations in their activity/rest pattern. Their yaw torque responses to stripe displacements were remarkably normal. We tested them for effects on cueing in SVA and found an intact ICE, but no CAE (Fig 7). Even heterozygous dDATfmn/+ flies showed the same defects as the homozygous mutant animals (see (D) in last figure). As it appears, not only a decrease of intracellular dopamine but also an increase of extracellular dopamine at the synapse shortens the time course of cueing in SVA. 10.1371/journal.pone.0161412.g007Fig 7 Dopamine re-uptake modulates the CAE. Flies with a lesion in the dDAT gene (fmn-/fmn-) still have an ICE but lack the CAE (N = 26). The same phenotype is found in flies heterozygous for this mutation. ICE and CAE are absent in rsh1 flies. Both defects can be rescued by inhibition of dopamine re-uptake via MPH (N = 21, 25). Pharmacological interference with dDAT function: Desipramine is an inhibitor of dDAT and its application leads to the dDATfmn phenotype (N = 26, 29). MPH is known to inhibit dopamine re-uptake via dDAT and causes a loss of CAE and ICE in wildtype CS (N = 41, 42). Error bars are SEMs (*P < 0.05, **P < 0.01, ***P < 0.001). We confirmed the dDATfmn result pharmacologically, treating wild type flies with the drug Desipramine [24], an inhibitor of dDAT. Flies kept on food with 1mg/ml Desipramine for 14h resembled dDATfmn flies, showing intact ICE but no CAE (Fig 7). We also applied methylphenidate hydrochloride (MPH), a drug prescribed against attention deficit and hyperactivity disorder (ADHD). MPH targets the dopaminergic system and, in particular, the re-uptake of dopamine via dDAT [25]. Interestingly, MPH treatment of wild type blocked both ICE and CAE (Fig 7). Mutant radish flies are impaired in cueing A further gene of interest for the investigation of SVA is radish (rsh). It was initially found to play a role in memory and has later also been implicated in attention-like processes by [26]. Exploiting novelty choice behavior later described at the behavioral level by [27], they found a characteristic modulation of a local field potential in response to the novel stimulus. This particular modulation was less sustained in rsh1 flies, indicating a defect in short-term choice processes. We investigated cueing in mutant rsh1 flies. They showed neither an ICE nor a CAE while yaw-torque responses to test stimuli where normal. Studying attention-like behavior, [26] had presented evidence that feeding rsh1 flies with MPH rescued defects in selection/suppression dynamics of this behavior. In our study as well, the mutant phenotype in SVA could be reverted by feeding MPH to rsh1 flies (Fig 7). These flies behaved much like the wild type in terms of having a significant ICE, which was slightly higher than the also significant CAE. The result suggests that in rsh1 flies the impairment of cueing is due to a local shortage of dopamine at the relevant synapses, which can be compensated by MPH inhibiting dDAT. Dopamine receptor DopR1 is involved in cueing SVA To exert its effect as a neurotransmitter, dopamine needs to bind to a receptor. Drosophila has three types of dopamine receptors, two D1-like (DopR1, DopR2) and one D2-like (DD2R). They are involved in a wide range of behaviors such as learning, wakefulness, arousal and locomotion [28–31]. Additionally, in Drosophila there is a β-adrenergic-like G-protein coupled receptor called DopEcR with high expression in the CNS, which responds to dopamine as well as to ecdyson[32]. We obtained mutants for three of the four receptor genes to find out which of them was involved in SVA and particularly cueing. Flies without functional DopEcR were compromised neither in ICE nor in CAE. Cueing seems not to require DopEcR (Fig 8A). Likewise, DopR2 mutant flies (DAMB) showed no defects in SVA, displaying wild type like ICE and CAE (Fig 8B). The third mutant was dumb2, a hypomorphic DopR1 allele. We found a significant ICE, but no CAE in dumb2 flies (Fig 8C; for an expression pattern of DopR1 receptor see [29]). To independently confirm this result, wild type flies were put for 14h on food enriched with 1mg/ml Butaclamol, a DopR antagonist with higher affinity for DopR1 than for DopR2 [33–36]. The effects of this treatment resembled the phenotype of dumb2 mutant flies: a normal ICE, but no significant CAE (Fig 8D). Given that DopR2 mutant flies showed normal cueing, DopR1 seems to be the relevant target of Butaclamol. Taking all these results together the findings point at DopR1 as the most likely candidate for mediating the dopamine effects on the time course of SVA. 10.1371/journal.pone.0161412.g008Fig 8 Dopamine receptors for SVA. (A-C) Out of three tested dopamine receptors only dDA1 (mutant dumb2) is critical for the cueing after-effect (N = 26, 35; N = 27, 30; N = 29, 35). (D) Pharmacological manipulation of dDA1 by the antagonist Butaclamol causes the same phenotype. Controls are CantonS (CS); mutants are homozygous, on CS background. All error bars are SEMs (*P < 0.05, **P < 0.01, ***P < 0.001). Mushroom bodies are required for cueing As the dopamine system is intimately linked to the MBs and several of the genes used in this study are preferentially expressed there, we wondered whether the MBs are involved in SVA and specifically in cueing. We fed newly hatched larvae with hydroxyurea (HU, [16]), to ablate the MB neuroblasts and to obtain flies with MBs consisting of only embryonic KCs. We tested these flies and found normal test responses but neither a significant ICE, nor a CAE (Fig 9). This finding links SVA and in particular cueing to the MBs. 10.1371/journal.pone.0161412.g009Fig 9 Mushroom bodies are required for cueing. After ablation of the MB neuroblasts in 1st instar larva, adult flies show ICE and CAE neither in negative ('6°|4°’; N = 22, 24) nor positive cueing (‘6°|2°’; N = 29, 20), while the controls behave like wild type. Error bars are SEMs (*P < 0.05, **P < 0.01, ***P < 0.001). Only certain mushroom body compartments matter So far we had shown that proper dopamine signaling, already 50% reduction of dDAT and the MBs each play a pivotal role in the CAE. Next, we probed the role of dDAT at the MBs as a combination of those three. Recently, a critical role of dDAT expression in the MBs has been reported for aversive odor memory as well as sleep regulation [37]. Here we use dDAT expression to assign the CAE to compartments of the MB. This analysis does not include the ICE since neither the transporter (Fig 7) nor the dDA1 receptor (Fig 8) appear to be required for it (ICE + CAE data shown as S2 Fig). We used various MB Gal4-lines to express dDAT in a dDATfmn/+ background in certain MB compartments. To test for necessity, we used RNAi against dDAT-mRNA in the wild type background to block the formation of dDAT protein exclusively in the Gal4-labeled cells. As dDAT is supposed to be a presynaptic dopamine transporter and Kenyon cells are not known to be dopaminergic, it is somewhat speculative to assume that wild type flies express dDAT in Kenyon cells. The results, however, suggest that this is the case. We started with OK107-Gal4, which strongly labels the complete MBs. Restoring dDAT-function in all OK107-labeled cells of dDATfmn/+ flies rescued the loss of CAE. Reciprocally, knocking down dDAT in those cells in wild type flies using RNAi suppressed the CAE (Fig 10A). Next, we drove dDAT-expression in a dDATfmn/+ background with MB247-Gal4, a line that–like OK107-Gal4—labels α/β (including the accessory calyx) and γ-lobes, but has no (or only marginal) expression in α’/β’. The results showed a rescue of the CAE. Again, RNAi against dDAT-mRNA in the same set of cells removed the CAE (Fig 10B). In combination the results so far gave first evidence that α’/β’ might not be required for SVA. Next, we examined the requirement of the γ-lobe. Local dDAT-expression in the Kenyon cells of only the αβ-lobes (c739-Gal4) in a dDATfmn/+ background restored the CAE and knocking down dDAT-function with RNAi in these same cells led to a dDATfmn-like phenotype, proving αβ to be necessary and sufficient for CAE (Fig 10C). As additional controls, dDAT-expression driven in the α’β’-lobes by the driver line c305a-Gal4 and, respectively in the γ–lobes by NP1131-Gal4 (Fig 10D) failed to rescue the CAE. Taken together, dDAT-expression in αβ-, not in α’β’- or γ-lobes is necessary and sufficient for the after-effect of cueing in SVA. 10.1371/journal.pone.0161412.g010Fig 10 Rescue and knockdown of dDAT function in mushroom body compartments. (A) The OK107-Gal4 line labels all Kenyon cells (KCs) of the MBs and dDAT expression in these cells rescues the suppression of the CAE of the heterozygous fmn- (dDATfmn/+) flies (N = 38, 25). Knockdown of dDAT function in the same set of KCs leads to the absence of CAE, emphasizing the importance of the MBs for this behavior (N = 32, 29). (B) MB247-Gal4 has only marginal expression in the α’β’-lobes, but otherwise labels the same KCs as OK107-Gal4. Providing dDAT function in the labeled cells is sufficient to rescue the defect in CAE of the genetic control (N = 35, 17). RNAi against dDAT-mRNA in MB247-Gal4 labeled KCs leads to a dDATfmn-like phenotype (N = 47, 29, 18). (C) c739-Gal4 driven expression of dDAT in the αβ-lobes is sufficient to rescue the dDATfmn-like phenotype of the genetic control (N = 24, 27). Necessity of the αβ-lobes for CAE is demonstrated by the loss of CAE, if RNAi against dDAT-mRNA is expressed in the same set of cells (N = 27, 21). (D) c305a-Gal4 and NP1131-Gal4 label α’β’ and γ-lobes, respectively. Ectopic expression of dDAT in these cells in mutant dDATfmn/+ background does not restore the CAE (N = 20, 25; N = 26, 19). (E) c708a-Gal4 has expression only in the αβp KCs. Also in these cells in dDATfmn/+ flies dDAT expression does not restore the CAE (N = 29, 23, 20). Knocking down dDAT function in these cells, however, is sufficient to cause a dDATfmn-like phenotype (N = 23, 21). All error bars are SEMs (*P < 0.05, **P < 0.01, ***P < 0.001). The Gal4-line c708a has an expression pattern that specifically addresses the αβp KCs. A rescue of dDAT function in these cells was insufficient to restore wild type cueing in dDATfmn/+ flies (Fig 10E). However, knocking down the dDAT protein in these 90 KCs in otherwise wild type flies caused a cueing defect (Fig 10E), ascribing a significant role for cueing in SVA to this small subset of cells. (Expression patterns of the above Gal4-lines (MB compartments) can be found in [9, 38].) Discussion Attractive and repulsive cueing The unexpected finding that cueing is attractive, repulsive or ineffective reveals that also in stimulus-response situations the fly evaluates the stimulus according to the putative consequences of its response options [39]. Like all behavior, cueing in SVA takes part in natural selection. It is generated (or not) to secure survival and possible offspring. Properties of cueing in SVA Having previously studied the attention span after endogenous shifts of the FoA in Drosophila [6], we now examined shifts of the FoA due to external cueing. The cue can be effective or ineffective, attractive or repulsive. It has no effect on the overall response frequency. It modulates the ratio of the response frequencies to the two stripes in relation to the cue. Sign and strength of the effect depend upon a variety of properties of the cue (e.g. visual field position, duration, salience, light intensity, S1 Fig). Repulsive responses after cueing are elicited with the same latency as attractive ones, require the same minimum cueing duration (for ICE) and in both there is no influence of the yaw-torque level prior to a response on the response polarity. Evidently, the fly takes the cue not just as a marker for a particular location in the visual field. It evaluates in addition the cue's potential significance. In our study the fly is tethered and flying in a highly artificial experimental environment. The same cueing event occurs every few seconds. Even for these special experimental conditions our understanding of cueing is still rudimentary. Presumably, the special sensory properties of the cueing process we observe want to tell us something about the challenges of flight control under more natural conditions. Cueing after-effect and attention span For the cueing after-effect (CAE) studied here and for the attention span in a previous study [6] the fly keeps the FoA for several seconds at a location to where it had shifted it. However, a cue is a signal from the outside world with a potential significance for the fly, whereas the attention span is a state maintained after the fly has actively shifted its FoA to a particular place. Still, similar processes at the neuronal or molecular level might underlie the two. For instance, both require a kind of working memory. Curiously, flies with a mutation in the radish gene showed a defect in both, the attention span and cueing. However, only the cueing is restored after feeding MPH (Fig 7 and [6]), indicating a different functional involvement of radish in the two processes. Are ICE and CAE the same process? We would like to assume that the ICE and the CAE are just the same process recorded at different times after the cue. If so, however, we need to find an explanation for the result of Fig 4B showing with repellent cueing in the upper visual field a CAE but no ICE. Whether indeed the fly has a special mechanism for repellent cueing in the UVF, which suppresses or delays the cueing effect for a second, remains to be confirmed. Genetic and pharmacological interventions in this study either interfered only with the CAE, or with both, CAE and ICE. Dopamine seemed to be mainly critical for the CAE while the ICE mostly remained undisturbed. In a few cases the manipulation of dopamine affected both, CAE and ICE. Two of them involved the mutant rsh- and/or the drug MPH. Both are known to have other targets besides dopamine [21, 24]. As complete genetic removal of dDAT in the fmn-/fmn- mutant (Fig 7) as well as reduced dopamine synthesis (Fig 6) leave the ICE intact one needs to search for other explanations for the suppression of the ICE in these cases. At present it has to remain open whether dopamine signaling is involved with cueing directly or just with its maintenance (CAE). SVA in mammals and the inverted-U hypothesis Selective visual attention in mammals has long been known to involve dopamine signaling (e.g. [22, 40]). We now provide evidence for the involvement of dopamine in cued shifts of attention in Drosophila. It will be interesting to work out and compare the basic principles as well as the behavioral properties of visual attention in mammals and flies. For instance, dopamine signaling in mammals is described as a balanced process between too much or too little transmitter. It is called the inverted-U hypothesis [41] describing the relationship between task-performance and dopamine levels. In the present study the cueing after-effect (CAE) and sometimes also immediate cueing (ICE) vanished with increasing local dopamine concentrations caused by blocking its re-uptake and also with decreasing dopamine levels blocking its synthesis. Superficially, these findings seem to comply with the inverted-U hypothesis. However, in the fly brain the increase is only extracellular whereas the decrease due to blocked synthesis is primarily intracellular and presynaptic. A further link between SVA in mammals and flies is the drug methylphenidate (MPH; commercial name: ritalin). [26] have argued that the altered short-term choice processes they found in the mutant rsh1 in a maze-walking paradigm might be a manifestation of impaired attention. Here we measured suppressed cueing of SVA in tethered flight in mutant rsh1 flies. The behavioral defects found in both paradigms could be rescued by feeding MPH, which is known to inhibit the re-uptake of dopamine from the synaptic cleft via dDAT [25, 26]. Once again, these findings suggest a parallel to attention in humans, where the drug is used to alleviate an attentional deficit (ADHD). The role of the αβp sub-compartment Several independent evidences support the conclusion that for the CAE a significant level of dDAT is required in the Kenyon cells of the αβ compartment of the MBs and that dDAT matters only there for the CAE. Already a reduction of dDAT in the 90 Kenyon cells of the αβp compartment seems to be sufficient to suppress the CAE. How the αβ compartment and in particular the αβp Kenyon cells are involved in the CAE remains to be studied in detail. Interestingly, in optophysiological recordings [38] showed that the αβc and αβs Kenyon cells respond to odors by depolarization whereas αβp neurons are hyperpolarized by the same stimuli. This is in line with the special anatomy of αβp. As mentioned in the Introduction olfactory input connections to the αβp fibers in the calyx are largely missing [10, 42]. Here for the first time we report a role of αβp in a visual task. Our data suggest that for the CAE the αβp sub-compartment is necessary but it might interact with one or both of the other two sub-compartments (αβc; αβs) as the restoration of dDAT in only the αβp neurons of the fmn-/+ heterozygote was not sufficient to rescue the CAE. In the present study dDAT was manipulated in the postsynaptic cells. The level of dDAT in these neurons has been shown to affect DA signaling at, and dDA1 receptor concentration in these cells [37]. The transporter is supposed to keep the synaptic cleft in dopaminergic synapses free of dopamine and to limit the sphere of dopaminergic influence [43]. The above conclusions should be confirmed using the dumb2 mutant and GAL4 driver lines for all three sub-compartments (αβc; αβs; αβp). Moreover, it should now be possible to identify the dopaminergic neurons innervating the relevant Kenyon cells and the pathway from the visual system to the MB [44, 45]. Selective visual attention equips visual systems of animals and humans with the fundamental ability to restrict their behavioral responses to stimuli in only part of their visual field. In flies one can now get at the underlying physiological and circuit mechanisms. Linking this process to dopamine in the αβp compartment of the MBs is a promising first step. Supporting Information S1 Fig With many visual stimuli cueing is repulsive. Broad stripes (w = 18°) and a large oscillation amplitude (Δψcue = 15°). Repulsive cueing is observed with inverted contrast, with male flies instead of females, black stripes on green background, flicker instead of oscillations, 5s oscillations instead of 1s, or just showing a different background color on one side for 1s as the cue. Flickering a grey stripe and just changing background color on one side elicit only an ICE but no CAE. (TIF) Click here for additional data file. S2 Fig Rescue and suppression of dDAT in mushroom body compartments has inconsistent effects on ICE. (A) OK107-Gal4 in heterozygous fmn- (dDATfmn/+) flies unexpectedly suppresses ICE. Additional dDAT expression in these cells rescues the suppression of the ICE (N = 38, 25). (B) Suppression of dDAT with RNAi in wild type in the same set of KCs leads to the absence of the ICE, emphasizing the importance of the MBs for this behavior (N = 32, 29). (C) MB247-Gal4 has only marginal expression in the α’β’-lobes. Otherwise the situation is the same as in (A). Ectopic dDAT expression in these cells rescues the suppression of the ICE (N = 47, 29, 18). (D) Same as in (B). (E) c739-Gal4 in dDATfmn/+ flies leaves ICE normal. Additional ectopic expression of dDAT in the αβ-lobes has no effect (N = 24, 27). (F) The necessity of the αβ-lobes for the ICE is demonstrated by the loss of ICE, if RNAi against dDAT-mRNA is expressed in the same set of cells in wild type (N = 27, 21). (G—J) dDAT modulation in c305-Gal4 (α’β’-lobes; N = 20, 25), NP1131-Gal4 (γ-lobes; N = 26, 19) and c708-Gal4 (αβp KCs; (N = 29, 23, 20; N = 23, 21) seems not to affect ICE. All error bars are SEMs (*P < 0.05, **P < 0.01, ***P < 0.001). (TIF) Click here for additional data file. This work was supported by the Rudolf Virchow Center and German Science Foundation, Reinhart Koselleck HE 986/20-1. We thank A. Haberberger (Rudolf-Virchow-Center) and S. Clemens-Richter (Department of Neurobiology and Genetics) for help with fly maintenance and experiments, K. Oechsner and H. Kaderschabek (Biocenter, University of Würzburg) for technical support, M. Krischke (University of Würzburg) for the HPLC measurements and J. Dubnau (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY), B. Gerber (Leibniz Institute for Neurobiology, Magdeburg, Germany), K. Kume (Nagoya University, Nagoya, Japan) and Scott Waddell (Oxford University, UK) for fly stocks. We thank the two reviewers for their constructive criticism. ==== Refs References 1 Wolf R , Heisenberg M . On the fine structure of yaw torque in visual flight orientation of Drosophila melanogaster. II. A temporally and spatially variable weighting function for the visual field (‘visual attention’) . J Comp Physiol A . 1980 ;140 (1 ):69 –80 . 10.1007/BF00613749 2 Heisenberg M , Wolf R . Vision in Drosophila—Genetics of Microbehavior . Braitenberg , editor. Berlin, Heidelberg New York, Tokyo : Springer Verlag ; 1984 . 252 p. 3 Tang S , Wolf R , Xu S , Heisenberg M . Visual pattern recognition in Drosophila is invariant for retinal position . Science . 2004 ;305 (5686 ):1020 –2 . 10.1126/science.1099839 .15310908 4 Tang S , Juusola M . Intrinsic activity in the fly brain gates visual information during behavioral choices . PLoS ONE . 2010 ;5 (12 ):e14455 10.1371/journal.pone.0014455 21209935 5 Sareen P , Wolf R , Heisenberg M . Attracting the attention of a fly . Proceedings of the National Academy of Sciences of the United States of America . 2011 ;108 (17 ):7230 –5 . 10.1073/pnas.1102522108 21482795 6 Koenig S , Wolf R , Heisenberg M . Vision in Flies: Measuring the Attention Span . PLoS ONE . 2016 ;11 (2 ):e0148208 10.1371/journal.pone.0148208 26848852 7 van Swinderen B . Attention in Drosophila . International review of neurobiology . 2011 ;99 :51 –85 . 10.1016/B978-0-12-387003-2.00003-3 .21906536 8 Warren RM , Warren RP . Helmholtz on perception: its physiology and development: Oxford, England : John Wiley + Sons ; 1968 . 277 p. 9 Aso Y , Grubel K , Busch S , Friedrich AB , Siwanowicz I , Tanimoto H . The mushroom body of adult Drosophila characterized by GAL4 drivers . Journal of neurogenetics . 2009 ;23 (1–2 ):156 –72 . 10.1080/01677060802471718 .19140035 10 Tanaka NK , Tanimoto H , Ito K . Neuronal assemblies of the Drosophila mushroom body . The Journal of comparative neurology . 2008 ;508 (5 ):711 –55 . 10.1002/cne.21692 .18395827 11 Martin JR , Ernst R , Heisenberg M . Mushroom bodies suppress locomotor activity in Drosophila melanogaster . Learn Mem . 1998 ;5 (1–2 ):179 –91 . 10454382 12 Liu L , Wolf R , Ernst R , Heisenberg M . Context generalization in Drosophila visual learning requires the mushroom bodies . Nature . 1999 ;400 (6746 ):753 –6 . 10.1038/23456 .10466722 13 Pascual A , Preat T . Localization of long-term memory within the Drosophila mushroom body . Science . 2001 ;294 (5544 ):1115 –7 . 10.1126/science.1064200 .11691997 14 Joiner WJ , Crocker A , White BH , Sehgal A . Sleep in Drosophila is regulated by adult mushroom bodies . Nature . 2006 ;441 (7094 ):757 –60 . 10.1038/nature04811 .16760980 15 Xi W , Peng Y , Guo J , Ye Y , Zhang K , Yu F , et al Mushroom bodies modulate salience-based selective fixation behavior in Drosophila . The European journal of neuroscience . 2008 ;27 (6 ):1441 –51 . 10.1111/j.1460-9568.2008.06114.x .18364023 16 de Belle JS , Heisenberg M . Associative odor learning in Drosophila abolished by chemical ablation of mushroom bodies . Science . 1994 ;263 (5147 ):692 –5 . .8303280 17 Heisenberg M , Wolf R . On the Fine Structure of Yaw Torque in Visual Flight Orientation of Drosophila melanogaster . J Comp Physiol A . 1979 ;130 (2 ):113 –30 . 10.1007/BF00611046 18 Amalric M , Moukhles H , Nieoullon A , Daszuta A . Complex deficits on reaction time performance following bilateral intrastriatal 6-OHDA infusion in the rat . The European journal of neuroscience . 1995 ;7 (5 ):972 –80 . .7613632 19 Elia J , Ambrosini PJ , Rapoport JL . Treatment of attention-deficit-hyperactivity disorder . N Engl J Med . 1999 ;340 (10 ):780 –8 . .10072414 20 Ye Y , Xi W , Peng Y , Wang Y , Guo A . Long-term but not short-term blockade of dopamine release in Drosophila impairs orientation during flight in a visual attention paradigm . The European journal of neuroscience . 2004 ;20 (4 ):1001 –7 . 10.1111/j.1460-9568.2004.03575.x .15305868 21 Volkow ND , Wang GJ , Kollins SH , Wigal TL , Newcorn JH , Telang F , et al Evaluating dopamine reward pathway in ADHD: clinical implications . Jama . 2009 ;302 (10 ):1084 –91 . 10.1001/jama.2009.1308 19738093 22 Van Swinderen B , Andretic R . Dopamine in Drosophila: setting arousal thresholds in a miniature brain . Proceedings Biological sciences / The Royal Society . 2011 ;278 (1707 ):906 –13 . 10.1098/rspb.2010.2564 21208962 23 Calipari ES , Beveridge TJ , Jones SR , Porrino LJ . Withdrawal from extended-access cocaine self-administration results in dysregulated functional activity and altered locomotor activity in rats . The European journal of neuroscience . 2013 ;38 (12 ):3749 –57 . 10.1111/ejn.12381 24118121 24 Porzgen P , Park SK , Hirsh J , Sonders MS , Amara SG . The antidepressant-sensitive dopamine transporter in Drosophila melanogaster: a primordial carrier for catecholamines . Molecular pharmacology . 2001 ;59 (1 ):83 –95 . .11125028 25 Iversen SD , Iversen LL . Dopamine: 50 years in perspective . Trends in neurosciences . 2007 ;30 (5 ):188 –93 . 10.1016/j.tins.2007.03.002 .17368565 26 van Swinderen B , Brembs B . Attention-like deficit and hyperactivity in a Drosophila memory mutant . J Neurosci . 2010 ;30 (3 ):1003 –14 . 10.1523/JNEUROSCI.4516-09.2010 .20089909 27 Solanki N , Wolf R , Heisenberg M . Central complex and mushroom bodies mediate novelty choice behavior in Drosophila . Journal of neurogenetics . 2015 ;29 (1 ):30 –7 . 10.3109/01677063.2014.1002661 .25585638 28 Draper I , Kurshan PT , McBride E , Jackson FR , Kopin AS . Locomotor activity is regulated by D2-like receptors in Drosophila: an anatomic and functional analysis . Dev Neurobiol . 2007 ;67 (3 ):378 –93 . .17443795 29 Kim YC , Lee HG , Han KA . D1 dopamine receptor dDA1 is required in the mushroom body neurons for aversive and appetitive learning in Drosophila . J Neurosci . 2007 ;27 (29 ):7640 –7 . 10.1523/JNEUROSCI.1167-07.2007 .17634358 30 Seugnet L , Suzuki Y , Vine L , Gottschalk L , Shaw PJ . D1 receptor activation in the mushroom bodies rescues sleep-loss-induced learning impairments in Drosophila . Current biology: CB . 2008 ;18 (15 ):1110 –7 . 10.1016/j.cub.2008.07.028 18674913 31 Lebestky T , Chang JS , Dankert H , Zelnik L , Kim YC , Han KA , et al Two different forms of arousal in Drosophila are oppositely regulated by the dopamine D1 receptor ortholog DopR via distinct neural circuits . Neuron . 2009 ;64 (4 ):522 –36 . 10.1016/j.neuron.2009.09.031 19945394 32 Srivastava DP , Yu EJ , Kennedy K , Chatwin H , Reale V , Hamon M , et al Rapid, nongenomic responses to ecdysteroids and catecholamines mediated by a novel Drosophila G-protein-coupled receptor . J Neurosci . 2005 ;25 (26 ):6145 –55 . 10.1523/JNEUROSCI.1005-05.2005 .15987944 33 Sugamori KS , Demchyshyn LL , McConkey F , Forte MA , Niznik HB . A primordial dopamine D1-like adenylyl cyclase-linked receptor from Drosophila melanogaster displaying poor affinity for benzazepines . FEBS letters . 1995 ;362 (2 ):131 –8 . .7720859 34 Feng G , Hannan F , Reale V , Hon YY , Kousky CT , Evans PD , et al Cloning and functional characterization of a novel dopamine receptor from Drosophila melanogaster . J Neurosci . 1996 ;16 (12 ):3925 –33 . .8656286 35 Kim YC , Lee HG , Seong CS , Han KA . Expression of a D1 dopamine receptor dDA1/DmDOP1 in the central nervous system of Drosophila melanogaster . Gene Expr Patterns . 2003 ;3 (2 ):237 –45 . .12711555 36 Karpova EK , Bogomolova EV , Romonova IV , Gruntenko NE , Raushenbakh I . [Role of DopR in the molecular mechanism of the dopamine control of juvenile hormone metabolism in female Drosophila ]. Genetika . 2012 ;48 (8 ):999 –1002 . .23035552 37 Ueno T , Kume K . Functional characterization of dopamine transporter in vivo using Drosophila melanogaster behavioral assays . Frontiers in behavioral neuroscience . 2014 ;8 :303 10.3389/fnbeh.2014.00303 25232310 38 Perisse E , Yin Y , Lin AC , Lin S , Huetteroth W , Waddell S . Different kenyon cell populations drive learned approach and avoidance in Drosophila . Neuron . 2013 ;79 (5 ):945 –56 . 10.1016/j.neuron.2013.07.045 24012007 39 Heisenberg M . Outcome learning, outcome expectations, and intentionality in Drosophila . Learn Mem . 2015 ;22 (6 ):294 –8 . 10.1101/lm.037481.114 25979991 40 Nieoullon A . Dopamine and the regulation of cognition and attention . Progress in neurobiology . 2002 ;67 (1 ):53 –83 . .12126656 41 Cools R , Robbins TW . Chemistry of the adaptive mind . Philos Trans A Math Phys Eng Sci . 2004 ;362 (1825 ):2871 –88 . 10.1098/rsta.2004.1468 .15539374 42 Lin HH , Lai JS , Chin AL , Chen YC , Chiang AS . A map of olfactory representation in the Drosophila mushroom body . Cell . 2007 ;128 (6 ):1205 –17 . 10.1016/j.cell.2007.03.006 .17382887 43 Cragg SJ , Rice ME . DAncing past the DAT at a DA synapse . Trends in neurosciences . 2004 ;27 (5 ):270 –7 . 10.1016/j.tins.2004.03.011 .15111009 44 Aso Y , Hattori D , Yu Y , Johnston RM , Nirmala A , Ngo T , et al The neuronal architecture of the mushroom body provides a logic for associative learning . eLife . 2014 :1 –47 . 45 Vogt K , Aso Y , Hige T , Knapek S , Ichinose T , Friedrich AB , et al Direct neural pathways convey distinct visual information to Drosophila mushroom bodies . eLife . 2016 ;5 10.7554/eLife.14009 27083044

PMC005xxxxxx/PMC5003350.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757151510.1371/journal.pone.0161345PONE-D-16-17555Research ArticleBiology and Life SciencesImmunologyVaccination and ImmunizationVaccinesMedicine and Health SciencesImmunologyVaccination and ImmunizationVaccinesMedicine and Health SciencesPublic and Occupational HealthPreventive MedicineVaccination and ImmunizationVaccinesBiology and Life SciencesImmunologyVaccination and ImmunizationMedicine and Health SciencesImmunologyVaccination and ImmunizationMedicine and Health SciencesPublic and Occupational HealthPreventive MedicineVaccination and ImmunizationMedicine and Health SciencesInfectious DiseasesViral DiseasesRotavirus InfectionBiology and Life SciencesBiogeographyPhylogeographyEcology and Environmental SciencesBiogeographyPhylogeographyEarth SciencesGeographyBiogeographyPhylogeographyBiology and Life SciencesEvolutionary BiologyPopulation GeneticsPhylogeographyBiology and Life SciencesGeneticsPopulation GeneticsPhylogeographyBiology and Life SciencesPopulation BiologyPopulation GeneticsPhylogeographyBiology and life sciencesOrganismsVirusesRNA virusesReovirusesRotavirusBiology and Life SciencesMicrobiologyMedical MicrobiologyMicrobial PathogensViral PathogensReovirusesRotavirusMedicine and Health SciencesPathology and Laboratory MedicinePathogensMicrobial PathogensViral PathogensReovirusesRotavirusBiology and Life SciencesOrganismsVirusesViral PathogensReovirusesRotavirusPeople and PlacesGeographical LocationsAsiaLebanonMedicine and Health SciencesGastroenterology and HepatologyGastroenteritisPeople and PlacesPopulation GroupingsAge GroupsChildrenPeople and PlacesPopulation GroupingsFamiliesChildrenRotavirus Genotypes and Vaccine Effectiveness from a Sentinel, Hospital-Based, Surveillance Study for Three Consecutive Rotavirus Seasons in Lebanon Gastroenteritis Burden of Rotavirus in ChildrenAli Zainab 12Harastani Houda 23Hammadi Moza 12Reslan Lina 23Ghanem Soha 12Hajar Farah 12Sabra Ahmad 23Haidar Amjad 23Inati Adlette 4Rajab Mariam 5Fakhouri Hassan 6Ghanem Bassam 7Baasiri Ghassan 8Gerbaka Bernard 9Zaraket Hassan 210Matar Ghassan M. 210Dbaibo Ghassan 123*1 Department of Pediatrics and Adolescent Medicine, American University of Beirut, Faculty of Medicine, Beirut, Lebanon2 Center for Infectious Diseases Research, American University of Beirut, Faculty of Medicine, Beirut, Lebanon3 Department of Biochemistry and Molecular Genetics, American University of Beirut, Faculty of Medicine, Beirut, Lebanon4 Department of Pediatrics, Nini Hospital, Tripoli, Lebanon5 Department of Pediatrics, Makassed General Hospital, Beirut, Lebanon6 Department of Pediatrics, Rafic Hariri University Hospital, Beirut, Lebanon7 Department of Pediatrics, Nabatieh Governmental Hospital, Nabatieh, Lebanon8 Department of Pediatrics, Hammoud Hospital, Saida, Lebanon9 Department of Pediatrics, Hôtel-Dieu de France University Hospital, Beirut, Lebanon10 Department of Experimental Pathology, Immunology and Microbiology, Faculty of Medicine, American University of Beirut, Beirut, LebanonSestak Karol EditorTulane University, UNITED STATESCompeting Interests: GD has received honoraria for lectures and grant funding through my institution from MSD, GSK, Sanofi-Pasteur, Hikma, and Pfizer. Rotateq® is commercialized by MSD and Rotarix® by GSK. This does not alter our adherence to PLOS ONE policies on sharing data and materials. Conceptualization: GD SG FH. Data curation: GD ZA MH AS AI MR HF B. Ghanem B. Gerbaka GB. Formal analysis: GD ZA HH LR. Funding acquisition: GD. Investigation: GD ZA MH AS AI MR HF B. Ghanem B. Gerbaka GB. Methodology: GD GM HZ HH LR. Project administration: GD AH. Resources: AH. Supervision: GD HZ GM. Validation: LR HH AH. Writing – original draft: GD ZA HH SG FH HZ LR. Writing – review & editing: GD GM HZ ZA HH LR AI. * E-mail: gdbaibo@aub.edu.lb; ghassa.dbaibo.2016@gmail.com29 8 2016 2016 11 8 e016134517 5 2016 3 8 2016 © 2016 Ali et al2016Ali et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Introduction Globally, rotavirus (RV) is the leading cause of gastroenteritis (GE) in children. Longitudinal data about changes in RV genotype distribution and vaccine effectiveness (VE) are scarce. This study was conducted in Lebanon over 3 consecutive RV seasons to estimate the rate of RVGE hospitalization, identify RV genotypes, determine the seasonal and geographical variations, and calculate RV VE. Materials and Methods This prospective, multicenter, hospital-based surveillance study was conducted between 2011 and 2013 and enrolled children (<5 years) admitted for GE. Socio-demographic and clinical data about the current episode of GE at admission were collected. Genotypes were determined from stool samples testing positive for RV by PCR. Results Of 1,414 cases included in the final analysis, 83% were <2 years old and 55.6% were boys. Median duration of hospitalization was 4 days and 91.6% of GE cases were severe (Vesikari score ≥11). PCR testing showed that 30.3% of subjects were RV-positive of which 62.1% had fever versus 71.1% of RV-negative subjects (P = 0.001). RV was predominantly detected in the cold season from November till March (69.9%). G and P genotype pairs for all RV-positive stool specimens showed a predominance of G1P[8] in 36% (n = 154) of specimens, G9P[8] in 26.4% (n = 113), and G2P[4] in 17.8% (n = 76). RV-negative subjects were more likely to be RV-vaccinated (21%) compared to the RV-positive subjects (11.3%) (P<0.001), with a vaccine breakthrough rate of 18.8%. The ratio of RV1-vaccinated for each RV5-vaccinated subject was 7.8 and VE against RV disease was 68.4% (95%CI, 49.6%-80.2%). Conclusion RV is a major cause of GE requiring hospitalization of children under 5 years of age in Lebanon. A few genotypes predominated over the three RV seasons studied. Mass RV vaccination will likely decrease the burden of hospitalization due to RV. VE is similar to what has been observed for other middle-income countries. http://dx.doi.org/10.13039/100004334MerckDbaibo Ghassan This study was funded by Merck Sharp & Dohme Corp. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityRaw data in the form of specimen collection records listed in a spreadsheet are provided in the supplementary files of this manuscript (S1 Table). IUCN Red List assessments of all described species are freely available at www.iucnredlist.org<http://www.iucnredlist.org>. Additional data will be submitted to Orthoptera Species File Online orthoptera.speciesfile.org at regular intervals in the coming years. All data are stored in PN's MANTIS database and are freely available upon request to pnaskrecki@oeb.harvard.edu.Data Availability Raw data in the form of specimen collection records listed in a spreadsheet are provided in the supplementary files of this manuscript (S1 Table). IUCN Red List assessments of all described species are freely available at www.iucnredlist.org<http://www.iucnredlist.org>. Additional data will be submitted to Orthoptera Species File Online orthoptera.speciesfile.org at regular intervals in the coming years. All data are stored in PN's MANTIS database and are freely available upon request to pnaskrecki@oeb.harvard.edu. ==== Body Introduction Rotavirus (RV) is the leading cause of severe gastroenteritis (GE) in infants and young children around the world [1]. Rotavirus is a double-stranded RNA virus belonging to the Reoviridae family. It is a complex particle with viral genome segments encoding 6 structural viral proteins (VPs), which make up the virus particles, and 6 non-structural proteins involved in viral virulence. The outer capsid is made of VP7 and VP4 proteins, the latter responsible for the capsid’s spikes. Both proteins are neutralizing antigens in the immune response to RV infection. In the early 1990’s, genotyping was introduced based on sequences of the VP7 and VP4 genes, and has now replaced serotyping as the preferred method of typing. P genotypes are denoted in brackets (P[8], P[4]…) whereas serotypes are denoted in parentheses, for example P(8) [2]. RV infection can range from being asymptomatic to a severe symptomatic one [3]. The main clinical symptoms of RV infections are fever, vomiting, and diarrhea in infants and young children. The outcome of the infection is affected by both viral and host factors, and the major host factor determinant is the patient’s age. The age group most susceptible to severe RV infection ranges from 3 months to 2 years. RV infection is generally asymptomatic in neonates, since they are protected by transplacental transfer of maternal antibodies [3]. As for adults, the incidence of RV infection is low and symptoms are unlikely to be severe although outbreaks in adolescents and adults have been described [3]. The first human rotaviruses were discovered in 1973 [4,5]. Soon afterwards, RV was found to be associated with 20 to 30% of severe diarrheal cases that required hospitalizations in children younger than 5 years of age worldwide [6]. Globally, it has been estimated that there are more than 114 million RV diarrheal episodes per year requiring home care with an additional 24 million clinic visits, 2.4 million hospitalizations, and more than 600,000 deaths in children younger than 5 years of age [7]. Additional epidemiological studies have estimated that by 5 years of age, one in 50 children is hospitalized and one in 205 children dies from RV-associated causes [7,8]. In Western Europe, prior to the vaccine era, RV was responsible of 230 deaths per year [9,10]. Pre-vaccine epidemiological studies also reported that 50% of cases of gastroenteritis in children younger than 5 years of age who were treated in the emergency room were caused by RV in the USA and Western Europe [9,10,11]. Consequently, one out of 150 children younger than 3 years of age would be hospitalized and 1 out of 11 would be seen as an outpatient in an emergency room for RV disease [11]. Importantly, RV is a common hospital-acquired infection where 20% of children admitted to the hospital for another cause would acquire a RV infection during their stay throughout the year [11,12]. Furthermore, the burden of the RV infection is not limited to morbidity and mortality, but is also associated with significant economic losses. The latter exceeds 1 billion US dollars per year in USA alone, prior to vaccine introduction [13]. RV is common in temperate and tropical climates, as well as in developed and less-developed social settings. Despite the policies to improve the hygiene levels in these countries, decreasing the transmission of RV is impossible because of many factors. First, rotaviruses are shed in feces in amounts up to 1010 particles per gram of stool [14]. Second, the infectious dose has been estimated to be as low as one infectious virus particle. RV is also highly resistant to commonly used disinfectants and will survive for several days on fomites [15]. RV infections vary throughout the year in many regions of the world. Nevertheless, RV disease seems to be seasonal in temperate climates, occurring in the dry and cool months and less likely in warm and humid months of the year [14,16]. No specific treatment of RV infections is available and only supportive care to prevent dehydration and its complications is provided where available. Vaccination is the only effective prevention strategy. Bovine rotavirus RIT 4237 vaccine was initially developed but was never marketed [17]. A simian RV reassortant vaccine (Rotashield®) was developed and marketed in August 1998. However, it was withdrawn from the market in October 1999 because of its association with a high incidence of intussusception [18]. In 2006 and 2007, two oral vaccines for RV GE (Rotateq® and Rotarix®) were licensed in many countries. Rotateq® (Rotavirus Vaccine, Live, Oral, Pentavalent, Merck Sharp & Dohme corp., USA) is a human-bovine reassortant vaccine which was found to be highly efficacious with a protection rate of 74% against diarrhea of any severity [19]. Rotarix® (Rotavirus Vaccine, Live, Oral, GlaxoSmithKline Biologicals, UK) is a human G1P[8] RV vaccine. Rotarix® was 85% effective against preventing severe diarrhea in a large cohort study [19,20]. Both vaccines were 100% effective against the most severe cases [19]. Epidemiological data about RV infections are scarce in Lebanon. We recently published the results of a hospital-based surveillance study in five large urban, private hospitals in North, Central and South Lebanon, conducted between April 2007 and September 2008 essentially covering one RV season [21]. Baseline data are essential to understand the burden of the disease and to make suitable recommendations to control RV infections in the country. Therefore, our current study is the second to report on the epidemiology of this disease in Lebanon but covering three consecutive RV seasons. Our main goal was to assess the rate of RV infections in a sentinel surveillance system in the Lebanese pediatric population (less than 5 years). The secondary objectives were (1) to estimate the rate of diarrheal hospitalization due to RV, (2) to determine the age and the seasonal distribution of RV-associated hospitalization, (3) to identify the most prevalent RV genotypes, (4) to determine the seasonal changes and year-to-year changes with respect to predominant RV genotypes in the target population, and (5) to assess the vaccine effectiveness (VE) of RV1 and RV5 against rotavirus disease. The results are expected to help policy makers understand the importance and magnitude of RV infections so that vaccine recommendations can be formulated accordingly. Material and Methods Study procedure Our study was a prospective, multicenter, hospital-based surveillance in seven medical centers distributed in North, Central, and South Lebanon. It was conducted over 30 months from January 2011 through June 2013. Eligible participants were children below 5 years of age, admitted mainly for GE, to the designated centers. Infants and children hospitalized more than once were considered as new participants at each new admission. Also, a stay in the emergency department longer than 12 hours was considered a hospitalization. To be enrolled in the study, the child’s parents or guardians had to be believed by the investigators to comply with the protocol’s requirements. A written informed consent was obtained from the parents or guardians. Infants and the children whose admission diagnosis did not include GE or whose onset of GE occurred more than 12 hours after admission to the hospital (hospital-acquired infections) were excluded. Data collection Data were collected by the investigators using a standardized English case report form (CRF) (S1 CRF). Information collected during hospitalization included data about age, gender, area of residence, height and weight of the patients, their past medical history and their current episode of GE at admission (body temperature, duration of diarrhea or vomiting). The severity of RV gastroenteritis was assessed using the Vesikari scale [22]. Information about any prior treatments, rehydration therapy, rotavirus vaccination, and antibiotics were also reported. Further, the date, the diagnosis at discharge and the outcome of GE (recovery, recovery with sequelae, death, transfer to another hospital, ongoing or unknown) were recorded. Laboratory analysis One stool sample was collected from each patient in sterile stool sample containers preferably within four days and not later than 10 days after the onset of GE symptoms. The detection of rotavirus antigen was tested using the SD Bioline Rotavirus rapid kit (Standard Diagnostics, INC., Republic of Korea). This test was performed according to the manufacturer’s specifications. All RV-positive samples were processed for RNA extraction and genes sequencing. Double-stranded viral RNA was extracted from 10% stool suspensions that were made by adding 0.5 g or 500 μl of the fecal sample to 5 ml of NaCl solution (0.89%). The homogenate was clarified by centrifugation at 4,000 g at 4°C for 20 minutes. The supernatant was re-centrifuged at 1,500 g at 4°C for 10 minutes. RNA was extracted from 420 μl of the clarified supernatant using the QIAamp® Viral RNA Mini Kit (Qiagen, Hilden, Germany) in accordance with the manufacturer’s spin protocol. RNA was eluted in 45 μl RNase-free distilled water and stored at -20°C. For specimens that were negative by antigen detection, every twentieth sample collected chronologically was subjected to real-time reverse-transcription polymerase chain reaction (RT-PCR) testing to confirm the absence of rotavirus. Double-stranded RNA was denatured at 97°C for 5 minutes and reverse-transcribed and amplified using the Qiagen One-Step RT-PCR Kit (Qiagen, Hilden, Germany). Previously published primers were used for the amplification of genes encoding for VP7 genes (nt: 37–932) and VP4 genes (nt: 11–887) [23]. The target region of each gene was reverse-transcribed at a temperature of 42°C for 30 minutes followed by deactivation of the reverse-transcriptase at 95°C for 15 minutes. Amplification of each gene was performed on a C1000 thermal cycler (Bio-Rad, Inc., Berkeley, California, USA) using the following cycle parameters: 30 cycles of 94°C for 30 seconds, 42°C for 30 seconds, and 72°C for 45 seconds, and final extension at 72°C for 7 minutes. Amplicons were detected by gel electrophoresis using a 1.5% agarose gel and analyzed and photographed by a gel documentation system (Gel doc XR, Bio-Rad, Berkeley, California, USA). RT-PCR products were cleaned using ExoSAP-IT® according to manufacturer’s instructions (USB Corp., Cleveland, OH, USA) and used directly for sequencing. Sequencing was performed at Macrogen institution (Seoul, Republic of Korea). The sequence data were analyzed by using BioEdit v7.2.5 software. The genotype assignment for each gene was performed using the BLAST (Basic Local Alignment Search Tool) server on the GenBank database at the National Center for Biotechnology Information (NCBI) and the online RotaC v2.0 rotavirus genotyping tool (available at: http://rotac.regatools.be). Data analysis The total population of children aged below 5 years of age in Lebanon was 219,150 during the study conduct, and about 21% of this population was to be served by the seven participating study centers as estimated by the Lebanese Ministry of Public Health and the Mundi Index [24]. Therefore, the total population of children in the same age category and served by the designated centers was estimated at 46,000. Moreover, the total number of GE cases admitted to the study centers were estimated as 800 per year. As a result, the incidence of GE hospitalization in the study sample was expected to be 1.74 (800 over 46,000). Based on 30-month enrollment period, 115,000 children-years were to be followed up, of whom 2,000 GE episodes were expected with all children participating in the study over a period of two and a half years. Assuming that 20 to 40% of these GE episodes would be RV-positive GE, approximately 400 to 800 RV-positive GE episodes would be possible to be captured by year. Thus, 2,000 children were targeted in this study considering that all episodes of GE hospitalizations at the participating hospitals were to be enrolled over two and a half years. Records for participants with incomplete, insufficient or conflicting data were excluded. Sample characteristics were summarized using the mean and the standard deviation (SD) for continuous variables such as age; frequency distributions for categorical variables such as gender were presented. Incidence rates were calculated using descriptive data, along with its corresponding 95% confidence interval (CI). Season-to-season variation in RV infections and predominant rotavirus genotypes were described by geographical areas and seasons. Season 1 extended from January 2011 to August 2011, season 2 from September 2011 to August 2012, and season 3 from September 2012 to June 2013. Characteristics of the children were also compared using the chi-square test and Fisher Exact test as appropriate. As for vaccine effectiveness, it was computed along with its 95% CI as (1-odds ratio [OR]) × 100%, by logistic regression. The logistic regression had RV-positive and RV-negative as outcome and vaccination status as explanatory variables. ORs were adjusted by geographic region and season. Patients were assumed to have the proper vaccination if they had two documented RV1 vaccination or 3 documented RV5 doses all 14 days before hospital admission. Significance level was two-sided and set at 5%. The statistical analysis was carried out using IBM-SPSS version 21 software for Windows Release (IBM Corp. Released 2012. IBM SPSS Statistics for Windows, Version 21.0. Armonk, NY: IBM Corp.). Ethical considerations The study was conducted according to Good Clinical Practice, and the 1996 version of the Declaration of Helsinki. It was also approved by the institutional review boards (IRBs) of each of the participating centers (IRB of the American University of Beirut, IRB of Nini Hospital, IRB of Makassed General Hospital, IRB of Rafic Hariri University Hospital, IRB of Nabatieh Governmental Hospital, Nabatieh, Lebanon, IRB of Hammoud Hospital, and Ethics Committee of Hôtel-Dieu de France University Hospital) before initiating the study. Information about the infants and the children whose parents or guardians accepted to participate in the study were collected after signing the written informed consent. Participants’ numbers were assigned sequentially to all children whose parents or guardians signed the consent form according to the range of the children’s numbers assigned to every center. Data collection and analysis were performed respecting the participants’ autonomy and anonymity. Results Socio-demographic and clinical characteristics of the patients diagnosed with GE The distribution of the demographic and clinical characteristics of the children hospitalized for GE according to their RV status is displayed in Table 1. Over the 30-month surveillance program, 1,555 children were hospitalized at the seven study centers for GE and were enrolled in the study but 141 were excluded due to inability to obtain a stool sample or due to loss of the case report form during transport to the main center. The remaining 1,414 children were included in the final analysis. Of the 1,414 fecal samples tested, 30.3% (n = 428, 95%CI, 27.9%-32.7%) were RV-positive by PCR. Of the samples that were negative for RV by immunoassay and were subsequently screened by PCR, none were positive. 10.1371/journal.pone.0161345.t001Table 1 Distribution of demographic and clinical outcomes of patients according to RV status. Variable Overall RV-positive RV-negative P-value* (n = 1,414†) (n = 428†) (n = 986†) Center 0.024*  AUBMC 157 (11.1%) 44 (28.0%) 113 (72.0%) 0.030*£  Hammoud Hospital 209 (14.8%) 83 (39.7%) 126 (60.3%)  Hôtel-Dieu de France University Hospital 5 (0.4%) 0 (0.0%) 5 (100.0%)  Makassed Hospital 189 (13.4%) 47 (24.9%) 142 (75.1%)  Nini Hospital 556 (39.3%) 166 (29.9%) 390 (70.1%)  Nabatieh Hospital 212 (15.0%) 65 (30.7%) 147 (69.3%)  Rafic Hariri University Hospital 86 (6.1%) 23 (26.7%) 63 (73.3%) Age (months) 0.984  0–2 93 (6.6%) 30 (32.3%) 63 (67.7%)  3–6 268 (19.0%) 79 (29.5%) 189 (70.5%)  7–12 461 (32.6%) 137 (29.7%) 324 (70.3%)  13–24 364 (25.7%) 112 (30.8%) 252 (69.2%)  25–60 228 (16.1%) 70 (30.7%) 158(69.3%) Male gender 786 (55.6%) 235 (29.9%) 551 (70.1%) 0.734 Living in Lebanon 1350 (95.5%) 413 (30.6%) 937 (69.4%) 0.252 Vesikari score>10 1,295 (91.6%) 398(93.0%) 898(91.1%) 0.231 Symptoms  Fever 966 (68.3%) 266 (62.1%) 700 (71.1%) 0.001*  Vomiting 889 (62.9%) 327 (76.4%) 562 (57.0%) <0.001*  Diarrhea 1,288 (91.1%) 385 (90.0%) 903 (91.7%) 0.295 Treatments during hospitalization and at discharge‡  Oral rehydration 234 (16.6%) 85 (19.9%) 149 (15.1%) 0.028*  IV rehydration 1,379 (97.5%) 416(97.2%) 963(97.7%) 0.600  Antibiotics 698 (49.4%) 173(40.4%) 525(53.2%) <0.001* Vaccination history  Yes (any type and dose) 255 (18.1%) 48(11.2%) 207 (21.0%) <0.001  RV1 (Rotarix®)  1 dose 54 (25.0%) 13 (32.5%) 41 (23%)  2 doses 165 (75.0%) 27 (67.5%) 138 (77%)  RV5 (Rotateq®) 0.973  1 dose 8 (28.5%) 2 (28.5%) 6 (28.6%)  2 doses 3 (10.7%) 1 (14.3%) 2 (9.5%)  3 doses 17 (60.8%) 4 (57.2%) 13 (61.9%)  Unknown vaccine 2 (0.1%) 2 (0.5%) 0 (0.0%)  Unvaccinated 1,157 (81.8%) 378 (88.3%) 779 (79.0%) Duration of hospitalization (days)‡ 0.106  0–3 537 (38.0%) 145 (33.9%) 392 (39.8%)  4–7 805 (56.9%) 261 (61.0%) 544 (55.2%)  >7 72 (5.1%) 22 (5.1%) 50 (5.1%) †The sample size per column might not add to the total of that column depending on whether there are some missing data or not. *Significant at the 5% level. £P-value excluding Hôtel-Dieu de France University Hospital from the comparison. ‡Percentages are computed within each column (i.e. percentage of treatment within each group of RV+ and RV-). Abbreviations: AUBMC = American University of Beirut Medical Center. Overall, age-specific rates of the patients with GE rose sharply from age zero to 7 months (25.6%), peaked in the 7–12 months age group (32.6%), and subsequently dropped to (25.7%) for those aged 13–24 months and (16.1%) for those aged 25–60 months with no significant difference in age-specific rates between the RV-positive and RV-negative groups. In both groups, 83% of the patients were younger than 2 years of age, 55.6% were boys and only 4.5% were not living in Lebanon. The median duration of hospitalization was 4 days and ranged between 1 and 73 days with vast majority of GE cases (91.6%) being severe according to Vesikari score of 11 or above. The difference in the duration of hospitalization between the RV-positive and RV-negative children was not statistically significant. Patients with RV-positive GE significantly differed from the RV-negative group by fever and vomiting, treatments during hospitalization and at discharge (oral rehydration and antibiotics), year of hospitalization, and RV season. As such, fewer RV-positive patients had fever compared to those who were RV-negative (62.1% versus 71.1%, P = 0.001); more RV-positive children (19.9%) were treated with oral rehydration while only 15.1% were offered the same treatment in the RV-negative group (P = 0.03) probably related to the increased incidence of vomiting in the former group (76.4% vs 57%, P <0.001). 53.2% of children with RV-negative GE were given antibiotics compared to 40.4% in the RV-positive group (P<0.001). Additionally, the incidence of RV-positive GE associated hospitalization reached its maximum (53.8%) in 2013 while RV-negative GE (75.7%) was more frequent in 2011 (P<0.001). Overall, total GE peaked in the month of July (12.3%). RV was encountered throughout the 30-month study period, but was predominantly detected in the cold season from November till March (69.9%) (Table 1 and Fig 1). 10.1371/journal.pone.0161345.g001Fig 1 Season-to-season variation in RV infections and predominant rotavirus genotypes (in frequency) across seasons between January 2011 and June 2013 (n = 428). RV genotyping G and P genotypes of rotavirus were determined in the 428 RV-positive stool specimens out of the total 1,414 samples (30.3%) by sequencing of the VP7 and VP4 genes, respectively. There was a predominance of G1P[8] RV accounting for 36% (n = 154) of specimens, G9P[8] RV was responsible for 26.4% (n = 113) of cases, G2P[4] for 17.8% (n = 76) and G4P[8] for 15.9% (n = 68) (Fig 2). 10.1371/journal.pone.0161345.g002Fig 2 Distribution of the RV genotypes among all RV-positive specimens (n = 428). Furthermore, RV genotype predominance differed significantly across seasons. As such, G9P[8] was predominant in seasons 1 and 3, while G1P[8] was the major genotype in season 2 (P<0.001). As for G2P[4], it was more frequent in season 2 compared to seasons 1 and 3 (P<0.001), and G4P[8] was more encountered in seasons 1 and 2 rather than season 3 (P = 0.004) (Fig 1). RV genotypes were also different by season and across geographical areas of the study centers. In season 1, G9P[8] (n = 25, 16.4%) and G1P[8] (n = 23, 34.3%) dominated in Beirut, while the most common genotypes were G1P[8] (n = 14, 46.6%) in South Lebanon, G4P[8] (n = 25, 41.6%) and G1P[8] (n = 20, 29%) in North Lebanon. G4P[8] disappeared from the North in seasons 2 and 3 and G2P[4] did not circulate in Beirut in season 3 although it was strongly present in South Lebanon during the same season (Fig 3). Overall, G1P[8] was significantly predominant in North Lebanon (P = 0.007) all seasons combined, while G2P[4] predominated in the South (P<0.001). In parallel, G9P[8] was the most frequent genotype in North Lebanon with no significant predominance (P = 0.186). The same pattern applies for G4P[8] in the South (P = 0.344). Overall, and despite the small area of Lebanon, there was seasonal and geographic variation in the distribution of genotypes (Fig 3). 10.1371/journal.pone.0161345.g003Fig 3 Distribution of rotavirus genotypes (in frequency) by season and geographical area of study center (n = 428). Geographical areas of study centers: Beirut: American University of Beirut Medical Center, Makassed Hospital and Rafic Hariri University Hospital; North: Nini Hospital; South: Hammoud Hospital and Nabatieh Hospital. Seasons: Season 1: January 2011 to August 2011, Season 2: September 2011 to August 2012, and Season 3: September2012 to June 2013. All genotypes appeared equally in all age-groups examined (Fig 4). The Vesikari score did not differ significantly between the different genotypes (Table 2). 10.1371/journal.pone.0161345.g004Fig 4 Distribution rotavirus genotypes by age group (n = 428). 10.1371/journal.pone.0161345.t002Table 2 Vesikari score for GE associated with the common rotavirus genotypes in vaccinated versus unvaccinated patients. Predominant rotavirus genotype Vaccinated Unvaccinated P-value* N N Mean ± standard deviation Mean ± standard deviation All genotypes 48 378 0.339 13.81 ± 2.70 14.19 ± 2.54 G1P[8] 11 141 0.950 14.45 ± 2.11 14.50 ± 2.50 G2P[4] 15 61 0.420 13.47 ± 2.77 13.89 ± 2.65 G4P[8] 9 59 0.088 12.56 ± 1.81 14.07 ± 2.52 G9P[8] 11 102 0.696 14.45 ± 3.42 14.03 ± 2.49 *Significant at the 5% level. Vaccination status of patients diagnosed with GE and vaccine effectiveness Only 18.1% (n = 255) of the children hospitalized for GE were vaccinated against rotavirus. Patients who were RV-negative were more likely to be vaccinated as compared to the RV-positive children (21% versus 11.3%, P<0.001). Moreover, 219 patients were vaccinated with RV1 and 28 with RV5 at a ratio of 7.8 RV1-vaccinated patient for each RV5-vaccinated patient. A 2-dose series of RV1 had been given to 6.4% (n = 27) of RV-positive patients versus 14.1% (n = 138) of RV-negative children (P<0.001) (Table 1). Regardless of their RV status, the 255 vaccinated patients hospitalized for GE significantly differed from the 1,157 unvaccinated cohort by study center, age category and duration of hospitalization. The highest rate of vaccination (20.6%) was in the category of 13–24 months whereas, understandably, the highest rate of unvaccinated children diagnosed and hospitalized for GE was in the category of 0–2 months (P = 0.004). The rate of RV GE in the vaccinated cohort was significantly lower than the unvaccinated cohort (18.8% vs 32.7%, P<0.001). We reasoned that vaccinated patients might have lower disease severity due to at least partial protection provided by the vaccine. Nevertheless, when the Vesikari severity score was compared between vaccinated and unvaccinated patients, there was no statistical difference when analyzed as individual genotypes or as the total genotypes (Table 3). 10.1371/journal.pone.0161345.t003Table 3 Vaccination status of patients diagnosed with gastroenteritis. Variable Vaccinated Not Vaccinated P-value (n = 255) (n = 1,157) Center <0.001*  AUBMC 61 (38.9%) 96 (61.1%) <0.001*£  Hammoud Hospital 57 (27.5%) 150 (72.5%)  Hôtel-Dieu de France University  Hospital 3 (60.0%) 2 (40.0%)  Makassed Hospital 18 (9.5%) 171 (90.5%)  Nini Hospital 59 (10.6%) 497 (89.4%)  Nabatieh Hospital 52 (24.5%) 160 (75.5%)  Rafic Hariri University Hospital 5 (5.8%) 81 (94.2%) Age (months) 0.008*  0–2 4 (4.3%) 89 (95.7%)  3–6 52 (19.4%) 216 (80.6%)  7–12 84 (18.3%) 375 (81.7%)  13–24 75 (20.6%) 289 (79.4%)  25–60 40 (17.5%) 188 (82.5%) Vesikari score 0.866  ≤10 22 (18.6%) 96 (81.4%)  >10 233 (18.0%) 1,060 (82.0%) Duration of hospitalization (days)‡ 0.001*  0–3 116 (45.5%) 420 (36.3%)  4–7 135 (52.9%) 669 (57.8%)  >7 4 (1.6%) 68 (5.9%) Year 0.308  2011 116 (17.0%) 567 (83.0%)  2012 104 (20.1%) 413 (79.9%)  2013 35 (16.5%) 177 (83.5%) *Significant at the 5% level. £P-value excluding Hôtel-Dieu de France University Hospital from the comparison. ‡ Percentages are computed within each column. Abbreviations: AUBMC = American University of Beirut Medical Center. Forty eight cases of rotavirus infection occurred in vaccinated patients, leading to a vaccine breakthrough rate of 18.8%. Importantly, 40 out of 219 patients (18.3%) vaccinated with RV1 and 7 out of 27 patients (26%) vaccinated with RV5 had not completed the recommended vaccination schedule with the relevant vaccine. We then analyzed the RV genotype distribution of RV1 and RV5 breakthrough cases (Table 4). 10.1371/journal.pone.0161345.t004Table 4 Vaccine breakthrough by vaccine type. Predominant rotavirus genotype Vaccine breakthrough n(%) Vaccine breakthrough per number of doses n(%) RV1 (Rotarix®) RV5 (Rotateq®) RV1 (Rotarix®) RV5 (Rotateq®) 1 dose 2 doses 1 dose 2 doses 3 doses G1P[8] 10 (25%) 1 (14.3%) 7 (17.5%) 3 (7.5%) 0 (0%) 0 (0%) 1 (14.3%) G2P[4] 13 (32.5%) 1 (14.3%) 4 (10%) 9 (22.5%) 0 (0%) 0 (0%) 1 (14.3%) G2P[8] 1 (2.5%) - 0 (0%) 2 (2.5%) - - - G4P[8] 6 (15%) 3 (42.9%) 0 (0%) 6 (15%) 2 (28.6%) 0 (0%) 1 (14.3%) G9P[8] 10 (25%) 1 (14.3%) 2 (5%) 8 (20%) 0 (0%) 0 (0%) 1 (14.3%) G9P[4] - 1 (14.3%) - - 0 (0%) 1 (14.3%) 0 (0%) The majority of the genotypes where completely or partially homotypic with the vaccine types with the heterotypic exceptions being G2P[4] in RV1 (G1P[8]) vaccinated patient and G9P[4] in RV5 (G1,G2, G3, G4, and P[8]) vaccinated patient. Vaccine effectiveness was computed for all vaccine types combined. Using RV-negative controls, the unadjusted VE was 62.7% (95%CI, 41.2%-76.3%). After adjustment by geographical area and season, VE against RV disease was 68.4% (95%CI, 49.6%-80.2%). Vaccine effectiveness was highest among the G1P[8] genotype; but for G2P[4] and G4P[8], VE was much lower with a wide confidence interval due to the small numbers (Table 5). 10.1371/journal.pone.0161345.t005Table 5 Vaccine effectiveness against rotavirus disease. Vaccine Effectiveness (confidence interval at 95%) Unadjusted 0.627 (0.412, 0.763) Adjusted for geographical area 0.648 (0.443, 0.778) Adjusted for geographical area and season 0.684 (0.496, 0.802) By geographical area  Beirut 0.647 (0.136, 0.855)  North 0.747 (0.273, 0.912)  South 0.598 (0.245, 0.786) By RV genotype  G1P[8] 0.921 (0.677, 0.981)  G9P[8] 0.587 (0.033, 0.824)  G2P[4] 0.209 (0.000, 0.617)  G4P[8] 0.415 (0.000, 0.753) Seasons: Season 1: January 2011 to August 2011, Season 2: September 2011 to August 2012, and Season 3: September 2012 to June 2013. Geographical areas of study centers: Beirut: American University of Beirut Medical Center, Makassed Hospital and Rafic Hariri University Hospital; North: Nini Hospital; South: Hammoud Hospital and Nabatieh Hospital. Discussion To our knowledge, our study is the second epidemiological study, and the first extending over three rotavirus seasons, estimating the disease burden and the prevalent genotypes of rotavirus causing gastroenteritis in children based on a sentinel hospital-based surveillance program in Lebanon. Further, it evaluated the difference in clinical outcomes between the vaccinated and unvaccinated cohorts with GE, and provided data about the effectiveness of the two oral vaccines, RV1 and RV5 against RV disease in Lebanon. Socio-demographic and clinical characteristics of patients diagnosed with GE Our analysis demonstrates that the burden of RV GE is significant in Lebanon with the rate of hospitalized GE caused by RV being 30.3% (n = 428, 95%CI, 27.9%-32.7%) between 2011 and 2013. This is lower than the findings reported in studies carried out in other developing countries in the Middle East. The percentage of rotavirus infection in hospitalized infants and young children was 37% in Iraqi Kurdistan [25], 39% in both Turkey [26,27] and Jordan [27], 40% in Kuwait [28], 45.2% in Yemen [29] and 49% in Oman [30]. The highest rate of rotavirus infection was recorded in Syria (61%) [27]. Contrarily, lower percentages of rotavirus infection were reported in our previous hospital-based surveillance study where GE attributable to RV was 27.7% [21]. Lower rates were also documented in Saudi Arabia (16%) [27] and Egypt (25.2%) [31]. Additionally, the rate of RV GE observed in this study was higher than those in developed countries such as the USA where RV was found to be responsible for 22% of childhood diarrheal hospitalizations [32]. These variations may reflect actual differences in RVGE rates but may also be related to variations in study design, the methods used in detecting rotavirus, and vaccination against RV. Indeed, high RV vaccination coverage rates were reported in the developed countries where mass vaccination programs have been implemented (72% in USA, 84%-87% in Australia, 87% in Austria and 85% in Belgium) leading to a significant drop in rates of RV-related hospitalizations [33]. In Lebanon, RV vaccination is currently not included in the national immunization program but is available in the private sector. The seven sentinel surveillance hospitals in our study serve families from a range of socioeconomic backgrounds, including those with very low to high income and this is reflected in the vaccination rates observed that ranged from 5.8% to 60%. Also, the present study found that, despite the use of vaccines, the rotavirus prevalence has not changed, but rather increased by a few points. This is likely due to suboptimal herd immunity as only 18.1% of subjects had received rotavirus vaccine. As for age, 83.6% of RV GE was in children aged below two years. Our findings corroborate the results of our previous study conducted in Lebanon, in the same settings as the present study, where 75% of the RV GE cases occurred in children under two years of age [21]. Further, the difference between male and female in this study was not significant (P = 0.734), this result is concordant with previous studies [29,34]. Most (61.0%) of the 428 RV-positive patients required hospitalization for 4 to 7 days and 93% of them had high severity (Vesikari score >10). When compared with RV-negative patients, no statistically significant difference was found for the duration of hospitalization or the severity score. Many previous studies illustrated the burden of hospitalization due to RV GE. Hospitalization attributed to RV rose from 13.5% in the period 1999–2000 to 17.1% in the study period (2001–2005) in Spain in children aged less than 5 years old [35]. RV also accounted for 40% of diarrheal in-patient consultations in Argentina [36] and, in USA, RV led to 55,000 to 70,000 hospitalizations in children younger than 5 years of age in the post-vaccine era [37]. The significant difference between the RV-positive and RV-negative patients concerning fever and vomiting are not conclusive in the clinical diagnosis of RV GE, because such symptoms are not distinguishable from other types of GE. Nonetheless, the frequent combination of vomiting (62.9%) with diarrhea (91.1%) leads to the risk of dehydration in RV GE more than other types of GE and therefore increases the need for hospitalization and rehydration. A concerning finding in our study is that 40.4% of the patients with RV GE still received antibiotics, an unnecessary measure in a viral infection. Over the 30-month-long surveillance period, the 1414 cases of GE occurred during all seasons but the majority was observed from December till March (41.1%), followed by July (12.3%). Also, RV GE was much more predominant in the cold season (69.9%). This is similar to the pattern seen in the literature including our previous study, which showed the same seasonal occurrence [21,27,35,38,39]. RV genotyping The most prevalent genotypes were G1P[8] (36%) followed by G9P[8] (26.4%), G2P[4] (17.8%) and G4P[8] (15.9%). Our most prevalent genotypes were recorded in a Yemeni study from 2014 where two of the major global human RV genotypes (G2P[4] 55% and G1P[8] 15%) were detected [29]. Our findings are also in agreement with a study from Iraq where G1P[8] (33%),G4P[8] (21%), G2P[4] (15%), G9P[8] (11%) and G1P[6] (11%) were the most frequently detected RV genotypes previously reported from Erbil in 2005 [24]. In Oman, the most frequent rotavirus genotypes in children aged less than 5 years and hospitalized for gastroenteritis, were G2P[4] (26%), G1P[8] (3%) and G1P[4] (3%) and G9P[4] (3%) [29]. Importantly, RV genotypes distribution in our study showed a different predominance profile across the three seasons, with G9P[8] being predominant in seasons 1 and 3, and G1P[8] in season 2 (P<0.001) (Fig 1). This may suggest that the same genotype is unlikely to predominate for two successive seasons due to the exhaustion of the susceptible infant population. A Brazilian study showed that G9P[8] was most prevalent in 2011 and G12P[8] in 2012 [40]. It was interesting to note that, despite the small area of Lebanon, there was significant geographic variation in the dominant genotypes. Whereas the four major genotypes circulated during each of the three seasons when all of Lebanon was considered, the relative predominance of specific genotypes varied between different regions. It is noteworthy that different genotypes prevailed in different countries, but the discrepancies with our findings could be due to differences in the year(s) covered in each study. Vaccination status of patients diagnosed with GE and vaccine effectiveness The vaccination and RV rates were significantly correlated (Pearson’s chi-square coefficient of 19.1, P<0.001) as RV infections decreased with vaccinations. Also, the length of hospitalization for 4–7 days was significantly lower in the vaccinated patients (52.9%) compared to the unvaccinated ones (57.8%) (P<0.001) possibly indicating that vaccinated patients are able to recover faster. The vaccine breakthrough rate of 18.8% and the vaccine effectiveness of 68.4% observed in our study is commensurate with those of previous post-marketing studies assessing the effectiveness of the pentavalent (Rotateq®) and monovalent (Rotarix®) rotavirus vaccines in real-world settings [41–44]. Our study was not designed to measure differences in effectiveness between the vaccines. The majority of patients in our study were vaccinated with Rotarix®. Our VE results against RV disease are reassuring for a middle-income country like Lebanon. They also highlight the effectiveness of vaccination against RV and the importance of mass immunization to decrease its burden. Vaccination rate is still very low despite vaccine effectiveness, consequently the frequencies of RV cases did not change between the two studies [21]. Study limitations and strengths One limitation of our study is the inability to measure the true burden of rotavirus disease. The number of patients cared for by the different hospitals will be inaccurate due to the multiplicity of hospitals serving the same population. Nonetheless, this study is the first Lebanese and regional one to look at three consecutive seasons of rotavirus. This has implications regarding season-to-season variation in the predominant genotype and the relevance of vaccine used. Importantly, our study was also designed to evaluate both the burden of RV disease in hospitalized children and vaccine effectiveness of the 2-dose RV1 (Rotarix®) and 3-dose pentavalent vaccine (RV5, Rotateq®) series against rotavirus disease. Conclusion In conclusion this study provides updated information on the extent of rotavirus-specific hospitalizations of RV GE, the prevalent genotypes of RV, vaccine effectiveness, and the impact of vaccination against the pathogen in Lebanon. Our results indicate that gastroenteritis caused by rotavirus in Lebanon is an important health problem, particularly among children aged under two years, mostly during the winter season. Also, our findings provide compelling evidence of the potential benefits of RV vaccination among infants and young children with acceptable VE. However, the data provided show a low rate of vaccination (18.1%), despite the availability of the two oral vaccines RV1 and RV5 in Lebanon but only in the private sector. Therefore, government supported vaccinations are of utmost importance to ensure mass immunization. The present results would help inform the health policymakers about the RV GE burden and underscore the need for national immunization programs to be implemented all over the country. Supporting Information S1 CRF Case Report Form (CRF)—Rotavirus study. (PDF) Click here for additional data file. We are grateful to all the children, their parents and guardians who participated in the study and Merck Sharp & Dohme (MSD) for funding. We thank Caline Balaa and Omar Kebbe Baghdadi for their help with the study monitoring. We also thank Racha Aaraj from Phoenix Clinical Research for help with writing the manuscript and Ziyad Mahfoud for help with the statistics. ==== Refs References 1 Cortese MM , Parashar UD . Centers for Disease Control and Prevention (CDC) . Prevention of rotavirus gastroenteritis among infants and children: recommendations of the Advisory Committee on Immunization Practices (ACIP) . MMWR Recomm Rep . 2009 ; 58 (RR-2 ): 1 –25 . 19194371 2 Prasad BV , Crawford S , Lawton JA , Pesavento J , Hardy M , Estes MK . Structural studies on gastroenteritis viruses . Novartis Found Symp . 2001 ; 238 : 26 –37 ; discussion 37–46. 11444031 3 Ray PG , Kelkar SD , Walimbe AM , Biniwale V , Mehendale S . Rotavirus immunoglobulin levels among Indian mothers of two socio-economic groups and occurrence of rotavirus infections among their infants up to six months . J Med Virol . 2007 ; 79 (3 ): 341 –349 . 17245723 4 Adams WR , Kraft LM . Epizootic diarrhea of infant mice: identification of the etiologic agent . Science . 1963 ; 141 (3578 ): 359 –360 . 14011026 5 Bishop RF , Davidson GP , Holmes IH , Ruck BJ . Virus particles in epithelial cells of duodenal mucosa from children with acute non-bacterial gastroenteritis . Lancet . 1973 ; 2 (7841 ): 1281 –1283 . 4127639 6 Bern C , Martines J , de Zoysa I , Glass RI . The magnitude of the global problem of diarrhoeal disease: a ten-year update . Bull World Health Organ . 1992 ; 70 (6 ): 705 –714 . 1486666 7 Parashar DU , Gibson CJ , Bresee JS , Glass RI . Rotavirus and severe childhood diarrhea. Centers for disease control and prevention (CDC) Atlanta, Georgia, USA . Emerging Infectious Diseases 2006 ; 12 (2 ): 304 –6 .16494759 8 Centers for Disease Control and Prevention (CDC) . Rotavirus surveillance—worldwide, 2001–2008 . MMWR . 2008 ; 57 (46 ): 1255 –1257 . 19023263 9 Pediatric ROTavirus European CommitTee (PROTECT) . The paediatric burden of rotavirus disease in Europe . Epidemiol Infect . 2006 ; 134 (5 ): 908 –916 . Epub 2006 Apr 4. Review. 16650331 10 Soriano-Gabarró M , Mrukowicz J , Vesikari T , Verstraeten T . Burden of rotavirus disease in European Union countries . Pediatr Infect Dis J . 2006 ; 25 (1 Suppl ): S7 –S11 . 16397431 11 Payne DC , Staat MA , Edwards KM , Szilagyi PG , Gentsch JR , Stockman LJ , et al Active, population-based surveillance for severe rotavirus gastroenteritis in children in the United States . Pediatrics . 2008 ; 122 (6 ): 1235 –1243 . 10.1542/peds.2007-3378 19047240 12 Ramig RF . Pathogenesis of intestinal and systemic rotavirus infection . J Virol . 2004 ; 78 (19 ): 10213 –10220 . 15367586 13 Chandran A , Heinzen RR , Santosham M , Siberry GK . Nosocomial rotavirus infections: a systematic review . J Pediatr . 2006 ; 149 (4 ): 441 –447 . 17011311 14 Glass RI , Bresee J , Jiang B , Gentsch J , Ando T , Fankhauser R , et al Gastroenteritis viruses: an overview . Novartis Found Symp . 2001 ; 238 : 5 –19 ; discussion 19–25. Review. 11444035 15 Ansari SA , Springthorpe VS , Sattar SA . Survival and vehicular spread of human rotaviruses: possible relation to seasonality of outbreaks . Rev Infect Dis . 1991 ; 13 (3 ): 448 –461 . 1866549 16 Crawford SE , Patel DG , Cheng E , Berkova Z , Hyser JM , Ciarlet M , et al Rotavirus viremia and extraintestinal viral infection in the neonatal rat model . J Virol . 2006 ; 80 (10 ): 4820 –4832 . 16641274 17 Vesikari T , Isolauri E , Delem A , D'Hondt E , André FE , Zissis G . Immunogenicity and safety of live oral attenuated bovine rotavirus vaccine strain RIT 4237 in adults and young children . Lancet . 1983 ; 2 (8354 ): 807 –811 . 6137646 18 Centers for Disease Control and Prevention. Rotavirus Vaccine (RotaShield®) and Intussusception. Historical information as RotaShield® was taken off U.S. market in 1999. Available: http://www.cdc.gov/vaccines/vpd-vac/rotavirus/vac-rotashield-historical.htm. 10403588 19 Greenberg HB , Estes MK . Rotaviruses: from pathogenesis to vaccination . Gastroenterology . 2009 ; 136 (6 ):1939 –51 . 10.1053/j.gastro.2009.02.076 19457420 20 Ruiz-Palacios GM , Pérez-Schael I , Velázquez FR , Abate H , Breuer T , Clemens SC , et al Human Rotavirus Vaccine Study Group . Safety and efficacy of an attenuated vaccine against severe rotavirus gastroenteritis . N Engl J Med . 2006 ;354 (1 ):11 –22 . 16394298 21 Dbaibo G , Rajab M , Inati A , Mikhael R , Choueiry E , Al-Tannir M , et al Hospital-based surveillance study of rotavirus gastroenteritis in children under 5 years of age in Lebanon . Trials in Vaccinology 2013 ; 2 : 25 –30 . 22 Lewis K. Vesikari Clinical Severity Scoring System Manual. 2 May 2011. Available: https://www.path.org/publications/files/VAD_vesikari_scoring_manual.pdf. Accessed: 19 April 2016. 23 Nirwati H , Wibawa T , Aman AT , Soenarto Y . Genotyping of Rotavirus by Using RT-PCR Methods . Indonesian Journal of Biotechnology 2013 ; 18 (1 ): 8 –13 . 24 Mundi Index. Lebanon Demographics Profile. Available: http://www.indexmundi.com/lebanon/demographics_profile.html Accessed 13 July 2016. 25 Ahmed HM , Coulter JB , Nakagomi O , Hart CA , Zaki JM , Al-Rabaty AA , et al Molecular characterization of rotavirus gastroenteritis strains, Iraqi Kurdistan . Emerg Infect Dis . 2006 ; 12 (5 ): 824 –826 . 16704845 26 Mitui MT , Bozdayi G , Ahmed S , Matsumoto T , Nishizono A , Ahmed K . Detection and molecular characterization of diarrhea causing viruses in single and mixed infections in children: a comparative study between Bangladesh and Turkey . Med Virol . 2014 ; 86 (7 ): 1159 –1168 . 27 Khoury H , Ogilvie I , El Khoury AC , Duan Y , Goetghebeur MM . Burden of rotavirus gastroenteritis in the Middle Eastern and North African pediatric population . BMC Infectious Diseases 2011 , 11 :9 10.1186/1471-2334-11-9 21214934 28 Marmash RW , Dalwai AK , Szucs G , Molla AM , Pacsa AS , Al Nakib W , et al Genotypic characterization of rotaviruses and prevalence of serotype-specific serum antibodies in children in Kuwait . Epidemiol Infect 2007 ; 135 (8 ): 1331 –1337 . 17291379 29 Al-Badani A , Al-Areqi L , Majily A , Al-Sallami S , Al-Madhagi A , Al-Kamarany MA . Rotavirus Diarrhea among Children in Taiz, Yemen: Prevalence—Risk Factors and Detection of Genotypes . Int J Pediatr . 2014 ; 2014 : 928529 10.1155/2014/928529 25197286 30 Al Awaidy SA , Bawikar S , Al Busaidy S , Baqiani S , Al Abedani I , Varghese R , et al Considerations for introduction of a rotavirus vaccine in Oman: rotavirus disease and economic burden . J Infect Dis . 2009 ; 200 Suppl 1 :S248 –53 . 10.1086/605339 19817605 31 Matson DO , Abdel-Messih IA , Schlett CD , Bok K , Wienkopff T , Wierzba TF , et al Rotavirus genotypes among hospitalized children in Egypt, 2000–2002 . J Infect Dis .2010 ; 202 Suppl :S263 –5 . 10.1086/653581 20684714 32 Parashar DU , Gibson CJ , Bresee JS , Glass RI . Rotavirus and severe childhood diarrhea. Centers for disease control and prevention (CDC) Atlanta, Georgia, USA . Emerging Infectious Diseases 2006 ; 12 (2 ): 304 –6 .16494759 33 Uhlig U , Kostev K , Schuster V , Uhlig HH . Rotavirus vaccination in Germany: analysis of nationwide surveillance data 2006 to 2010 . Pediatr Infect Dis J . 2011 ; 30 (12 ): e244 –7 . 21817956 34 Nguyen TV , Le Van P , Le Huy C , Weintraub A . Diarrhea caused by rotavirus in children less than 5 years of age in Hanoi, Vietnam . J Clin Microbiol . 2004 ; 42 (12 ): 5745 –5750 . 15583308 35 López-de-Andrés A , Jiménez-García R , Carrasco-Garrido P , Alvaro-Meca A , Galarza PG , de Miguel AG . Hospitalizations associated with rotavirus gastroenteritis in Spain, 2001–2005 . BMC Public Health . 2008 ; 8 : 109 10.1186/1471-2458-8-109 18397512 36 Bok K , Castagnaro N , Borsa A , Nates S , Espul C , Fay O , et al Surveillance for rotavirus in Argentina . J. Med. Virol . 2001 ; 65 : 190 –198 . 11505463 37 Centers for disease control and prevention. Rotavirus in the US. Available: http://www.cdc.gov/rotavirus/surveillance.html. Accessed: 19 April 2016. 38 Bicer S , Sahin GT , Koncay B , Gemici H , Siraneci R , Ozturk NY , et al Incidence assessment of rotavirus and adenovirus associated acute gastroenteritis cases in early childhood . Infez Med . 2011 ; 19 (2 ): 113 –119 . 21753251 39 Hassine-Zaafrane M , Sdiri-Loulizi K , Ben Salem I , Kaplon J , Ayouni S , Ambert-Balay K , et al The molecular epidemiology of circulating rotaviruses: three-year surveillance in the region of Monastir, Tunisia . BMC Infect Dis . 2011 ; 11 : 266 10.1186/1471-2334-11-266 21967503 40 Luchs A , Cilli A , Morillo SG , Carmona RCC , Timenetsky MCST . Rotavirus genotypes circulating in brazil, 2007–2012: implications for the vaccine program . Rev Inst Med Trop Sao Paulo . 2015 ; 57 (4 ): 305 –313 . 10.1590/S0036-46652015000400006 26422154 41 Adlhoch C , Hoehne M , Littmann M , Marques AM , Lerche A , Dehnert M , et al Rotavirus vaccine effectiveness and case-control study on risk factors for breakthrough infections in Germany, 2010–2011 . Pediatr Infect Dis J . 2013 ; 32 (2 ): e82 –9 . 10.1097/INF.0b013e3182720b71 23334342 42 Chang WC , Yen C , Wu FT , Huang YC , Lin JS , Huang FC , et al Effectiveness of 2 rotavirus vaccines against rotavirus disease in Taiwanese infants . Pediatr Infect Dis J . 2014 ; 33 (3 ): e81 –6 . 10.1097/INF.0000000000000105 24569388 43 Payne DC , Boom JA , Staat MA , Edwards KM , Szilagyi PG , Klein EJ , et al Effectiveness of pentavalent and monovalent rotavirus vaccines in concurrent use among US children <5 years of age, 2009–2011 . Clin Infect Dis . 2013 ; 57 (1 ): 13 –20 . 10.1093/cid/cit164 23487388 44 Cortese MM , Immergluck LC , Held M , Jain S , Chan T , Grizas AP , et al Effectiveness of monovalent and pentavalent rotavirus vaccine . Pediatrics . 2013 ; 132 (1 ):e25 –33 . Epub 2013 Jun 17. 10.1542/peds.2012-3804 23776114

PMC005xxxxxx/PMC5003351.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757156510.1371/journal.pone.0161915PONE-D-16-14916Research ArticleResearch and analysis methodsMathematical and statistical techniquesStatistical methodsMonte Carlo methodPhysical sciencesMathematicsStatistics (mathematics)Statistical methodsMonte Carlo methodBiology and Life SciencesCell BiologyCell ProcessesCell ProliferationBiology and Life SciencesCell BiologyCell ProcessesCell Cycle and Cell DivisionResearch and Analysis MethodsBiological CulturesCell CulturesBiology and Life SciencesCell BiologyCell ProcessesCell GrowthResearch and Analysis MethodsSpecimen Preparation and TreatmentStainingCell StainingBiology and Life SciencesBiophysicsBiophysical SimulationsPhysical SciencesPhysicsBiophysicsBiophysical SimulationsBiology and Life SciencesComputational BiologyBiophysical SimulationsPhysical SciencesChemistryChemical ElementsOxygenSpheroid Formation of Hepatocarcinoma Cells in Microwells: Experiments and Monte Carlo Simulations Cellular Spheroid Formation and Monte Carlo SimulationsWang Yan 12Kim Myung Hee 12Tabaei Seyed R. 12Park Jae Hyeok 12Na Kyuhwan 3Chung Seok 3Zhdanov Vladimir P. 4http://orcid.org/0000-0002-8692-8955Cho Nam-Joon 125*1 School of Materials Science and Engineering, Nanyang Technological University, Singapore2 Centre for Biomimetic Sensor Science, Nanyang Technological University, Singapore3 School of Mechanical Engineering, Korea University, Seoul, Korea4 Boreskov Institute of Catalysis, Russian Academy of Sciences, Novosibirsk, Russia5 School of Chemical and Biomedical Engineering, Nanyang Technological University, SingaporeMassoumi Ramin EditorLund University, SWEDENCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: NJC. Formal analysis: YW MHK SRT JHP KN SC VPZ NJC. Funding acquisition: NJC. Investigation: YW MHK JHP KN SC VPZ NJC. Methodology: MHK JHP KN SC VPZ NJC. Resources: MHK JHP. Supervision: NJC. Validation: YW MHK SRT JHP KN SC VPZ NJC. Writing – original draft: YW MHK JHP. Writing – review & editing: YW MHK JHP KN SC VPZ NJC. * E-mail: njcho@ntu.edu.sg29 8 2016 2016 11 8 e016191512 4 2016 15 8 2016 © 2016 Wang et al2016Wang et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.The formation of spherical aggregates during the growth of cell population has long been observed under various conditions. We observed the formation of such aggregates during proliferation of Huh-7.5 cells, a human hepatocarcinoma cell line, in a microfabricated low-adhesion microwell system (SpheroFilm; formed of mass-producible silicone elastomer) on the length scales up to 500 μm. The cell proliferation was also tracked with immunofluorescence staining of F-actin and cell proliferation marker Ki-67. Meanwhile, our complementary 3D Monte Carlo simulations, taking cell diffusion and division, cell-cell and cell-scaffold adhesion, and gravity into account, illustrate the role of these factors in the formation of spheroids. Taken together, our experimental and simulation results provide an integrative view of the process of spheroid formation for Huh-7.5 cells. http://dx.doi.org/10.13039/501100001381National Research Foundation SingaporeNRF-NRFF2011-01http://orcid.org/0000-0002-8692-8955Cho Nam-Joon http://dx.doi.org/10.13039/501100001381National Research Foundation SingaporeNRF-CRP10-2012-07http://orcid.org/0000-0002-8692-8955Cho Nam-Joon This work was funded by National Research Foundation Singapore (www.nrf.gov.sg), NRF-NRFF2011-01 and NRF-CRP10-2012-07, awarded to NJC. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper.Data Availability All relevant data are within the paper. ==== Body Introduction The study of cell culture in three-dimensional (3D) scaffolds is of considerable intrinsic interest and is also important in the context of numerous applications including, e.g., tissue engineering, disease modeling and drug screening platforms [1–3]. The structure and size of the corresponding scaffolds vary in a broad range from two-dimensional (2D) arrays of sub-millimeter wells to complex 3D structures aiming at mimicking specific organs [2, 3]. Chemically, the scaffolds are often fabricated by using natural hydrogels [2], synthetic polymers [1], or combination of such materials [4]. Cells growing in scaffolds typically aggregate. The shape and morphology of aggregates may be different, depending on various factors including the cell type, design of a scaffold and the corresponding fabrication material [1]. Cellular spheroids represent the most common shape of cell assembly [5, 6]. Aggregates of this shape were created, e.g., by concave microwell method [7], hanging drop method [5, 8], or rotating-wall vessel technique [9, 10]. The size (diameter) of spheroids may reach ~1 cm as observed in experiments with human colon adenocarcinoma cells [9] and rat hepatocytes [11] (the latter cells displayed liver-like morphology or, more specifically, a compact structure with tight cell-cell junctions, smooth and rough endoplasmic reticulum and bile canaliculi lined with the microvilli). Often, the size is smaller. For example, the size of spheroids composed of mammary epithelial cells was reported to be ~100 μm (these spheroids can produce and secrete milk proteins upon hormonal stimulation) [5], while in the case of hepatocytes the size was ~200 μm [7]. The growth of cell cultures in scaffolds is of interest also in the context of theoretical biology and statistical physics (for general introduction into this area, see reviews [12–16]). The corresponding models are usually based on the mean-field (MF) kinetic equations or Monte Carlo (MC) simulations. The MF approach is convenient in the situations where the geometry is simple. Such models were used to scrutinize the limitations in the nutrient supply and oxygen transport in porous scaffolds on the coarse-grained level without or with explicit description of single pores (see e.g. references [4, 17, 18] and [18, 19], respectively, and references therein). MC simulations, based often on the lattice approximation and describing evolution of an ensemble of individual cells, are efficient in the situations with complex geometry and/or in the cases when the focus is on aggregation of cells (as in our present study). The available generic 2D and 3D MC simulations have been focused on the growth and differentiation of stem cells [20], cell seeding [21], and formation of cell sheets [4]. Related theoretical studies concern stem-cell niches [22–25] and scaffold-less biofabrication [26]. Herein, we report the results of our study of culturing Huh-7.5 cells in microfabricated low-adhesion microwells. These cells belonging to a human hepatocarcinoma cell line are widely used as a liver cell model for the exploration of HCV infection [27]. Earlier, we observed the formation of Huh-7.5 cell spheroids in PEG-based hydrogels [28] and multilayer cell sheets in a biofunctionalized 3D scaffold [4, 29]. Our present work is focused on the same cells and has three novel ingredients. First, we use a recently designed microwell platform for direct observation of the proliferation of cells. Its advantages include: (i) The microwell has a total depth that is two times of its diameter, and walls formed of triangular flat fragments are used to separate adjacent wells. So in contrast to conventional microfabricated semi-circular wells, this mechanical stress (shear force)-free design prevents the cells from slipping during medium exchange, and the method of fluid delivery is diffusion based. (ii) Compared to the hanging drop method [5, 8], the microwell system enables flexible medium exchange during incubation, and due to the small radius of curvature of individual wells (smaller than the size of a water drop), we are able to achieve fairly similar distribution of cell aggregates in different wells. (iii) Another feature distinguishing it from the conventional plastic round bottom wells is that the base (fabricated from silicone elastomer) is easily oxygen-permeable. The latter allows us to reduce hypoxia of cells in the centers of aggregates that is inevitable in conventional plastic scaffolds. Second, the use of microwell platform described above allowed us to observe explicitly in detail the formation of spherical Huh-7.5 aggregates on the length scale up to 500 μm, with both light microscopy and confocal fluorescence microscopy. Third, our experimental results are complemented by 3D Monte Carlo simulations to illustrate the role of various factors (e.g., cell diffusion, cell-surface adhesion, and gravity) in the process of cell aggregation and spheroid formation. Materials and Methods Pretreatment of 3D SpheroFilm™ for cell culture The 3D SpheroFilm™ microwell was obtained from Incyto Co. (Chonan, Korea). The inner diameter of the hemispheres in the microwell is 500 μm, and the total well depth is 1000 μm (Fig 1). The microwell was made of mass-producible silicone elastomer. For proper cell culture, the microwell was cut to 1-cm2 squares, and was placed at the bottom of wells in a 24 well plate. 100% ethanol was added into the plate and repeatedly pipetted to remove the air bubbles from the wells. Once ethanol was removed, the wells were washed three times with phosphate-buffered saline (PBS), and then incubated with cell culture medium for at least 24 hours. The cell culture medium was removed prior to cell seeding. 10.1371/journal.pone.0161915.g001Fig 1 Scheme of SpheroFilmTM scaffold and spheroid formation there. (a) Top view of SpheroFilmTM containing 361 (19 × 19) microwells; [(b) and (c)] shape and dimensions of microwells with a diameter of 500 μm and the total well depth of 1000 μm; (d) cell seeding and attachment in and outside of microwells; (e) side and top views of spheroid formation in microwells; (f) side and top views of cell growth in and outside of a microwell. Cell culture and cell seeding in 3D SpheroFilm Human hepatocarcinoma Huh-7.5 cells, purchased from Apath (NY, USA), were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin and 100 μg/ml streptomycin (Life Technologies) in a humidified atmosphere with 5% CO2 at 37°C. Medium was changed every three days. Cells were detached with 0.25% trypsin-EDTA solution (Life Technologies) from the tissue culture plate, counted, and adjusted to 0.05, 0.1 and 0.2 × 106 cell/ml. 1 ml cell suspension was seeded in each piece of microwell square, cut and placed in 24 well plate. After 10 minutes of cell seeding, suspending cells were removed by aspiration, and the remaining cells were washed and incubated in fresh growth media until time of assay. Immunofluorescence staining and imaging Cells in the microwells were collected at various stages for immunocytochemistry. Cells were washed twice with PBS, fixed with 4% paraformaldehyde (PFA) for 10 minutes, permeabilized with 0.1% Triton X-100 in PBS for 30 minutes, washed again with PBS and incubated in blocking buffer (3% bovine serum albumin (BSA) in PBS) for 1 hour. Cells were stained with mouse primary antibody against Ki-67 (Life Technologies) by overnight incubation at 4°C, and then washed three times with PBS to remove unbound primary antibody. The cells were then incubated with anti-mouse secondary antibody conjugated with Alexa Fluor® 488 (Life Technologies). Meanwhile, filamentous actin (F-actin) was stained with Alexa Fluor® 555 labelled phalloidin (Life Technologies) for 2 hours at room temperature (protected from light). After two washes with PBS, the nuclei were stained with 10 μg/ml DAPI (Life Technologies) for 30 minutes. Fluorescent cell images were taken on a LSM 710 confocal microscope with ZEN program (Carl Zeiss). Results and Discussion Cell growth and spheroid formation The shape and dimensions of 3D SpheroFilmTM microwells are schematically illustrated in Fig 1. Huh-7.5 cells were seeded in the microwells at density of 0.05 × 106 cell/ml. These cells underwent several stages of morphological changes in the following 10 days (Fig 2). Initially they were attached to the bottom of the hemispheres and randomly dispersed. Later on (Fig 2, day 1), multiple cells started to form clusters due to the seemingly random cell migration and cell-cell adhesion. After day 1, several small cell clusters as well as a number of individual cells gradually merged into large cell aggregates at the center of each well of the microwell surface (Fig 2, day 4). These pre-mature spheroids kept growing bigger and denser, and finally became mature with significant thickness and clear 3D structure at day 7 and onward (Fig 2, day 7 and day 10). When Huh-7.5 cells were seeded at higher densities (0.1 and 0.2 × 106 cell/ml, respectively), similar stages of morphological changes can be observed with earlier cell aggregation and denser spheroids formation (Fig 2). 10.1371/journal.pone.0161915.g002Fig 2 Cell growth and spheroid formation in the microwells. Three different densities of Huh-7.5 cells (0.05, 0.1 and 0.2 × 106 cell/ml) were seeded into the platform and the cells were observed under light microscope over 10-day period. Scale bar is 100 μm. The cell growth patterns were further examined by fluorescent staining of Ki-67 and F-actin. Ki-67 is used as a marker of proliferating cells [30–32]. The expression of Ki-67 is known to be upregulated in the G1, S, G2 and M phases of cell cycle, and reaches maximum in G2 and M phases [33, 34]. In spheroid cell culture (Fig 3), Huh-7.5 cells showed intense Ki-67 staining (green fluorescence) at day 1 after cell seeding, but the green fluorescence was markedly reduced at day 4 compared to day 1. The expression of Ki-67 at day 7 and 10 was similar to day 4. This observation indicates that the rate of cell proliferation was fastest at day 1, and it slowed down at day 4 and the rate was somewhat held constant till day 10. Another interesting observation is the spatial distribution of Ki-67 expression across the spheroid. At day 1, Ki-67 staining is evenly distributed across the cell aggregates; however, at days 4, 7 and 10, strong Ki-67 cells were found scattered on the periphery of spheroids (see white arrows), meaning that cell proliferation was globally suppressed after day 4, but some cells on the periphery still maintained their proliferation activity. Interestingly, although cells in the spheroids showed suppressed growth rate, the cells growing outside the microwells (those attached to the flat surface) demonstrated consistent high expression of Ki-67 (Fig 4). 10.1371/journal.pone.0161915.g003Fig 3 Ki-67 and F-actin staining of Huh-7.5 cells in the microwells. Huh-7.5 cells were seeded at 0.05 × 106 cell/ml density, and after 1, 4, 7 or 10 days of incubation, cells were stained with DAPI for nucleus (blue), Alexa Fluor 488 for Ki-67 (green) and Alexa Fluor 555 for F-actin (red). DIC stands for differential interference contrast images. The bright green and red circles observed at days 4, 7 and 10 are cells that strongly express Ki-67 and F-actin (see white arrows). Scale bar is 100 μm. 10.1371/journal.pone.0161915.g004Fig 4 Ki-67 and F-actin staining of Huh-7.5 cells on flat surface outside the microwells. Scale bar is 100 μm. The expression of F-actin displayed similar trend with Ki-67 (Figs 3 and 4). Concerning this aspect, we note that Chang and Hughes-Fulford earlier observed remarkable difference between actin cytoskeleton of cells in monolayers and spheroids [35]. In particular, F-actin stress fibers were found in cells of monolayer culture, while cells in spheroids featured cortical actin that clearly distributed at the outline of the cells [35]. They also reported the upregulation of structural genes including cytoskeletal molecules in monolayers. Thus, it seems to be reasonable to attribute the strong F-actin staining of cells in the microwells at day 1 (Fig 3) as well as cells outside microwells at all the time points (Fig 4) to the cell-substrate interaction and monolayer morphology, whereas when there is more cell-cell interaction in cell spheroid (Fig 3), the F-actin expression is weaker. Monte Carlo simulations In reality, the proliferation of cells is usually accompanied by their aggregation due to cell-cell adhesion. In microwells, the aggregation may take place near the microwell walls due to adhesion of cells to the wall surface, and it may result in the formation of cell layers attached to the walls. The size and structure of cell aggregates depend on various factors including the adhesion strength, balance between the rates of cell division and diffusion, structure of the microwell, cell-cell communication, and limitations in the nutrient or oxygen supply, etc. Our present 3D lattice MC simulations are focused on the proliferation of cells in a single well of the microwell system (Fig 1). The emphasis is on adhesion, diffusion and the likely role of gravity in this process. The corresponding parameters are varied in a wide range in order to provide a general view on the pattern formation in the system under consideration. Some other general factors are ignored in our simulations (which is inevitable due to the complexity of the system). In particular, the cell-cell communication is not taken into account primarily because at the moment we have no data specifying this factor. Concerning this aspect, we may note that the use of the Ki-67 marker indicates that in our experiments the cell growth occurs mainly at the periphery of aggregates. It may be related to cell-cell communication and also to spatial constraints on the cell division. Our simulations take the latter factor into account, and accordingly the results are expected to be robust even if the former factor is ignored. The limitations of the nutrient or oxygen supply are not taken into account either. Such limitations are expected to be significant roughly on the aggregate length scale larger than 200–400 μm [36]. The aggregates we observe (Fig 3) are smaller than or comparable to this length scale, and accordingly the nutrient or oxygen supply is not expected to terminate the cellular growth. To mimic the hemispherical well, we use a 60 × 60 × 80 slab of a cubic lattice. This slab is cut down to a hemisphere in the lower part (at z < 30, where z is the vertical coordinate measured in the lattice spacing units) and a cylinder in the upper part (at 30 ≤ z ≤ 80). Each lattice site can be either vacant or occupied by one cell. Cells can diffuse and divide. As usual in the lattice models, the diffusion of cells is realized via jumps of monomers to vacant nearest-neighbour (nn) sites. Diffusion (sedimentation) of aggregates is not taken into account partly due to the lack of simple suitable algorithms and partly because in our case (during the cell proliferation) this process is expected to be important only at the initial stage of the whole process (when the aggregates are small and their number is large) while we are more interested in the situation when the aggregates are relatively large. Division of a cell is also considered to be possible provided it has a vacant nn site. After division, the cell remains on the same site and another (newly born) cell occupies a vacant nn site. Both for diffusion and division, we use the so-called no-flux boundary conditions, i.e., a jump or division is possible if an nn vacant site belongs to the array of sites representing the well. To mimic the cell-cell and cell-surface adhesion, we introduce the effective attractive dimensionless (i.e., normalized to kBT) interaction, ϵcc < 0, between two cells located in nn sites and the effective attractive dimensionless cell-surface interaction, ϵcs ≤ 0, for cells contacting the boundary representing the wall of microwells. In this case, the aggregation is known to occur at |ϵcc| > 0.89 [37]. In our calculations, we set ϵcc = −1.2. The value of the other interaction, ϵcs, is varied to illustrate various situations. To specify diffusion of monomers in the lattice models, one can use various prescriptions, e.g., those corresponding to the initial-state, Metropolis, or Kawazaki dynamics [38]. We employ the initial-state dynamics which reasonably predict rapid diffusion of single cell and slow diffusion inside aggregates. In this case, the jump rate is reduced by exp(nϵcc) if a cell has n neighbours. If a cell contacts the boundary, the jump rate is reduced by the additional factor, exp(mϵcs), where m is the number of the corresponding contacts. To characterize the relative rates of division and diffusion, we use the dimensionless parameter pdiv. The rates of division and diffusion are considered to be proportional to pdiv and 1 –pdiv, respectively. With the specification above, our MC algorithm consists of sequential trials to realize diffusion or division events. A site is chosen at random. If the site is vacant, the trial ends. Otherwise, a cell located in the site tries to diffuse if ρ > pdiv or to divide if ρ < pdiv, where ρ (0 ≤ ρ ≤ 1) is a random number. In both cases, one of the nn site is selected at random and if the latter site is vacant, the division is performed with unit probability, while the diffusion jump is realized with the probability exp(nϵcc) or exp(nϵcc + mϵcs) depending on the local arrangement. After each trial, the time is incremented by |ln(ρ)| / N, where N is the number of lattice sites, and ρ (0 ≤ ρ ≤ 1) is another random number. With the latter prescription, the unit time is, as usual, identified with MC step (MCS). On average, one MCS corresponds to N MC trials. All the MC runs were started with five cells located on the lattice at random (for the representative results with rapid diffusion, the details of the initial arrangement of cells are insignificant). To characterize the cell concentration, we use the average occupation of sites, ϑ. The MC runs were performed up to reaching ϑ = 0.4. In reality, the diffusion of cells is relatively rapid, and accordingly the value of pdiv should be small. To get an integrated view on the process, it is instructive, however, to show typical patterns and kinetics at various rates of diffusion. Following this line, we first show patterns for pdiv = 1, 10−2, and 10−4 in the absence of the cell-surface interaction, i.e. ϵcs = 0 (Figs 5–7). If the diffusion is negligible (pdiv = 1), the growth takes place around the seed cells and there are only a few aggregates (Fig 5). In the case of slow diffusion (pdiv = 10−2), the aggregates are numerous and their size is small (Fig 6). In the case of relatively rapid diffusion (pdiv = 10−4), the number of aggregates becomes smaller and their size is larger (Fig 7). 10.1371/journal.pone.0161915.g005Fig 5 Monte Carlo simulation of cell proliferation and spheroid formation in the microwells. Cross sections of a microwell along the main axis at ϑ = 0.2 (a) and 0.4 (b) in the absence of cell diffusion and cell-surface adhesion (pdiv = 1, ϵcs = 0). The well and cells are represented by black and red circles, respectively. 10.1371/journal.pone.0161915.g006Fig 6 Monte Carlo simulation of cell proliferation and spheroid formation in the microwells. Cross sections of a microwell along the main axis at ϑ = 0.2 (a) and 0.4 (b) with slow cell diffusion (pdiv = 10−2) and no cell-surface adhesion (ϵcs = 0). 10.1371/journal.pone.0161915.g007Fig 7 Monte Carlo simulation of cell proliferation and spheroid formation in the microwells. Cross sections of a microwell along the main axis at ϑ = 0.2 (a) and 0.4 (b) with relatively rapid cell diffusion (pdiv = 10−4) and no cell-surface adhesion (ϵcs = 0). Secondly, using pdiv = 10−4 and ϵcc = ϵcs = –1.2, we show the likely role of the cell-surface adhesion (Fig 8). In this case, the cell-surface adhesion is appreciable and the cells are located primarily near the walls of microwells. 10.1371/journal.pone.0161915.g008Fig 8 Monte Carlo simulation of cell proliferation and spheroid formation in the microwells. Cross sections of a microwell along the main axis at ϑ = 0.2 (a) and 0.4 (b) in the presence of adhesion (ϵcs = –1.2) of cells to the bottom and side area of the microwell. The adhesion to the top boundary is neglected because in reality the corresponding plane crosses the solution. The diffusion of cells is relatively rapid (pdiv = 10−4). The well and cells are represented by black and red circles, respectively. In the simulations described above, the aggregates are located either at random (Figs 5–7) or near the walls (Fig 8). In our experiments, however, the cells aggregate primarily at the bottom of microwells near the center (Fig 2). This feature can be reproduced by taking the gravity into account. This factor is usually ignored in MC simulations of cell proliferation, and it was ignored in our simulations above as well. To mimic the effect of gravity, we reduce the probability of jumps upwards by a factor of exp(ϵg), where ϵg < 0 is the corresponding effective energy. The results of MC simulations with pdiv = 10−4, ϵcc = –1.2, ϵcs = 0, and ϵg = –0.03 show that due to the gravity the proliferation takes place near the bottom of a microwell (Fig 9). 10.1371/journal.pone.0161915.g009Fig 9 Monte Carlo simulation of cell proliferation and spheroid formation in the microwells. Cross sections of a well along the main axis at ϑ = 0.2 (a) and 0.4 (b) in the absence of cell-surface adhesion (ϵcs = 0). Cell diffusion is relatively rapid (pdiv = 10−4). In addition, the gravity is taken into account (ϵg = –0.03). Figs 5–9 exhibit typical cell growth and aggregation patterns obtained in simulations. The corresponding kinetics of cell proliferation are shown in Fig 10. In the absence of diffusion, the kinetics are slow. With diffusion, the kinetics become faster. The growth of the cell population is predicted to be exponential initially and then becomes close to linear. 10.1371/journal.pone.0161915.g010Fig 10 Kinetics of cell growth in the Monte Carlo simulations. Lines 1–5 represent the cell growth kinetics of MC runs shown in Figs 5–9, respectively. Conclusion In this work, using a new microwell system for cell proliferation and MC simulations, we have demonstrated that the combination of experimental observation and modeling is a powerful tool for the study of cell spheroid formation. The results suggest that in the 3D cell culture platform used in our study, the cell spheroids would form on the basis of strong cell-cell interaction and weak cell-surface adhesion, and growth of spheroids mainly depends on the cell proliferation on their periphery. The likely role of the gravity in the spheroid formation was demonstrated as well. We believe that this strategy could potentially be applied to simulate even more complicated cell systems in the future. Furthermore, the capability to direct specific cells to the formation of aggregates of different types on diverse platforms, including cell spheroids in the microwells (as demonstrated in this study), cell spheroids in hydrogels of different stiffness [28], and multilayer cell sheets in biofunctionalized 3D microscaffold [4], could have significant implications for the construction and development of in vitro tissue models. This work was supported by National Research Foundation of Singapore. ==== Refs References 1 Lee J , Cuddihy MJ , Kotov NA . Three-dimensional cell culture matrices: state of the art . Tissue Eng Part B Rev . 2008 ;14 (1 ):61 –86 . 10.1089/teb.2007.0150 18454635 2 Varner VD , Nelson CM . Toward the directed self-assembly of engineered tissues . Annu Rev Chem Biomol Eng . 2014 ;5 :507 –26 . 10.1146/annurev-chembioeng-060713-040016 24797818 3 O'Brien CM , Holmes B , Faucett S , Zhang LG . Three-dimensional printing of nanomaterial scaffolds for complex tissue regeneration . Tissue Eng Part B Rev . 2015 ;21 (1 ):103 –14 . 10.1089/ten.TEB.2014.0168 25084122 4 Kim MH , Kumar SK , Shirahama H , Seo J , Lee JH , Zhdanov VP , et al Biofunctionalized hydrogel microscaffolds promote 3D hepatic sheet morphology . Macromol Biosci . 2016 ;16 (3 ):314 –21 . 10.1002/mabi.201500338 26612190 5 Timmins NE , Harding FJ , Smart C , Brown MA , Nielsen LK . Method for the generation and cultivation of functional three-dimensional mammary constructs without exogenous extracellular matrix . Cell Tissue Res . 2005 ;320 (1 ):207 –10 . 15714278 6 Kim C , Chung S , Kim YE , Lee KS , Lee SH , Oh KW , et al Generation of core-shell microcapsules with three-dimensional focusing device for efficient formation of cell spheroid . Lab Chip . 2011 ;11 (2 ):246 –52 . 10.1039/c0lc00036a 20967338 7 Wong SF , No da Y , Choi YY , Kim DS , Chung BG , Lee SH . Concave microwell based size-controllable hepatosphere as a three-dimensional liver tissue model . Biomaterials . 2011 ;32 (32 ):8087 –96 . 10.1016/j.biomaterials.2011.07.028 21813175 8 Kelm JM , Timmins NE , Brown CJ , Fussenegger M , Nielsen LK . Method for generation of homogeneous multicellular tumor spheroids applicable to a wide variety of cell types . Biotechnol Bioeng . 2003 ;83 (2 ):173 –80 . 12768623 9 Goodwin TJ , Jessup JM , Wolf DA . Morphologic differentiation of colon carcinoma cell lines HT-29 and HT-29KM in rotating-wall vessels . In Vitro Cell Dev Biol . 1992 ;28A (1 ):47 –60 . 1730571 10 Becker JL , Prewett TL , Spaulding GF , Goodwin TJ . Three-dimensional growth and differentiation of ovarian tumor cell line in high aspect rotating-wall vessel: morphologic and embryologic considerations . J Cell Biochem . 1993 ;51 (3 ):283 –9 . 8501130 11 Brown LA , Arterburn LM , Miller AP , Cowger NL , Hartley SM , Andrews A , et al Maintenance of liver functions in rat hepatocytes cultured as spheroids in a rotating wall vessel . In Vitro Cell Dev Biol Anim . 2003 ;39 (1–2 ):13 –20 . 12892522 12 Deisboeck TS , Wang Z , Macklin P , Cristini V . Multiscale cancer modeling . Annu Rev Biomed Eng . 2011 ;13 :127 –55 . 10.1146/annurev-bioeng-071910-124729 21529163 13 Swat MH , Thomas GL , Belmonte JM , Shirinifard A , Hmeljak D , Glazier JA . Multi-scale modeling of tissues using CompuCell3D . Methods Cell Biol . 2012 ;110 :325 –66 . 10.1016/B978-0-12-388403-9.00013-8 22482955 14 Urdy S . On the evolution of morphogenetic models: mechano-chemical interactions and an integrated view of cell differentiation, growth, pattern formation and morphogenesis . Biol Rev Camb Philos Soc . 2012 ;87 (4 ):786 –803 . 10.1111/j.1469-185X.2012.00221.x 22429266 15 Scianna M , Bell CG , Preziosi L . A review of mathematical models for the formation of vascular networks . J Theor Biol . 2013 ;333 :174 –209 . 10.1016/j.jtbi.2013.04.037 23684907 16 Wu J , Rostami MR , Tzanakakis ES . Stem cell modeling: From gene networks to cell populations . Curr Opin Chem Eng . 2013 ;2 (1 ):17 –25 . 23914346 17 Jeong D , Yun A , Kim J . Mathematical model and numerical simulation of the cell growth in scaffolds . Biomech Model Mechanobiol . 2012 ;11 (5 ):677 –88 . 10.1007/s10237-011-0342-y 21830072 18 Makhaniok A , Haranava Y , Goranov V , Panseri S , Semerikhina S , Russo A , et al In silico prediction of the cell proliferation in porous scaffold using model of effective pore . Biosystems . 2013 ;114 (3 ):227 –37 . 10.1016/j.biosystems.2013.10.001 24141144 19 Ehsan SM , George SC . Nonsteady state oxygen transport in engineered tissue: implications for design . Tissue Eng Part A . 2013 ;19 (11–12 ):1433 –42 . 10.1089/ten.TEA.2012.0587 23350630 20 Zhdanov VP , Kasemo B . Simulation of the growth and differentiation of stem cells on a heterogeneous scaffold . Phys Chem Chem Phys . 2004 ;6 (17 ):4347 –50 . 21 Robu A , Aldea R , Munteanu O , Neagu M , Stoicu-Tivadar L , Neagu A . Computer simulations of in vitro morphogenesis . Biosystems . 2012 ;109 (3 ):430 –43 . 10.1016/j.biosystems.2012.06.002 22732329 22 Zhdanov VP . Simulation of proliferation and differentiation of cells in a stem-cell niche . Physica A . 2008 ;387 (24 ):6126 –36 . 23 Barrio RA , Romero-Arias JR , Noguez MA , Azpeitia E , Ortiz-Gutierrez E , Hernandez-Hernandez V , et al Cell Patterns Emerge from Coupled Chemical and Physical Fields with Cell Proliferation Dynamics: The Arabidopsis thaliana Root as a Study System . Plos Comput Biol . 2013 ;9 (5 ):e1003026 10.1371/journal.pcbi.1003026 23658505 24 de Rooij DG , van Beek MEAB . Computer Simulation of the Rodent Spermatogonial Stem Cell Niche . Biol Reprod . 2013 ;88 (5 ):131 10.1095/biolreprod.113.108639 23536371 25 Ovadia J , Nie Q . Numerical Methods for Two-Dimensional Stem Cell Tissue Growth . J Sci Comput . 2014 ;58 (1 ):149 –75 .24415847 26 Sun Y , Yang XF , Wang Q . In-silico analysis on biofabricating vascular networks using kinetic Monte Carlo simulations . Biofabrication . 2014 ;6 (1 ):015008 10.1088/1758-5082/6/1/015008 24429898 27 Blight KJ , McKeating JA , Rice CM . Highly permissive cell lines for subgenomic and genomic hepatitis C virus RNA replication . J Virol . 2002 ;76 (24 ):13001 –14 . 12438626 28 Lee BH , Kim MH , Lee JH , Seliktar D , Cho NJ , Tan LP . Modulation of Huh7.5 spheroid formation and functionality using modified PEG-based hydrogels of different stiffness . PLoS One . 2015 ;10 (2 ):e0118123 10.1371/journal.pone.0118123 25692976 29 Kim MH , Kumar SK , Shirahama H , Seo J , Lee JH , Cho N-J . Phenotypic regulation of liver cells in a biofunctionalized three-dimensional hydrogel platform . Integrative Biology . 2016 ;8 :156 –66 . 10.1039/c5ib00269a 26792030 30 Gerdes J , Schwab U , Lemke H , Stein H . Production of a mouse monoclonal antibody reactive with a human nuclear antigen associated with cell proliferation . Int J Cancer . 1983 ;31 (1 ):13 –20 . 6339421 31 Hall PA , Woods AL . Immunohistochemical markers of cellular proliferation: achievements, problems and prospects . Cell Tissue Kinet . 1990 ;23 (6 ):505 –22 . 2276170 32 Kaita KD , Pettigrew N , Minuk GY . Hepatic regeneration in humans with various liver disease as assessed by Ki-67 staining of formalin-fixed paraffin-embedded liver tissue . Liver . 1997 ;17 (1 ):13 –6 . 9062874 33 Gerdes J , Lemke H , Baisch H , Wacker HH , Schwab U , Stein H . Cell cycle analysis of a cell proliferation-associated human nuclear antigen defined by the monoclonal antibody Ki-67 . J Immunol . 1984 ;133 (4 ):1710 –5 . 6206131 34 Sasaki K , Murakami T , Kawasaki M , Takahashi M . The cell cycle associated change of the Ki-67 reactive nuclear antigen expression . J Cell Physiol . 1987 ;133 (3 ):579 –84 . 3121642 35 Chang TT , Hughes-Fulford M . Monolayer and spheroid culture of human liver hepatocellular carcinoma cell line cells demonstrate distinct global gene expression patterns and functional phenotypes . Tissue Eng Part A . 2009 ;15 (3 ):559 –67 . 10.1089/ten.tea.2007.0434 18724832 36 Lee J , Cuddihy MJ , Cater GM , Kotov NA . Engineering liver tissue spheroids with inverted colloidal crystal scaffolds . Biomaterials . 2009 ;30 (27 ):4687 –94 . 10.1016/j.biomaterials.2009.05.024 19524294 37 Argyrakis P , Groda YG , Bokun GS , Vikhrenko VS . Thermodynamics and diffusion of a lattice gas on a simple cubic lattice . Phys Rev E . 2001 ;64 (6 ):066108 . 38 Zhdanov VP , Kasemo B . From a pluripotent stem cell to an ensemble of differentiated cells: Elements of theoretical tissue engineering . Phys Chem Chem Phys . 2004 ;6 (1 ):138 –43 .

PMC005xxxxxx/PMC5003352.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757135010.1371/journal.pone.0155203PONE-D-15-55425Research ArticlePhysical sciencesChemistryChemical compoundsOrganic compoundsVitaminsVitamin DPhysical sciencesChemistryOrganic chemistryOrganic compoundsVitaminsVitamin DPhysical SciencesMaterials ScienceMaterials by AttributeSurfactantsMedicine and Health SciencesSurgical and Invasive Medical ProceduresObstetric ProceduresCesarean SectionBiology and Life SciencesPhysiologyPhysiological ParametersBody WeightBirth WeightMedicine and Health SciencesPhysiologyPhysiological ParametersBody WeightBirth WeightBiology and Life SciencesBiochemistryLipidsPhospholipidsMedicine and Health SciencesWomen's HealthMaternal HealthPregnancyMedicine and Health SciencesWomen's HealthObstetrics and GynecologyPregnancyBiology and life sciencesNutritionNutritional deficienciesVitamin D deficiencyMedicine and health sciencesNutritionNutritional deficienciesVitamin D deficiencyBiology and Life SciencesNutritionDietMedicine and Health SciencesNutritionDietVitamin D Depletion in Pregnancy Decreases Survival Time, Oxygen Saturation, Lung Weight and Body Weight in Preterm Rat Offspring Maternal Vitamin D Depletion and PrematurityLykkedegn Sine 12Sorensen Grith Lykke 3Beck-Nielsen Signe Sparre 12Pilecki Bartosz 3Duelund Lars 4Marcussen Niels 5Christesen Henrik Thybo 12*1 Hans Christian Andersen Children’s Hospital, Odense University Hospital, Odense, Denmark2 Clinical Institute, Faculty of Health Sciences, University of Southern Denmark, Odense, Denmark3 Institute of Molecular Medicine, Department of Cancer and Inflammation, Faculty of Health Sciences, University of Southern Denmark, Odense, Denmark4 MEMPHYS, University of Southern Denmark, Odense, Denmark5 Institute of Pathology, Odense University Hospital, Odense, DenmarkKanellopoulos-Langevin Colette EditorXavier Bichat Medical School, INSERM-CNRS - Université Paris Diderot, FRANCECompeting Interests: The authors confirm that the financial support by Takeda Pharma does not alter the authors' adherence to PLOS ONE policies on sharing data and material. Conceptualization: SL GLS SSBN HTC. Data curation: SL. Formal analysis: SL GLS SSBN HTC. Funding acquisition: SL HTC. Investigation: SL. Methodology: SL GLS SSBN HTC. Project administration: SL GLS HTC. Resources: SL GLS BP LD NM. Supervision: GLS SSBN HTC. Validation: SL NM LD BP. Visualization: SL HTC. Writing – original draft: SL. Writing – review & editing: SL GLS SSBN LD BP NM HTC. * E-mail: henrik.christesen@rsyd.dk29 8 2016 2016 11 8 e015520322 12 2015 28 7 2016 © 2016 Lykkedegn et al2016Lykkedegn et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Animal studies suggest a role of vitamin D in fetal lung development although not studied in preterm animals. We tested the hypothesis that vitamin D depletion aggravates respiratory insufficiency in preterm rat offspring. Furthermore, the effects of vitamin D depletion on growth and lung surfactant were investigated. Female Sprague-Dawley rats were randomly assigned low vitamin D (VDL) or control diet before mating and followed with serum 25-hydroxyvitamin D (s-25(OH)D) determinations. After cesarean section at gestational day 19 (E19) or day 22 (E22), placental weight, birth weight, crown-rump-length (CRL), oxygenation (SaO2) at 30 min and survival time were recorded. The pup lungs were analyzed for phospholipid levels, surfactant protein A-D mRNA and the expression of the vitamin D receptor (VDR). S-25(OH)D was significantly lower in the VDL group at cesarean section (12 vs. 30nmol/L, p<0.0001). Compared to the controls, E19 VDL pups had lower birth weight (2.13 vs. 2.29g, p<0.001), lung weight (0.09 vs. 0.10g, p = 0.002), SaO2 (54% vs. 69%, p = 0.002) as well as reduced survival time (0.50 vs. 1.25h, p<0.0001). At E22, the VDL-induced pulmonary differences were leveled out, but VDL pups had lower CRL (4.0 vs. 4.5cm, p<0.0001). The phospholipid levels and the surfactant protein mRNA expression did not differ between the dietary groups. In conclusion, Vitamin D depletion led to lower oxygenation and reduced survival time in the preterm offspring, associated with reduced lung weight and birth weight. Further studies of vitamin D depletion in respiratory insufficiency in preterm neonates are warranted. http://dx.doi.org/10.13039/100007403Dagmar Marshalls Fond (DK)Lykkedegn Sine http://dx.doi.org/10.13039/501100003035Aase og Ejnar Danielsens Fond (DK)Lykkedegn Sine Takeda Pharma (DK)Lykkedegn Sine http://dx.doi.org/10.13039/501100006197Fonden til Lægevidenskabens FremmeLykkedegn Sine http://dx.doi.org/10.13039/501100004196Odense Universitetshospital (DK)Christesen Henrik Thybo All studies were financially supported by: Dagmar Marshalls foundation, Aase & Ejner Danielsens foundation, Takeda Pharma, The A.P. Moeller Foundation for the Advancement of Medical Science and Odense University Hospital. The funding sources had no involvement in collection, analysis and interpretation of data, writing of the report, or decision to submit the article for publication. Data AvailabilityAll relevant data are presented within the paper thus fully available. Additional data may be requested by contacting the corresponding author (henrik.christesen@rsyd.dk).Data Availability All relevant data are presented within the paper thus fully available. Additional data may be requested by contacting the corresponding author (henrik.christesen@rsyd.dk). ==== Body Introduction At birth, adequate lung development and growth is the key to successful adaptation to extrauterine life [1]. Thus, appropriate lung function depends on both proper development of the lung structure as well as sufficient surfactant synthesis and secretion [1, 2]. In preterm neonates inadequate lung maturation leads to respiratory distress syndrome (RDS) characterized by structural immaturity and inadequate synthesis and secretion of surfactant [2]. Antenatal corticosteroids [3, 4], intratracheal surfactant [5] and nasal continuous positive airway pressure (nCPAP) or mechanical ventilation [6, 7] are important in modern treatment of RDS. Yet, respiratory insufficiency due to lung immaturity is still a major cause of mortality in extremely preterm neonates [6], and research in other treatment modalities are needed. Placenta plays a vital role in normal fetal development, and fetal or neonatal disorders may be a result of placental insufficiency. Hypovitaminosis D, most commonly defined as s-25(OH)D levels below 50 nmol/L, is frequent in both pregnant women and preterm neonates [8–13]. Some human studies have shown an association between low s-25(OH)D levels during pregnancy and reduced placental development and weight [14–16], while others have not [17, 18]. Both placental weight and placental weight/birth weight (PW/BW) ratio have been used to describe the growth conditions for the fetus [19–22]. In humans, a high PW/BW ratio has been associated with RDS, low apgar scores and increased risk of admission to the neonatal intensive care unit [21, 23]. Only one human study [17] has approached the association between vitamin D and PW/BW ratio, however, without positive findings. The impact of calcitriol (1,25(OH)2D), the active metabolite of vitamin D) on early lung development and maturation as well as on development of lung diseases in early life is an emerging field of research. A recent human study found an association between s-25(OH)D levels below 30 nmol/L and increased oxygenation requirement and greater need of assisted ventilation in preterm neonates [24]. Yet, the human evidence on the impact of 1,25(OH)2D on lung development and maturation remains sparse. In contrary, both animal and laboratory studies, primary based on vitamin D enrichment, have provided detailed insights into the mechanisms through which 1,25(OH)2D stimulates the development and maturation of the lung [3]. In vivo studies in mice [25] and rats [26, 27] have shown that vitamin D deficiency during pregnancy causes alterations in lung growth and structure in term offspring. In vitro studies of cells from fetal rat lung explants have shown an impact of vitamin D on the embryogenesis and cellular growth and differentiation, including surfactant synthesis and secretion [28–35]. Laboratory studies on human pulmonary adenocarcinoma-derived cell lines support these findings [34, 36, 37]. However, in vivo studies of the role of vitamin D in lung development in preterm offspring have not been performed. We therefore tested the hypothesis that vitamin D depletion in pregnancy does not aggravate respiratory insufficiency in the preterm rat offspring. Materials and Methods Animals Sprague-Dawley rat dams were purchased from Taconic Biosciences (Denmark) and housed in a facility without windows and light without UV-B radiation. All fluorescence lamps were shatterproof and had built-in UV-filters (Phillips MASTER TL-D Secura 58W/830 (150cm)). Room air was kept at 21–24 degrees Celsius and the animals were exposed to day-night cycles alternating every 12 hours. The animals (n = 34) were randomly assigned to two different dietary groups; a low vitamin D (VDL) diet group (n = 18) (<5 IU/kg cholecalciferol, purified vitamin D3 deficient diet, Art No E15312-24) and a control diet group (n = 16) (1500 IU/kg cholecalciferol, purified control diet, Art No E15000-04). Both diets were designed for maintenance, were identical in terms of all other components including calcium and phosphorus and were obtained from Ssniff, Soest, Germany. Adult ten-week-old females were fed ad libitum for five weeks with the assigned diet before mating with adult males and continued throughout the pregnancy. The males were maintained on a standard rat chow (600 IU/kg cholecalciferol, Art No #1320, Altromin Spezialfutter, Lage, Germany). The pregnant females from each dietary group were randomly assigned to either the E19 or E22 subgroup immediately after the appearance of vaginal plug (gestational day 0 (E0)). Blood samples were collected from the tail vein before mating and after cesarean section before euthanizing. To observe the welfare, the weight of the pregnant rats was followed throughout the study. In a supplementary study, a whole-body dual-energy x-ray absorptiometry (DXA) scan of some of the pups from each dietary group was performed to describe the bone mineral content. The Danish National Animal Experiments Inspectorate approved the study (permit number: 2012-15-2934-00243). Cesarean section and survival On gestational day 19 (E19) or 22 (E22)—roughly equivalent to human gestational week 24 and term, pregnant mothers were anesthetized with isoflurane inhalation and the pups were recovered by cesarean section. Maternal oxygen saturation (SaO2) was monitored during the procedure and all efforts were made to minimize suffering. After removal of the pups, serum samples from the mothers were collected by intracardiac puncture followed by euthanization with pentobarbital sodium. Newborn pups were immediately wiped dry with a cloth and placed on a heating pad. After the cesarean section, the placenta and pups were weighed and the pup crown-rump length (CRL) was assessed within 30 minutes after birth. SaO2 was measured at 30 minutes after birth, and the pups were randomly, selected to be sacrificed immediately hereafter or kept with foster mother rats. The lungs from the euthanized pups were weighed and dissected for qPCR (quantitative real-time PCR) analysis, immunohistochemical analysis and phospholipid analysis. An investigator blinded to the dietary group monitored the survival of the non-sacrificed pups with a frequency of 15 minutes for the first two hours. The pups were euthanized if no movements, low heart rate and undetectable low SaO2 were observed. To observe the welfare and describe the survival time of the E22 pups, the blinded investigator registered daily weight, the ability to suckle and the activity level the first seven days after birth. Our human endpoints were observed distress and/or affected welfare described as weight loss > 20% within the first two days of life, inactivity and/or ruffled fur. In case of doubt, a veterinarian examined the pups. Oxygen saturation SaO2 was measured by using Nonin Pulse Oximeter® on the rats and pups in room air. The sensor was attached to the precordial site, and the animals were placed in a prone position to maintain a good attachment of the sensor. During the measurements, the pups were kept on a heating pad to avoid hypothermia. The readings of SaO2 were accepted as valid when the simultaneously monitored heart rate was stable. Each measurement was finished within 10–15 seconds to avoid desaturation secondary to prolonged measurement. For each animal, three to five measurements were taken within 2 minutes and the highest value of SaO2 was used for analysis. Serum 25(OH)D, calcium and phosphorus levels Serum samples were centrifuged and kept frozen at -20°C until analysis. All analyses were made after the study was finished. While the levels of s-25(OH)D were measured by liquid chromatography-tandem mass spectrometry (LC-MS/MS) [38], serum phosphorus and calcium levels were measured using a Roche Cobas 8000 Autoanalyzer Spectrophotometric (Roche Diagnostics®). qPCR analysis To quantify mRNA expression of the surfactant protein A-D genes (sftpa, sftpb, sftpc, sftpd), total RNA was extracted from homogenized lung tissue using TRIzol reagent (Life Technologies) according to the manufacturer’s instructions. Two μg RNA were used for cDNA production. Reverse transcription was performed using M-MLV Reverse Transcriptase (Sigma), and the reaction product was diluted to a final concentration of 10 ng/μl. Real-time PCR was performed in duplicates using the TaqMan Universal PCR Master Mix and TaqMan Gene Expression Assays specific for the given gene. The assay kits used were as follows: Gapdh, Rn01775763_g1; Sftpa1, Rn00824545_m1; Sftpb, Rn00684778_m1; Sftpc, Rn00569225_m1; Sftpd, Rn00563557_m1. The results were calculated using the 2ΔΔCt method. Phospholipid levels Isolated lungs were snap-frozen in liquid nitrogen and stored at -80°C until analysis. Lipid extraction was performed using a MTBE/methanol/ammonium acetate protocol [39]. Afterwards, the phospholipid concentration was determined in duplicates using a modification of the Bartlett method [40]. Briefly, 0.65 ml 70% perchloric acid was added to the dried samples followed by heating at 190°C for 60 minutes. Thereafter, the cooled samples were mixed with 3.3 ml water, 0.5 ml 2.5% ammonium molybdate and 0.5 ml 10% ascorbic acid followed by incubation at 100°C for 10 minutes. Finally, 250 μl from each cooled sample was transferred to a 96-well microtiter plate and absorbance was read at 800nm. Immunohistochemical detection of VDR Isolated lungs were immersed in 4% paraformaldehyde at room temperature for 24 hours fixation, processed, embedded in paraffin wax and cut into sections of 1.5 μm thickness. Before nuclear staining with Mayer’s hematoxylin, the sections were pretreated with MBO/TEG for 15 minutes followed by D-6 monoclonal antibody against VDR (dilution 1:500, Santa Cruz Biotechnologies cat. no. sc-13133). The sections were scanned, and microscopic analysis was performed using newCAST software (Visiopharm, Hoersholm, Denmark). Each section was analyzed according to previously described procedures [41] by an investigator blinded to the dietary group as well as the gestational age of the pups. The expression of VDR in the lung was estimated as the density of VDR-positive alveolar type II cells (cells/μm2 lung tissue). Statistics Shapiro-Wilks test revealed non-normal distribution of s-25(OH)D, maternal SaO2, CRL and lung weight. Between-group comparisons were made using two-way ANOVA and two-tailed unpaired t-tests, or non-parametric Mann-Whitney U test when appropriate. Survival analysis was performed using log-rank Mantel-Cox test. Normally distributed data were presented as mean ± SEM and non-normally distributed data as median and range. Spearman’s correlation was used to study within group correlations between SaO2 and LW, BW, LW/BW ratio and PW/BW ratio. A power calculation assuming an E19 SaO2 SD of 10, two-sided alpha 0.05 and beta 0.20, showed a need of 34 pups to detect a true difference in SaO2 of 10 between groups. As we only expected to be able to measure SaO2 in one pup per litter, mother rat n = 34 were chosen. Due to a substantial overlap in s-25(OH)D levels between the two dietary groups, we further performed a post hoc analysis. In this analysis associations between maternel s-25(OH)D at cesarean section and the measured phenotypical outcomes (SaO2, PW, BW, LW, LW/BW ratio, PW/BW ratio and CRL) were analyzed using linear uni- and multivariate regression models disregarding of initial dietary group. Moreover, we also performed a post hoc survival analysis comparing pups of mothers with s-25(OH)D <25nmol/L to pups of mothers with s-25(OH)D ≥25 nmol/L disregarding of initial dietary group. A p-value <0.05 was considered statistically significant; p = 0.05–0.10 was considered a trend. Data were analyzed using the Prism software package (version 6.0, GraphPad) and STATA software, version 13.1 (StataCorp, College Station, TX, USA). Results Maternal data Before mating, the median s-25(OH)D in the VDL group was lower compared to the controls, but no significant differences in s-total calcium or s-phosphorus were found between the two dietary groups, Table 1. However, the finding of lower s-25(OH)D values in the VDL group only reached significance in the E22 subgroup, S1 Table. 10.1371/journal.pone.0155203.t001Table 1 Maternal data. VDL group (n = 18) Controls (n = 16) p-value S-25(OH)D before mating (nmol/L) * 43 (24–130) 64 (42–118) 0.009 S-25(OH)D after CS (nmol/L) ** 12 (8–27) 30 (17–137) < 0.0001 S-total calcium (nmol/L) before mating * 2.64 ± 0.03 2.62 ± 0.02 0.547 S-total calcium (nmol/L) after CS** 2.72 ± 0.04 2.92 ± 0.07 0.018 S-phosphorus (nmol/L) before mating * 2.61 ± 0.08 2.84 ± 0.15 0.153 S-phosphorus (nmol/L) after CS** 1.84 ± 0.05 2.06 ± 0.08 0.028 Weight (g) at arrival to the laboratory 262 ± 3 263 ± 4 0.778 Weight (g) before CS 386 ± 9 386 ± 13 0.967 Maternal weight gain (g) 124 ± 9 122 ±13 0.896 Duration of maternal anesthesia (min) 7.59 ± 0.40 6.58 ± 0.60 0.158 Maternal SaO2 (%) 100 (98–100) 99 (95–100) 0.292 Litter size (n) 13 ± 0.8 12 ± 1.3 0.345 *Tail vein. **Intracardial puncture. Normal distributed data presented as mean ± SEM. Non-normal distributed data presented as median (range). Abbreviations: S-25(OH)D: serum 25-hydroxyvitamin D, CS: cesarean section; SaO2: oxygen saturation. At cesarean section, both s-25(OH)D levels and s-phosphorus levels had decreased in both dietary groups and s-total calcium levels increased, Table 1. At time of cesarean section, s-25(OH)D was significantly lower in the VDL animals compared to controls both at E19 and at E22, S1 Table. In the E19 control subgroup, the s-25(OH)D had a median of 25 nmol/L, e.g. an unintended moderate s-25(OH)D deficiency level, whereas the median in the E22 subgroup was 53 nmol/L, e.g. a normal s-25(OH)D level. S-total calcium and s-phosphorus became significantly lower in the VDL animals compared to controls at cesarean section in the E19 subgroup, but not at E22, S1 Table. Both s-total calcium and s-phosphorus were within the laboratory reference range for adult rodents at all times. All mother rats achieved a weight gain during the experiment with no significant difference in gestational weight gain between the groups. At the cesarean section, no differences were observed in maternal SaO2, the duration of maternal anesthesia or in the number of pups in each litter. Offspring data Half of the animals from each dietary group had cesarean section at E19, n = 65 (VDL) vs. n = 56 (controls): the other half at E22, n = 75 (VDL) vs. n = 49 (controls). When comparing preterm and term offspring in general, a significant difference was observed in all measurements (lung, birth and placental weights and CRL) as the results of maturation, Table 2. 10.1371/journal.pone.0155203.t002Table 2 Offspring data. E19 (n = 121) E22 (n = 124) p-value (E19 vs E22) VDL group (n = 65) Control group (n = 56) p-value VDL group (n = 75) Control group (n = 49) p-value VDL group Control group PW (g) 0.58 ± 0.01 0.58 ± 0.02 0.671 0.71 ± 0.01 0.76 ± 0.03 0.074 <0.001 <0.001 BW(g) 2.13 ± 0.03 2.29 ± 0.03 <0.001 5.21 ± 0.08 5.42 ± 0.18 0.245 <0.001 <0.001 LW(g) 0.09 ± 0.001 0.10 ± 0.002 0.002 0.15 ± 0.003 0.15 ± 0.004 0.543 <0.001 <0.001 CRL (cm) 2.8 (2.2–3.2) 2.9 (2.6–3.1) 0.072 4.0 (3.5–4.5) 4.5 (3.9–5.0) <0.001 <0.001 <0.001 PW/BW ratio 0.28 ± 0.006 0.25 ± 0.01 0.006 0.14 ± 0.003 0.14 ± 0.003 0.314 <0.001 <0.001 LW/BW ratio 0.04 ± 0.001 0.04 ± 0.001 0.011 0.03 ± 0.008 0.03 ± 0.001 0.060 <0.001 <0.001 Normal distributed data presented as mean ± SEM. Non-normal distributed data presented as median (range). Abbreviations: CRL; Crown-Rump Length, LW; lung weight, BW; birth weight, PW; placenta weight. When comparing the two dietary groups at E19 or E22, respectively, there was no difference between the weights of placenta, but a significant difference in PW/BW ratio was observed at E19, giving a higher ratio in the VDL pups compared to control pups due to lower body weight. At E19, birth weight, lung weight and lung weight/birth weight ratio (LW/BW ratio) were significantly lower in the pups born of VDL mothers than in the controls. While only insignificant difference in CRL was observed between the two dietary groups at E19, pups born at E22 of VDL mothers were significantly shorter than controls. Potential alterations in bone mineral content were not detectable using whole-body DXA scans (data not shown). All E22 weight measurements showed no difference between the dietary groups. Pups born at E19 had significantly (p < 0.001) lower 30 min SaO2 values than pups born at E22 (Fig 1). At E19, the pups of VDL mothers further had significantly (p = 0.002) lower SaO2 values than controls, whereas no significant difference was observed between the dietary groups at E22. 10.1371/journal.pone.0155203.g001Fig 1 SaO2 measured in rat pups 30 minutes after birth. SaO2 was measured 30 minutes after birth at E19 (VDL group (n = 13); control group (n = 23)) and E22 (VDL group (n = 28); control group (n = 16)). Within 2 minutes 3–5 measurements were made and the highest value of SaO2 was used for analysis. Each measurement was finished within 10–15 seconds to avoid desaturation secondary to prolonged measurement. At E19, the survival-rate was significantly (p < 0.0001) reduced in the VDL pups compared to controls and none of the pups survived more than 1.5 hours after birth (Fig 2). During the first 24 hours the majority of the E22 pups survived, leading to a 24h survival time of 74% in the VDL group and 84% in the control group. Further 3% of the pups in the VDL group died within day 2, after which all pups survived, leading to no significant difference in the long-term survival time between the groups born at E22 (Fig 2). 10.1371/journal.pone.0155203.g002Fig 2 Survival-rate in hours after birth. Kaplan-Meier survival analysis of both A) E19 (VDL group (n = 58); control group (n = 28)) and B) E22 (VDL group (n = 38); control group (n = 31)) pups. Comparison of survival curves using log-rank (Mantel-Cox) test. Results from correlation analysis are presented in Table 3. In the E19 VDL pups, we found a significant positive correlation between SaO2 and birth weight and a trend towards a negative correlation between PW/BW ratio and SaO2, but no correlation between lung weight and SaO2. In E19 control pups, a significant negative correlation between LW/BW ratio and SaO2 was observed. 10.1371/journal.pone.0155203.t003Table 3 Correlations between saturation and lung weight, birth weight, LW/BW ratio and PW/BW ratio. SaO2 E19 pups SaO2 E22 pups VDL group Control group VDL group Control group r p-value r p-value r p-value r p-value LW -0.05 0.864 0.35 0.349 0.08 0.663 0.12 0.655 BW 0.63 0.001 -0.18 0.558 -0.26 0.182 0.13 0.625 LW/BW ratio 0.09 0.782 -0.53 0.009 0.23 0.227 -0.08 0.763 PW/BW ratio -0.55 0.051 -0.37 0.080 0.34 0.076 0.04 0.892 r = Spearman’s coefficient. Abbreviations: LW; lung weight, BW; birth weight, PW; placenta weight. Pulmonary surfactant protein expression, phospholipids and VDR levels The qPCR analysis showed a significantly lower surfactant protein A-D mRNA levels (Fig 3) in lung tissue obtained from pups born at E19 compared to E22. The concentration of pulmonary phospholipids was also significantly lower in the E19 pups compared to the E22 pups (Fig 4). No significant differences in surfactant protein mRNA expression or phospholipid concentrations were observed when comparing the two dietary groups at either E19 or E22. 10.1371/journal.pone.0155203.g003Fig 3 Effects of vitamin D on mRNA levels of surfactant proteins. Quantitative real-time PCR analysis of surfactant protein A-D mRNA transcripts in fetal rat lung at E19, VDL group (n = 15); control group (n = 12) and at E22, VDL group (n = 15); control group (n = 11). Data were normalized against GAPDH. Results were calculated as mean ± SEM values. 10.1371/journal.pone.0155203.g004Fig 4 Effects of vitamin D on phospholipid levels. Phospholipid levels in fetal rat lung at E19, VDL group (n = 13); control group (n = 12) and at E22, VDL group (n = 15); control group (n = 9). Phospholipid levels were determined using a modification of the Bartlett method followed by absorbance read at 800nm. Immunohistochemical analysis showed that VDR was expressed in the ATII cells during the last days of gestation in both dietary groups. The density of VDR-positive ATII cells in lung tissue in both E19 dietary groups was significantly (p < 0.001) higher compared to E22. At E19, a trend (p = 0.068, Fig 5) towards lower levels of VDR was observed in the VDL group compared to controls. The expression of VDR in both E19 groups was significantly higher compared to E22 whereas no difference in VDR expression was found between the dietary groups at E22. 10.1371/journal.pone.0155203.g005Fig 5 Immunohistochemical analysis of VDR in fetal rat lung. Immunohistochemistry was performed on lung tissues from each dietary group at both E19 (VDL group (n = 31); control group (n = 27)) and E22 (VDL group (n = 35); control group (n = 19)) using the monoclonal D-6 antibody and counterstained with Mayer’s hematoxylin. Representative images of lung tissue from E19 (A) and E22 (B) (Magnification x28). Arrows indicate positive ATII cells. Positive (C) and negative (D) control sections with omission of the monoclonal D-6 antibody in the duodenum (Magnification x5). The expression of VDR in the lung was evaluated as the density of positive ATII cells (cells/μm2) (E). S-25(OH)D associations with pulmonary outcomes and survival Due to a substantial overlap in maternal s-25(OH)D between the two dietary groups at cesarean section, post hoc analyses on the association between s-25(OH)D and outcomes was performed for all animals disregarding initial dietary group, Table 4. 10.1371/journal.pone.0155203.t004Table 4 Uni- and multivariate linear regression models of the association between maternal s-25(OH)D at cesarean section and SaO2, PW, BW, LW, LW/BW ratio, PW/BW ratio and CRL. Univariate analyses Adjusted analyses s-25(OH)D at cesarean section s-25(OH)D at cesarean section E19 E22 E19 E22 β p-value β p-value β p-value β p-value SaO2* 1.87 <0.001 0.62 0.090 2.39 <0.001 1.13 0.113 BW** 0.01 0.001 0.03 <0.001 0.01 0.002 0.03 <0.001 PW** -0.00 0.334 0.01 <0.001 -0.00 0.100 0.01 <0.001 LW** 0.00 <0.001 -0.00 0.056 0.00 <0.001 -0.00 0.183 PW/BW** -0.00 0.001 0.00 <0.001 -0.00 <0.001 -0.00 0.302 LW/BW** 0.00 0.006 -0.00 <0.001 0.00 0.006 -0.00 <0.001 CRL** 0.00 0.168 -0.01 <0.001 -0.00 0.867 0.01 <0.001 All animals disregarding initial dietary group were included in the analyses. Adjusted for: * Maternal duration of anesthesia, ** Littersize and maternal weight. Abbreviations: β: β-coefficient, LW; lung weight, BW; birth weight, PW; placenta weight. At E19, s-25(OH)D was positively associated with SaO2, birth weight, lung weight and LW/BW ratio, and negatively associated with PW/BW ratio. At E22, maternal s-25(OH)D was positively associated with CRL, birth weight, placental weight and PW/BW ratio and negatively associated with LW/BW ratio. Adjustment for maternal weight and littersize did not change the associations at E19. But at E22 only birth weight, placental weight and CRL remained positive associated with s-25(OH) Moreover, a survival analysis comparing offspring of mothers with s-25(OH)D levels <25nmol/L at cesarean section to offspring of mothers with s-25(OH)D levels ≥25 nmol/L was performed (Fig 6). While the survival-rate was significantly (p = 0.0001) lower in the E19 pups of mothers with s-25(OH)D <25 nmol/L compared to s-25(OH)D ≥25 nmol/L, there was no significant difference between the pups of mothers with s-25(OH)D <25 nmol/L compared to s-25(OH)D ≥25 nmol/L at E22. 10.1371/journal.pone.0155203.g006Fig 6 Survival-rate of pups from mothers stratified by s-25(OH)D <25 nmol/L. Kaplan-Meier survival analysis of both A) E19 (s-25(OH)D <25 nmol/L (n = 58); ≥25 nmol/L (n = 43)) and B) E22 (s-25(OH)D <25 nmol/L (n = 58); ≥25 nmol/L (n = 28)) pups. Comparison of survival curves using log-rank (Mantel-Cox) test. Discussion We found that premature rat offspring from mothers with severe vitamin D deficiency had reduced oxygenation and survival-rate compared to offspring of mothers with moderate vitamin D deficiency. Potential explanatory findings were significantly lower lung weight, birth weight, LW/BW ratio higher PW/BW ratio and lower s-total calcium in the premature VDL pups, whereas no reduction in the expression of surfactant protein A-D mRNA, pulmonary phospholipid concentration, or VDR expression, was found. Furthermore, birth weight and PW/BW ratio correlated significantly to SaO2 levels at 30 min in E19 VDL pups. The VDL-induced pulmonary outcomes were only detectable in the premature pups and not in fully matured pups, but were supported by direct multivariate associations between s-25(OH)D levels and lung weight and SaO2 and decreased survival with s-25(OH)D deficiency in the premature. Our obtained results support that 25(OH)D is implicated in early fetal growth but are in contrast to the previously described positive impact of vitamin D on the synthesis and secretion of surfactant proteins and phospholipids [3]. We used an animal model of vitamin D deficiency during pregnancy, while most others either studied the in vitro synthesis of surfactant by incubating ATII cells with 1,25(OH)2D [29, 30, 33], or in vivo by vitamin D enrichment before sacrifice [26, 27]. Our model was designed to resemble the clinical setting of hypovitaminosis D in humans and we achieved a s-25(OH)D level of 12 nmol/L in the VDL mothers at cesarean section, which corresponds to severe vitamin D deficiency. In normal pregnant rats, levels of s-25(OH)D gradually decrease during the last period of gestation [42], as did in our control animals, which explains the substantial overlap in low s-25(OH)D between dietary groups. Our negative results on VDL-mediated surfactant protein A-D mRNA expression and pulmonary phospholipid levels may be attributed to failure of remaining s-25(OH)D concentrations >50 nmol/L in the E19 control group. Moreover, pups for lung studies were randomly chosen within the dietary groups, which had overlapping maternal s-25(OH)D concentration ranges. In rats, the synthesis of surfactant evolves in the saccular phase only between the 19th gestational day and term (day 21–22) [43]. In accordance, our observed surfactant protein A-D mRNA expression and phospholipid levels at E19 were very low, which may have induced a detection error of a true difference. However, at E22, we still found no difference between VDL pups and control pups. The observed lack of impact of vitamin D on these outcomes does not rule out that at higher s-25(OH)D concentrations in the saccular phase can have effects as observed by others [30, 33]. We further described the immunohistochemical localization and relative expression of VDR in the preterm pup and term pup lung. Previous studies have shown that the widely used antibody 9A7γ [32–34] not only binds to VDR, but also possesses non-specific interactions with yet unidentified proteins [44, 45]. Based on a parallel comparison of a large selection of VDR antibodies [44], we chose to use the mouse monoclonal antibody D-6, which possesses the highest specificity, sensitivity and versatility. Our study brings further evidence that the VDR is present in the ATII cells during the last period of gestation, but with unaffected expression by s-25(OH)D in the studied concentration range. Furthermore, our study showed a significantly higher expression of VDR in the E19 pups compared to E22 confirming previous observations of reduced VDR expression in the last days before term in rat offspring [34, 36]. Despite these null-findings for the surfactant synthesis and VDR expression, premature pups with the lowest s-25(OH)D levels had decreased oxygenation and survival time. In humans, the postmortem diagnosis of pulmonary hypoplasia is based on a low LW/BW ratio [46], and we found a strong negative correlation between LW/BW ratio and oxygenation in the premature pups. The premature VDL pups had both significantly reduced lung weight and LW/BW ratio, supporting that pulmonary hypoplasia was likely to be responsible for the decreased survival. The significant positive correlation between birth weight and oxygenation may further indicate a reduced muscle mass leading to earlier muscular fatigue. Moreover, muscular weakness associated to the well-known vitamin D deficiency-associated myopathy in both humans and rodents [47] and lower levels of s-total calcium may have been a contributing factor. We were, however, not able to observe differential signs of muscular fatigue, or increased respiratory distress with increased respiratory muscular work between the pup groups. E19 pups hardly moved and showed only very little respiratory effort at all and E22 pups showed no visual signs of muscular fatigue or respiratory distress. Decreased mineralization and hence more soft ribs as a further explanatory factor was not supported by our whole-body DXA-scans, which showed unchanged bone mineral content between vitamin D depleted offspring and controls at both E19 and E22 (data not shown). This finding is in accordance with findings by Zosky et al. [25]. The significantly reduced BW in the premature VDL pups may indicate a more general affection of organ development in utero as a result of the severe vitamin D deficiency as previously shown [3, 13, 48, 49]. In humans, a high PW/BW ratio has been associated with RDS, low Apgar scores and increased risk of admission to the neonatal intensive care unit [21, 23]. We showed that the reduced birth weight was not secondary to a reduced placental weight, which was unchanged. Although there are conflicting results in both animal and human data assessing the role of vitamin D in somatic growth [25, 48, 50, 51], there are several biological mechanisms possibly connecting maternal vitamin D to fetal growth. The vitamin D conversion enzyme gene, CYP27B1, and VDR are expressed in both placenta and lungs [50, 52]. In placenta, 1,25(OH)2D stimulates the secretion of placental hormones, which supports fetal growth and energy needs through delivery of calcium as well as glucose and fatty acids [17, 50]. In the lungs, 1,25(OH)2D stimulates pulmonary artery endothelial cell (PAEC) growth in the same manner as vascular endothelial growth factor (VEGF) [52]. Furthermore, 1,25(OH)2D regulates transforming growth factor β1 (TGF-β1), a known mediator of airway remodeling [53, 54]. The observed lack of differences in pulmonary and survival outcomes between dietary groups in mature pups may have several explanations: The VDR expression in the lungs was lower at E22 compared to E19 suggesting that pulmonary vitamin D effects are of minor importance at term compared to the preterm period. Moreover, term offspring do not suffer from respiratory insufficiency, why an effect on oxygenation and survival-rate should not be suspected in the mature lung. We observed reduced CRL in the VDL pups compared to control pups at E22, but not at E19. Direct association analyses across dietary groups between maternal s-25(OH)D and CRL at E22, but not at E19, supported these findings. Vitamin D deficiency per se, as well as VDR mutations and VDR knock-out, has shown not to affect serum mineral concentrations, bone mineralization and fetal longitudinal bone growth [55, 56]. Moreover, animal studies in severely vitamin D deficient rats [57–60] and VDR-null mice [61, 62] showed a normal increase in intestinal calcium absorption during pregnancy and normal mineral content of the term offspring. Thus, a low calcium effect on bone growth could not readily explain the significant VDL dependent shortening of CRL at E22. However, CRL is a measure of skull and spine growth rather than longitudinal bone growth and bone mineralization and our association analyses showed not only increased CRL, but also BW and PW at E22 with higher s-25(OH)D, suggesting a general growth-promoting effect of vitamin D at term. Strengths of the present study include the randomized, blinded design, the obtained severe deficiency level of s-25(OH)D in the VDL group, the between-group similarity in maternal weight gain, duration of anesthesia and maternal SaO2 during cesarean section. The difference in s-25-(OH)D between the dietary groups, was further designed to mimic differences in vitamin D status in the clinical setting rather than extreme vitamin D depletion versus pharmacological supplementation. Limitations included the large with-in group variation resulting in overlapping s-25(OH)D concentrations between the dietary groups. Yet, s-25(OH)D was associated with pup anthropometrics, SaO2 and survival time in our post hoc analyses, which supported the findings between the dietary groups. The females were ten weeks old and may not have been housed under conditions similar to ours, in terms of lighting and diet, before purchase potentially causing variation in s-25(OH)D which have a long half-life of 2–3 weeks [63, 64]. Variations in appetite may also have contributed to the relatively high within-group variation. To reduce the within group variations and ensure a larger difference between the dietary groups, future studies should assign the diets to young females immediately after the weaning from their mothers. We did not measure pup s-25(OH)D, because of technical volume limitations. However, maternal s-25(OH)D correlates strongly to offspring s-25(OH)D concentrations [65]. Lastly, the fragility of the lung tissue, especially at E19, did not allow in situ fixation of the premature lungs under physiological pressure enabling comparison of lung morphology between groups. The perspectives of our study include the emphasis of whole body animal models to study the effects of vitamin D on the lungs. Our results indicate that aggravation of respiratory failure may occur due to reduced lung and birth weight as the result of severe vitamin D depletion. However, we were unable to show a direct effect of vitamin D deficiency on surfactant measures or VDR expression. Future studies should pursue to evaluate our results using an optimized version of our model with animals assigned to the diets immediately after weaning. In conclusion, vitamin D depletion during pregnancy led to a lower SaO2 and shorter survival-rate in premature rat offspring despite no reduction in lung surfactant constituents. Explanatory factors include reduced lung weight, which may imply a reduced total lung diffusion area, and decreased birth weight, which may indicate a reduced muscle mass leading to earlier muscular fatigue. In support of the hypothesis, vitamin D alleviates respiratory insufficiency at preterm birth. Studies of vitamin D effects in human preterm neonates regarding respiratory insufficiency are warranted. Supporting Information S1 Table (DOCX) Click here for additional data file. The authors wish to thank animal technician Anne Mette Durand and medical student Kathrine Work Havelund for very helpful assistance. Abbreviations CRLcrown-rump length LW/BW ratiolung weight/birth weight ratio PW/BW ratioplacental weight/birth weight ratio S-25(OH)Dserum 25-hydroxyvitamin D SaO2oxygen saturation VDLlow vitamin D diet VDRvitamin D receptor ==== Refs References 1 Vannucchi C , Silva L , Lúcio C , Regazzi F , Veiga G , Angrimani D . Prenatal and Neonatal Adaptations with a Focus on the Respiratory System . Reproduction in Domestic Animals . 2012 ;47 :177 –81 . 2 Hermansen CL , Lorah KN . Respiratory Distress in the Newborn . American Family Physician . 2007 ;76 :987 –94 . 17956068 3 Lykkedegn S , Sorensen GL , Beck-Nielsen SS , Christesen HT . The impact of vitamin D on fetal and neonatal lung maturation. A systematic review . American journal of physiology Lung cellular and molecular physiology . 2015 ;308 :L587 –L602 . 10.1152/ajplung.00117.2014 25595644 4 Ballard PL , Ballard RA . Scientific basis and therapeutic regimens for use of antenatal glucocorticoids . Am J Obstet Gynecol . 1995 ;173 (1 ):254 –62 . 7631700 5 Halliday HL . Surfactants: past, present and future . Journal of perinatology: official journal of the California Perinatal Association . 2008 ;28 Suppl 1 :S47 –56 . Epub 2008/06/18. 10.1038/jp.2008.50 .18446178 6 Cools F , Offringa M , Askie LM . Elective high frequency oscillatory ventilation versus conventional ventilation for acute pulmonary dysfunction in preterm infants (Review) . Cochrane database Syst Rev . 2015 ;(3 ). 7 Verder H , Bohlin K , Kamper J , Lindwall R , Josson B . Nasal CPAP and surfactant for treatment of respiratory distress syndrome and prevention of bronchopulmonary dysplasia . Acta paediatrica . 2009 ;98 :1400 –8 . 10.1111/j.1651-2227.2009.01413.x 19572989 8 Andersen LB , Abrahamsen B , Dalgard C , Kyhl HB , Beck-Nielsen SS , Frost-Nielsen M , et al Parity and tanned white skin as novel predictors of vitamin D status in early pregnancy: a population-based cohort study . Clinical endocrinology . 2013 ;79 (3 ):333 –41 . Epub 2013/01/12. 10.1111/cen.12147 .23305099 9 Dawodu A , Nath R . High prevalence of moderately severe vitamin D deficiency in preterm infants . Pediatrics International . 2011 ;53 :207 –10 . 10.1111/j.1442-200X.2010.03209.x 20667028 10 Kaushal M , Magon N . Vitamin D in pregnancy: A metabolic outlook . Indian journal of endocrinology and metabolism . 2013 ;17 (1 ):76 –82 . Epub 2013/06/19. 10.4103/2230-8210.107862 23776856 11 Kazemi A , Sharifi F , Jafari N , Mousavinasab N . High Prevalence of Vitamin D Deficiency among Pregnant Women and their Newborns in an Iranian Population . Journal of Womens Health . 2009 ;18 (6 ):835 –9 . 12 Vandevijvere S , Amsalkhir S , Van Oyen H , Moreno-Reyes R . High Prevalence of Vitamin D Deficiency in Pregnant Women: A National Cross-Sectional Survey . PloS one . 2012 ;7 (8 ):e43868 Epub 2012/09/01. 10.1371/journal.pone.0043868 22937114 13 Hart PH , Lucas RM , Walsh JP , Zosky GR , Whitehouse AJO , Zhu K , et al Vitamin D in Fetal Development: Findings From a Birth Cohort Study . Pediatrics . 2015 ;135 (1 ):e167 –e73 . 10.1542/peds.2014-1860 25511121 14 Bodnar LM , Klebanoff MA , Gernand AD , Platt RW , Parks WT , Catov JM , et al Maternal Vitamin D Status and Spontaneous Preterm Birth by Placental Histology in the US Collaborative Perinatal Project . American Journal of Epidemiology . 2013 :1 –9 . 10.1093/aje/kwt237 15 Evans KN , Bulmer JN , Kilby MD , Hewison M . Vitamin D and Placental-Decidual Function . Journal of the Society for Gynecologic Investigation . 2004 ;11 :263 –71 . 10.1016/j.jsgi.2004.02.002 10.1016/j~jsgi.2004.02.002. 15219879 16 Evans KN , Nguyen L , Chan J , Innes BA , Bulmer JN , Kilby MD , et al Effects of 25-Hydroxyvitamin D3 and 1,25-Dihydroxyvitamin D3 on Cytokine Production by Human Decidual Cells . Biology of reproduction . 2006 ;75 (6 ):816 –22 . Epub 2006/09/08. 10.1095/biolreprod.106.054056 .16957024 17 Gernand AD , Simhan HN , Klebnoff MA , Bodnar LM . Maternal Serum 25-Hydroxyvitamin D and Measures of Newborn and Placenta Weight in a U.S. Multicenter Cohort Study . J Clin Endocrinol Metab . 2013 ;98 (1 ):398 –404 . 10.1210/jc.2012-3275 23162094 18 Javaid MK , Crozier SR , Harvey NC , Gale CR , Dennison EM , Boucher BJ , et al Maternal vitamin D status during pregnancy and childhood bone mass at age 9 years: a longitudinal study . Lancet . 2006 ;367 (9504 ):36 –43 . Epub 2006/01/10. 10.1016/s0140-6736(06)67922-1 .16399151 19 Godfrey K . The Role of the Placenta in Fetal Programming—A Review . Placenta . 2002 ;23 (suppl A ):s20 –7 . 11978056 20 Salafia C , Charles A , Maas E . Placenta and Fetal Growth Restriction . Clinical obstetrics and gynecology 2006 ;49 (2 ):236 –56 . 16721104 21 Shehata F , Levin I , Shrim A , Ata B , Weisz B , Gamzu R , et al Placenta/birthweight ratio and perinatal outcome: a retrospective cohort analysis . BJOG . 2011 ;118 :741 –7 . 10.1111/j.1471-0528.2011.02892.x 21332633 22 Myatt L . Placental adaptive responses and fetal programming . J Physiol . 2006 ;572 1 :25 –30 .16469781 23 Madkar C , Musale J , Deshpande H , Shitole R . A Study of placental weight and birth weight ratio (PW/BW) and it's effects on perinatal outcome . Indian Journal of Obstetrics and Gynaecology . 2015 ;2 (1 ):1 –6 . 24 Onwuneme C , Martin F , McCarthy R , Carroll A , Segurado R , Murphy J , et al The Association of Vitamin D Status with Acute Respiratory Morbidity in Preterm Infants . The Journal of Pediatrics . 2015 ;166 (5 ):1175 –80.e1 . 10.1016/j.jpeds.2015.01.055 25919726 25 Zosky GR , Berry LJ , Elliot JG , James AL , Gorman S , Hart PH . Vitamin D Deficiency Causes Deficits in Lung Function and Alters Lung Structure . American journal of respiratory and critical care medicine . 2011 ;183 :1336 –43 . 10.1164/rccm.201010-1596OC 21297070 26 Mandell E , Seedorf G , Gien J , Abman SH . Vitamin D Treatment Improves Survival and Infant Lung Structure After Intra-Amniotic Endotoxin Exposure in Rats: Potential Role for the Prevention of Bronchopulmonary . American journal of physiology Lung cellular and molecular physiology . 2014 ;306 (5 ):L420 –8 . Epub 2014/01/15. 10.1152/ajplung.00344.2013 24414254 27 Yurt M , Liu J , Sakurai R , Gong M , Husain SM , Siddiqui MA , et al Vitamin D Supplementation Blocks Pulmonary Structural and Functional Changes in a Rat Model of Perinatal Vitamin D deficiency . Am j Physiol lung Cell Mol Physiol 2014 ;307 :L859 –L67 . 10.1152/ajplung.00032.2014 25305247 28 Edelson JD , Chan S , Jassal D , Post M , Tanswell KA . Vitamin D stimulates DNA synthesis in alveolar type-II cells . Biochimica et Biophysica Acta . 1993 ;1221 :159 –66 . 29 Marin L , Dufour ME , Nguyen TM , Tordet C , Garabedian M . Maturational changes induced by 1 alpha,25-dihydroxyvitamin D3 in type II cells from fetal rat lung explants . American Journal of Physiology . 1993 ;265 :L45 –L52 . 8338181 30 Marin L , Dufour ME , Tordet C , Nguyen TM . 1,25(OH)2D3 stimulates Phospholipid Biosynthesis and Surfactant Release in Fetal Rat Lung Explants . Biol Neonate . 1990 ;57 :257 –60 . 2322608 31 Nguyen M , Guillozo H , Garabedian M , Balsan S . Lung as a Possible Additional Target Organ for Vitamin D during Fetal Life in the Rat . Biol Neonate . 1987 ;52 :232 –40 . 2823916 32 Nguyen TM , Guillozo H , Marin L , Dufour ME , Tordet C , Pike JW , et al 1,25-Dihydroxyvitamin D3 Receptors in Rat Lung during the Perinatal Period: Regulation and Immunohistochemical Localization . Endocrinology . 1990 ;127 :1755 –62 . 2169401 33 Nguyen TM , Guillozo H , Marin L , Tordet C , Koite S , Garabedian M . Evidence for a vitamin D paracrine system regulating maturation of developing rat lung epithelium . American Journal of Physiology-Lung Cellular and Molecular Physiology . 1996 ;15 :L392 –L9 . 34 Nguyen M , Trubert CL , Rizk-Rabin M , Rehan VK , Besancon F , Cayre YE , et al 1,25-Dihydroxyvitamin D3 and fetal lung maturation: immunogold detection of VDR expression in pneumocytes type II cells and effect on fructose 1,6 bisphosphatase . The Journal of steroid biochemistry and molecular biology . 2004 ;89-90 (1–5 ):93 –7 . Epub 2004/07/01. 10.1016/j.jsbmb.2004.03.054 .15225753 35 Sakurai R , Shin E , Fonseca S , Sakurai T , Litonjua AA , Weiss ST , et al 1α,25(OH)2D3 and its 3-epimer promote rat lung alveolar epithelial-mesenchymal interactions and inhibit lipofibroblast apoptosis . American journal of physiology Lung cellular and molecular physiology . 2009 ;297 (3 ):L496 –505 . Epub 2009/07/04. 10.1152/ajplung.90539.2008 19574420 36 Phokela SS , Peleg S , Moya FR , Alcorn JL . Regulation of human pulmonary surfactant protein gene expression by 1 α,25-dihydroxyvitamin D3 . American Journal of Physiology-Lung Cellular and Molecular Physiology . 2005 ;289 :617 –26 . 37 Rehan VK , Torday JS , Peleg S , Gennaro L , Vouros P , Padbury J , et al 1a,25-Dihydroxy-3-epi-vitamin D3, a natural metabolite of 1a,25-dihydroxy vitamin D3: production and biological activity studies in pulmonary alveolar type II cells . Molecular Genetics and Metabolism . 2002 ;76 :46 –56 . 12175780 38 Vogeser M , Parhofer KG . Liquid Chromatography Tandem-mass Spectrometry (LC-MS/MS)–Technique and Applications in Endocrinology . Exp Clin Endocrinol Diabetes . 2007 ;115 :559 –70 . 17943689 39 Abbott SK , Jenner AM , Mitchell TW , Brown SH , Halliday GM . An Improved High-Throughput Lipid Extraction Method for the Analysis of Human Brain Lipids . Lipids . 2013 ;48 (3 ):307 –18 . 10.1007/s11745-013-3760-z 23355308 40 Bartlett GR . Phosphorus Assay in Column Chromatography . J Biol chem . 1959 ;234 :466 –8 . 13641241 41 Ochs M , Mühlfeld C . Quantitative microscopy of the lung: a problem-based approach . Part 1: basic principles of lung stereology American journal of physiology Lung cellular and molecular physiology 2013 ;305 :L 15 –22 . 42 Clements MR , Fraser DR . Vitamin D Supply to the Rat Fetus and Neonate . J Clin Invest . 1988 ;81 :1768 –73 . 2838521 43 Samtani MN , Pyszczynski NA , Dubois DC , Almon RR , Jusko WJ . Modeling Glucocorticoid-Mediated Fetal Lung Maturation: I. Temporal Patterns of Corticosteroids in Rat Pregnancy . The Journal of pharmacology and experimental therapeutics . 2005 ;317 (1 ):117 –26 . Epub 2005/12/24. 10.1124/jpet.105.095851 .16371449 44 Wang Y , Becklund BR , DeLuca HF . Identification of a highly specific and versatile vitamin D receptor antibody . Arch Biochem Biophys . 2010 ;494 (2 ):166 –77 . Epub 2009/12/03. 10.1016/j.abb.2009.11.029 .19951695 45 Wang Y , Zhu J , DeLuca HF . Where is the vitamin D receptor? Arch Biochem Biophys . 2012 ;523 (1 ):123 –33 . Epub 2012/04/17. 10.1016/j.abb.2012.04.001 .22503810 46 De Paepe M , Friedman RM , Gundogan F , Pinar H . Postmortem Lung Weight/Body Weight Standards for Term and Preterm Infants . Pediatric Pulmononology . 2005 ;45 :445 –8 . 47 Polly P , Tan TC . The role of vitamin D in skeletal and cardiac muscle function . Frontiers in Physiology . 2014 ;5 :1 –6 .24478714 48 O'Loan J , Eyles DW , Kesby J , Ko P , McGrath JJ , Burne THJ . Vitamin D deficiency during various stages of pregnancy in the rat; its impact on development and behaviour in adult offspring . Psychoneuroendocrinology . 2007 ;32 :277 –34 . 49 Eyles DW , Burne THJ , McGrath JJ . Vitamin D, effects on brain development, adult brain function and the links between low levels of vitamin D and neuropsychiatric disease . Frontiers in Neuroendocrinology . 2013 ;34 :47 –64 . 10.1016/j.yfrne.2012.07.001 22796576 50 Lapillonne A . Vitamin D deficiency during pregnancy may impair maternal and fetal outcomes . Medical Hypotheses . 2010 ;74 :71 –5 . 10.1016/j.mehy.2009.07.054 19692182 51 Galthen-Sørensen M , Andersen L , Sperling L , Christesen H . Maternal 25-hydroxyvitamin D level and fetal bone growth assessed by ultrasound: a systematic review . Ultrasound Obstet Gynesol . 2014 ;44 :633 –40 . 52 Mandell E , Seedorf GJ , Ryan S , Gien J , Cramer SD , Abman SH . Antenatal Endotoxin Disrupts Lung Vitamin D Receptor and 25-hydroxyvitamin D 1-alpha Hydroxylase in the Developing Rat . American journal of physiology Lung cellular and molecular physiology . 2015 10.1152/ajplung.00253.2015 53 Chen G , Khalil N . TGF-beta1 increases proliferation of airway smooth muscle cells by phosphorylation of map kinases . Respiratory Research . 2006 ;7 (2 ):1 –10 .16390543 54 Foong RE , Shaw NC , Berry LJ , Hart PH , Gorman S , Zosky GR . Vitamin D deficiency causes airway hyperresponsiveness, increases airway smooth muscle mass, and reduces TGF-B expression in the lungs of female BALB/c mice . Physiological Reports . 2014 ;2 (3 ):1 –12 . 55 Kovacs C . The Role of Vitamin D in Pregnancy and Lactation: Insights from Animal Models and Clinical Studies .. The annual review of nutrition . 2012 ;(32 ):9.1 –9.27 . 56 Kovacs CS . Bone Development and Mineral Homeostasis in the Fetus and Neonate: Roles of the Calciotropic and Phosphotropic Hormones . Physiol Rev . 2014 ;94 :1143 –218 . 10.1152/physrev.00014.2014 25287862 57 Halloran BP , De Luca HF . Effect of vitamin D deficiency on skeletal development during early growth in the rat . Arch Biochem Biophys . 1981 ;209 (1 ):7 –14 . Epub 1981/06/01. .7283447 58 Brommage R , DeLuca HF . Placental transport of calcium and phosphorus is not regulated by vitamin D . The American journal of physiology . 1984 ;246 (4 Pt 2 ):F526 –9 . Epub 1984/04/01. .6720905 59 Miller SC , Halloran BP , DeLuca HF , Jee WS . Studies on the role of vitamin D in early skeletal development, mineralization, and growth in rats . Calcified tissue international . 1983 ;35 (4–5 ):455 –60 . Epub 1983/07/01. .6616315 60 Glazier JD , Mawer EB , Sibley CP . Calbindin-D9K gene expression in rat chorioallantoic placenta is not regulated by 1,25-dihydroxyvitamin D3 . Pediatr Res . 1995 ;37 (6 ):720 –5 . Epub 1995/06/01. 10.1203/00006450-199506000-00008 .7651755 61 Lieben L , Stockmans I , Moermans K , Carmeliet G . Maternal hypervitaminosis D reduces fetal bone mass and mineral acquisition and leads to neonatal lethality . Bone . 2013 ;57 (1 ):123 –31 . Epub 2013/07/31. 10.1016/j.bone.2013.07.029 .23895994 62 Kovacs CS , Woodland ML , Fudge NJ , Friel JK . The vitamin D receptor is not required for fetal mineral homeostasis or for the regulation of placental calcium transfer in mice . American journal of physiology Endocrinology and metabolism . 2005 ;289 (1 ):E133 –44 . Epub 2005/03/03. 10.1152/ajpendo.00354.2004 .15741244 63 Kovacs C . Maternal vitmain D deficiency: Fetal and neonatal implications . Seminars in Fetal and Neonatal Medicine . 2013 ;18 :129 –35 . 64 Zerwekh JE . Blood biomarkers of vitamin D status . Am j Clin Nutr . 2008 ;87 (suppl ):1087S –91S .18400739 65 Kovacs C . Maternal Mineral and Bone Metabolism During Pregnancy, Lactation, and Post-Weaning Recovery . Physiol Rev . 2016 ;96 :449 –547 . 10.1152/physrev.00027.2015 26887676

PMC005xxxxxx/PMC5003353.txt

==== Front PLoS Negl Trop DisPLoS Negl Trop DisplosplosntdsPLoS Neglected Tropical Diseases1935-27271935-2735Public Library of Science San Francisco, CA USA 2757125410.1371/journal.pntd.0004960PNTD-D-15-01358Research ArticleMedicine and Health SciencesTropical DiseasesNeglected Tropical DiseasesChikungunya InfectionMedicine and Health SciencesInfectious DiseasesViral DiseasesChikungunya InfectionBiology and Life SciencesPhysiologyImmune PhysiologyAntibodiesMedicine and Health SciencesPhysiologyImmune PhysiologyAntibodiesBiology and Life SciencesImmunologyImmune System ProteinsAntibodiesMedicine and Health SciencesImmunologyImmune System ProteinsAntibodiesBiology and Life SciencesBiochemistryProteinsImmune System ProteinsAntibodiesBiology and life sciencesOrganismsVirusesRNA virusesTogavirusesAlphavirusesChikungunya VirusBiology and Life SciencesMicrobiologyMedical MicrobiologyMicrobial PathogensViral PathogensTogavirusesAlphavirusesChikungunya VirusMedicine and Health SciencesPathology and Laboratory MedicinePathogensMicrobial PathogensViral PathogensTogavirusesAlphavirusesChikungunya VirusBiology and Life SciencesOrganismsVirusesViral PathogensTogavirusesAlphavirusesChikungunya VirusResearch and Analysis MethodsImmunologic TechniquesImmunoassaysEnzyme-Linked ImmunoassaysBiology and Life SciencesBiochemistryGlycobiologyGlycoproteinsE2 GlycoproteinsBiology and Life SciencesAnatomyBody FluidsBloodBlood SerumImmune SerumMedicine and Health SciencesAnatomyBody FluidsBloodBlood SerumImmune SerumBiology and Life SciencesPhysiologyBody FluidsBloodBlood SerumImmune SerumMedicine and Health SciencesPhysiologyBody FluidsBloodBlood SerumImmune SerumMedicine and Health SciencesHematologyBloodBlood SerumImmune SerumBiology and Life SciencesBiochemistryGlycobiologyGlycoproteinsBiology and Life SciencesBiochemistryProteinsRecombinant ProteinsAntigenic Variation of East/Central/South African and Asian Chikungunya Virus Genotypes in Neutralization by Immune Sera Antigenic Variation of Chikungunya Virus Genotypes in Neutralization by Immune SeraChua Chong-Long 1Sam I-Ching 1*Merits Andres 2Chan Yoke-Fun 1*1 Department of Medical Microbiology, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia2 Institute of Technology, University of Tartu, Tartu, EstoniaKlimstra William B. EditorThe University of Pittsburgh, UNITED STATESThe authors have declared that no competing interests exist. Conceived and designed the experiments: CLC ICS AM YFC. Performed the experiments: CLC AM. Analyzed the data: CLC ICS AM YFC. Contributed reagents/materials/analysis tools: ICS AM YFC. Wrote the paper: CLC ICS AM YFC. * E-mail: jicsam@ummc.edu.my (ICS); chanyf@ummc.edu.my (YFC)29 8 2016 8 2016 10 8 e00049603 8 2015 8 8 2016 © 2016 Chua et al2016Chua et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Background Chikungunya virus (CHIKV) is a re-emerging mosquito-borne virus which causes epidemics of fever, severe joint pain and rash. Between 2005 and 2010, the East/Central/South African (ECSA) genotype was responsible for global explosive outbreaks across India, the Indian Ocean and Southeast Asia. From late 2013, Asian genotype CHIKV has caused outbreaks in the Americas. The characteristics of cross-antibody efficacy and epitopes are poorly understood. Methodology/Principal Findings We characterized human immune sera collected during two independent outbreaks in Malaysia of the Asian genotype in 2006 and the ECSA genotype in 2008–2010. Neutralizing capacity was analyzed against representative clinical isolates as well as viruses rescued from infectious clones of ECSA and Asian CHIKV. Using whole virus antigen and recombinant E1 and E2 envelope glycoproteins, we further investigated antibody binding sites, epitopes, and antibody titers. Both ECSA and Asian sera demonstrated stronger neutralizing capacity against the ECSA genotype, which corresponded to strong epitope-antibody interaction. ECSA serum targeted conformational epitope sites in the E1-E2 glycoprotein, and E1-E211K, E2-I2T, E2-H5N, E2-G118S and E2-S194G are key amino acids that enhance cross-neutralizing efficacy. As for Asian serum, the antibodies targeting E2 glycoprotein correlated with neutralizing efficacy, and I2T, H5N, G118S and S194G altered and improved the neutralization profile. Rabbit polyclonal antibody against the N-terminal linear neutralizing epitope from the ECSA sequence has reduced binding capacity and neutralization efficacy against Asian CHIKV. These findings imply that the choice of vaccine strain may impact cross-protection against different genotypes. Conclusion/Significance Immune serum from humans infected with CHIKV of either ECSA or Asian genotypes showed differences in binding and neutralization characteristics. These findings have implications for the continued outbreaks of co-circulating CHIKV genotypes and effective design of vaccines and diagnostic serological assays. Author Summary Chikungunya virus (CHIKV) has caused large epidemics of fever, rash, and joint pain around the world in recent years. Three different CHIKV genotypes exist. Infection with one genotype is likely to lead to immune protection (or cross-protection) against future infections with a different genotype. However, little is known about the nature of this cross-protection. In this study, we used serum from Malaysian patients infected with CHIKV of either Asian or East/Central/South African (ECSA) genotypes. We compared the ability of the serum antibodies to bind to and neutralize two different viruses, from either Asian or ECSA genotypes. We found that both Asian and ECSA serum were more effective in binding and neutralizing ECSA virus. We identified the key amino acids/epitopes within the E1-E2 surface glycoprotein, and showed that variation of these impacts the efficacy of antiserum in cross-neutralizing different genotypes of CHIKV. We showed how sequence variation of a known linear neutralizing epitope could alter the cross-neutralization efficacy. This study aids understanding of the importance of different circulating genotypes within a country and has implications for the design of vaccines and diagnostic antibody tests. This study was funded by grants from University of Malaya (University Malaya Research Grant RG526-13HTM, High Impact Research grant E000013-20001 and Postgraduate Research Fund PG114-2012B), Fundamental Research Grant Scheme FP035-2015A of the Ministry of Higher Education, Malaysia and the European Union Seventh Framework Program (Integrated Chikungunya Research, grant agreement no. 261202). CLC was supported by the MyBrain15 (MyPhD) scheme of the Ministry of Higher Education, Malaysia and the James S. Porterfield Prize in International Virology. AM was supported by grant 20-27 from Estonian Research Council. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Chikungunya virus (CHIKV) is a re-emerging, mosquito-borne arbovirus which has caused unprecedented worldwide epidemics in recent years [1]. There are three major CHIKV genotypes circulating: West African, East/ Central/ South African (ECSA) and Asian [2]. After the global outbreaks of ECSA between 2005 and 2010, the Asian genotype has re-emerged to cause large outbreaks in the Americas and the Pacific islands [3, 4]. Malaysia has experienced CHIKV outbreaks due to two different genotypes, Asian and ECSA. The endemic Asian CHIKV strain was responsible for small, geographically-restricted outbreaks in 1998 and 2006 [5–7]. An imported ECSA outbreak was reported in 2006 prior to an explosive nationwide outbreak which affected over 15,000 people across different states in 2008 [8, 9]. CHIKV is an alphavirus from the family Togaviridae. A CHIKV virion is 60-70nm in diameter, with a single-stranded positive RNA genome of approximately 11.8 kb in a capsid with a phospholipid envelope carrying glycoproteins E1 and E2. Its genome has 2 open reading frames encoding the non-structural (nsP1-nsP2-nsP3-nsP4) and structural polyproteins (C-E3-E2-6K-E1) [10]. The E1 and E2 glycoproteins form heterodimers which enable interaction with cellular receptors and fusion of the virion envelope with the cell membrane to initiate infection [11], while the capsid protein is required during virus assembly [12]. These proteins are highly immunogenic, and most CHIKV-infected patients develop antibodies targeting the structural proteins (particularly E2) and, to a lesser extent, nsP3 [13, 14]. After the initial induction of type I interferon [15], CHIKV-specific antibodies have been shown as the major effector in immunity to control infection [16]. Among other immune factors, T cells may play a secondary role in suppressing infection [17], although others have found that CD4+ T cells are more important in orchestrating joint inflammation [18]. Currently, treatment for CHIKV is supportive and no licensed vaccine or antiviral are available. Phase I clinical trials have demonstrated the safety and efficacy of vaccination with virus-like particles using structural proteins derived from the West African genotype [19], and a recombinant measles virus-based CHIKV vaccine derived from the ECSA genotype [20]. Cross-reactivity can be achieved against heterogenous genotypes, by which CHIKV seropositive individuals infected with either ECSA or Asian CHIKV have cross-protection against both CHIKV genotypes [9]. However, the cross-neutralizing efficacy of CHIKV-specific antibodies against Asian and ECSA genotypes, which are both currently circulating in Malaysia, Brazil [21] and the Asian region [22], is poorly understood. A distinct antigenic relationship has been established between West African and ECSA genotypes, in which mice and hamsters immunized with the ECSA genotype had 4- to 8-fold differences in neutralizing capacity when tested against a West African strain [2]. In a Singaporean cohort, CHIKV-immune sera exhibited differential antibody binding and neutralizing capacity against isolates with a naturally occurring K252Q amino acid change in the E2 glycoprotein [14]. Given the ability of CHIKV to rapidly spread across different parts of the world with displacement of one genotype with another, the understanding of cross-neutralizing antibody and antigenic variation of different genotypes will have implications for both continued outbreaks and vaccine development. In this study, we analyzed the neutralizing capacity of CHIKV ECSA and Asian immune sera against representative clinical isolates and rescued viruses of ECSA and Asian CHIKV. We demonstrated that both sets of serum panels have stronger neutralizing capacity against the ECSA isolate, which corresponded to strong epitope-antibody interaction. E1-E211K enhances the neutralization activity of ECSA serum, while E2-I2T, H5N, G118S and S194G within linear epitopes improve the neutralization activity of both sets of sera panels. Rabbit polyclonal antibody targeting a known linear neutralizing epitope (LP1) from ECSA virus could only neutralize homotypic virus, but not heterotypic Asian virus due to sequence variation. These findings indicate the antigenic variation of ECSA or Asian CHIKV genotypes in naturally-acquired infection alters the spectrum of cross-genotype protective antibody immunity. Materials and Methods CHIKV immune serum panels This study included 63 human samples from two independent outbreaks in Malaysia. The Asian serum panel comprised 40 samples collected from patients 11–14 months after an Asian CHIKV outbreak in Bagan Panchor in 2006 [7]. The ECSA serum panel consisted of 23 samples from patients infected by ECSA strains in 2008–2010, collected 1–6 months after onset of symptoms, who were seen at the University Malaya Medical Centre in Kuala Lumpur [9]. Healthy controls (n = 15) with no past infection of CHIKV served as negative controls. Serum neutralization assay was performed on all the sera. To determine the neutralizing activity due to IgG, heat-inactivated sera were treated for 1 hour with dithiothreitol (DTT) (Life Technologies) at a final concentration of 5mM at 37°C. Ethics statement This study was approved by the Medical Ethics Committee of the University Malaya Medical Centre (reference no. 800.70). Our institution does not require informed consent for retrospective studies of archived and anonymized samples. Cells and viruses Baby hamster kidney (BHK-21) cells (ATCC no. CCL-10) were maintained in Glasgow minimum essential medium (GMEM) (Life Technologies) supplemented with 5% heat-inactivated fetal bovine serum (Flowlab), 10% tryptose phosphate broth, 20mM HEPES, 5mM L-glutamine, 100 U/ml penicillin and 100μg/ml streptomycin. Infected cells were maintained in GMEM containing 2% FBS. The clinical isolates used, which have been previously characterized [23], were MY/06/37348, an Asian genotype strain isolated from a patient in Bagan Panchor in 2006 (accession number FN295483), and MY/08/065, an ECSA virus isolated from a patient in Kuala Lumpur in 2008 (accession number FN295485). Both isolates had been passaged two times in Vero cells (ATCC no. CCL-81) before propagated in BHK-21 cells. Virus passage (P3) of clinical isolates was used for subsequent work. To study the neutralizing epitopes, viruses rescued from two different infectious clones, derived from ECSA and Asian genotypes of CHIKV, were included. The plasmid vectors capable of producing infectious viruses were constructed under the control of the human cytomegalovirus immediate-early promoter. The CHIKV infectious clone derived from the ECSA genotype was based on LR2006-OPY1, isolated in Reunion Island in 2006,and has been described previously [24]. The full-length infectious cDNA (icDNA) clone from the Asian genotype was engineered by gene synthesis and assembled by the restriction enzymes approach based on the consensus sequence for strain 3462, isolated in Yap State in 2013 (accession no. KJ451623); however, the protein coding regions in the non-structural and structural proteins were changed to be identical to isolate CNR20235 from the Caribbean outbreak, which was isolated in Saint Martin Island in 2013 (http://www.european-virus-archive.com/article147.html). Both molecular clones have ZsGreen gene incorporated as reporter and duplication of the subgenomic promoter. The ECSA molecular clone was named “ICRES1”, while the Asian molecular clone was designated as “CAR”. For construction of the chimeric viruses, the ectodomain regions of envelope glycoprotein genes E1 (amino acids 1–381) and E2 (amino acids 1–341) in the ICRES1 backbone were replaced with those of Semliki Forest virus (SFV) E1 (amino acids 1–381) and E2 (amino acids 1–340) from icDNA SFV6 [25] using NEBuilder HiFi DNA Assembly Master Mix (NEB). In order to study the effects of point mutations on the neutralizing epitopes, conventional PCR-based site-directed mutagenesis was performed on the CAR construct using Q5 High-Fidelity DNA polymerase (NEB) with designed primers (S1 Table). The sequences of all the constructs were verified by control restrictions and sequence analysis. Primers and sequences for infectious clone constructions are available upon request. The viruses were rescued from icDNA by electroporation. Stocks of rescued viruses (P0) were harvested and titrated by plaque assay on BHK-21 cells. To obtain P1 stocks, confluent BHK-21 cells grown in T75-cm2 flasks were infected with P0 stocks at a multiplicity of infection (MOI) of 1 plaque forming unit/cell and maintained in 2% FBS GMEM. P1 stocks were harvested after 24 or 48 hours, titrated and used for the neutralization assay. Infectious center assay was performed on all the viruses rescued from icDNA. Details on virus rescue and related protocols are shown in S1 Text and S2 Table. Serum neutralization assay Seroneutralization was performed with a previously described immunofluorescence-based cell infection assay in BHK-21 cells [26, 27], with minor modifications. The DTT-treated sera underwent 2-fold serial dilutions (1:100 to 1:6400) in 1X Dulbecco’s PBS prior to mixing with CHIKV pre-diluted with 2% FBS GMEM. Cells were infected with clinical isolates at an MOI of 10. The virus-antibody mixture was incubated for 2 hours at 37°C before inoculation into 104 cells in 96-well CellCarrier-96 optic black plates (Perkin Elmer), and further incubated for 1.5 hours at 37°C. The inocula were decanted and 2% FBS GMEM was added to the plates. The plates were fixed with 4% paraformaldehyde after 6 hours of incubation at 37°C, permeabilized with 0.25% Triton X-100 for 10 minutes, and immunostained using anti-CHIKV E2 monoclonal antibody B-D2(C4) [28] at 1μg/ml followed by rabbit anti-mouse IgG-FITC (Thermo Scientific) at 1:100 dilution. Cell nuclei were counter-stained with DAPI. Fluorescence intensity was analyzed with a Cellomics High Content Screening (HCS) ArrayScan VTI (Thermo Fisher) over 9 different fields at 5X magnification. Percentage of infectivity was calculated according to the following equation: % infectivity = (mean average fluorescence intensity from serum sample/mean average fluorescence intensity from virus control) × 100. The neutralizing titer (NT50) was expressed as the serum dilution that reduced infectivity by 50% using non-linear regression fitting in GraphPad Prism 5 (GraphPad Software). For seroneutralization using rescued viruses, diluted sera were mixed with viruses pre-diluted with 2% FBS GMEM (with infection performed at an MOI of 50), followed by the steps described above. The plates were fixed after 7 hours of incubation at 37°C. The plates were only counter-stained with DAPI prior to acquisition of ZsGreen fluorescence. To investigate the cross-reactivity of CHIKV sera against another alphavirus, SFV was rescued from icDNA SFV6 as previously described [25]. Diluted sera (1:25 and 1:100 dilutions) were mixed with SFV pre-diluted with 2% FBS GMEM (with infection performed at an MOI of 10), followed by the steps described above. The plates were fixed after 6 hours of incubation at 37°C, and stained with mouse anti-alphavirus monoclonal antibody (Santa Cruz) at 1:100 dilution. To investigate the effect of sequence variation of neutralizing epitopes in ECSA and Asian genotypes, polyclonal rabbit anti-LP1 (STKDNFNVYKATRPY), anti-LP1A (SIKDHFNVYKATRPY) and anti-LP47 (NHKKWQYNSPLVPRN) were produced commercially (GenScript). LP1 is similar to E2EP3, an immunogenic peptide (from an ECSA virus) previously reported to elicit neutralizing antibodies [26]. LP1A is the corresponding variant peptide with Asian genotype sequences. The LP47 peptide sequence is conserved in both genotypes. Seroneutralization was performed with purified antibody at 25μg/ml against the rescued viruses. Whole virus antigens and recombinant proteins For indirect IgG ELISA (antibody end-point assay) and Western blot, the antigen was partially purified virus prepared by sucrose-cushion ultra-centrifugation, treated with 1% Triton X-100 in TE buffer, clarified by centrifugation, and stored in 50% glycerol at -20°C. For production of native recombinant proteins of E1 (rE1, from amino acids 1–412) and E2 (rE2, from amino acids 1–362), viral RNA was extracted from clinical isolates (Asian MY/06/37348 and ECSA MY/08/065). cDNA was synthesized using reverse-transcription, and the genes were amplified using high fidelity Platinum Taq (Invitrogen) with designed primers (S3 Table). The transmembrane regions and cytoplasmic tails of the glycoproteins were not included in the expression cassette, to ensure solubility of the recombinant proteins. The amplicons were ligated into a pIEX-5 vector (Novagen) directionally at BamH1 and Not1 restriction sites. Each plasmid construct together with a pIE1-neo vector were co-transfected into Sf9 cells (Novagen) using Cellfectin II reagent (Invitrogen) [29]. Stable clones expressing rE1 and rE2 were generated under selection with G418 sulfate at 1000μg/ml. The proteins secreted from stable clones were purified under native conditions with activated Profinity IMAC resins (Bio-Rad) or HisTrap FF (GE). The eluates were concentrated with an Amicon centrifugal unit and the buffer was exchanged with sodium phosphate buffer (50mM NaH2PO4, 300mM NaCl, pH 8.0). The proteins were stored at -20°C in 50% glycerol, except for the proteins used in the competitive protein blocking assay, which were filter-sterilized and kept at 4°C. Fusion sequences expressing for rE2 and rE1 was generated by overlapping PCR; recombinant proteins encoded by obtained sequence were linked via linker with sequence GGGS-His (8X)-GGGG (S1 Text). The fusion glycoprotein constructs were transfected into TriExSf9 cells (Novagen) by TransIT-Insect transfection reagent (Mirus Bio). Western blot The proteins were resolved with 12% SDS-PAGE under reducing and non-reducing conditions and electro-transferred onto a nitrocellulose membrane (GE). The membrane was blocked with 10% skimmed milk in 0.05% PBS-Tween 20 (PBST). The immunoreactivity of recombinant proteins was evaluated with pools of CHIKV immune sera applied at indicated dilutions in the blocking buffer. The bound antigen-antibody complex was detected by anti-human IgG-HRP (DakoCytomation) at 1:5000 dilution in 1% bovine serum albumin (BSA)-0.05% PBST. The membrane was visualized by chemiluminescence (Bio-Rad) and images were acquired with a BioSpectrum AC imaging system (UVP). Mouse anti-His tag antibody (Merck Millipore) was included as a loading control. Mouse anti-E2 monoclonal antibody (clone: B-D2(C4); EIEVHMPPDT) [28] was also included as a control. Enzyme-linked immunosorbent assay (ELISA) All incubation steps were performed at 37°C for 1 hour, using 1% BSA-0.05% PBST as diluent for serum and antibodies. The plates were washed 4 times with 0.05% PBST after each incubation step. To determine the relative level of anti-E2 antibodies, the plates were coated with 250 ng of virus antigen or 100 ng of rE2 in 0.05M carbonate-bicarbonate buffer (pH 9.6). The antigens were normalized with monoclonal antibody B-D2(C4) to determine the relative level of anti-E2 antibodies. The plate was blocked with 3% BSA in 0.05% PBST. The sera were tested at 2-fold serial dilutions from 1:512 to 1:1,048,000 or 1:640 to 1:655,000. The IgG end-point titer was determined as the reciprocal of the highest dilution that produced an optical density (OD) reading of three times greater than that of the negative control. Anti-human IgG-HRP at 1:5000 dilution was added to detect the bound antibodies. TMB substrate (KPL) was added to each well and the plates were incubated at room temperature for 5 min. The reaction was terminated by adding 1M phosphoric acid. The absorbance was measured at 450nm with 630nm as the reference wavelength using an automated ELISA reader (Biotek Instruments). The cut-off value was established as the OD obtained from healthy controls sera plus three standard deviations (SD). The relative level of anti-rE2 antibodies was calculated with the following formula: (end point titer for rE2/end point titer for whole virus antigen) × 100. Competitive protein/peptide blocking assay Soluble recombinant CHIKV proteins (15μg) were mixed with heat-inactivated immune sera diluted at 1:200, and incubated for 1 hour at 37°C. CHIKV (MY/08/065) in amounts corresponding to an MOI of 10 was mixed with the samples, which were incubated for a further 2 hours at 37°C. Synthetic peptides were obtained from GenScript (LP1, STKDNFNVYKATRPY; LP24, TDSRKISHSCTHPFH; LP38, GNVKITVNGQTVRYK); 60μg of each peptide was mixed with immune sera diluted with 1X DPBS at 1:100 and incubated for 1.5 hours at 37°C. All the synthetic peptides for the blocking assay have a purity grade greater than 95% and are soluble in high-grade water. ICRES1 (sucrose-cushion purified virus in TE buffer pre-diluted using 2% FBS GMEM) at an amount corresponding to an MOI of 1 was mixed with the samples, which were incubated for a further 2 hours at 37°C prior to infection of BHK-21 cells. The plate was replenished with plaque medium (2% FBS GMEM containing 0.8% of carboxymethylcellulose), fixed with 4% paraformaldehyde after 15 hours of incubation, and this was followed by ZsGreen fluorescence acquisition. Infectivity corresponded to the fluorescence intensity acquired with a Cellomics HCS reader. The effect on infectivity of antibodies in the presence and absence of blocking peptides was compared. Peptide-based ELISA and epitopes analysis Biotinylated synthetic peptides covering the E2 glycoprotein sequence from amino acids 1–362 from a previous study [28] were used to screen CHIKV immune sera for binding to linear epitopes. The length of each peptide is 15-mer with a 10-mer overlap based on the CHIKV MY/08/065 sequence (accession no. FN295485; S4 Table). Similar steps were performed as described above except that the plates were washed 6 times after incubation with human sera and secondary antibody. The plates were coated with 20μg/ml streptavidin (NEB) and blocked with 5% BSA-PBST. The dissolved peptides in dimethyl sulphoxide were further diluted to a working concentration of approximately 150μg/ml in 1% BSA-PBST. CHIKV immune sera and healthy control sera were diluted at 1:1000, and screened against peptides in duplicate. The peptides with the highest OD reading from 2 adjacent overlapping synthetic peptides were considered as identified B-cell epitopes. Computational analysis and epitope localization were performed on structural data retrieved from Protein Data Bank (PDB, ID 3J2W) with UCSF CHIMERA software [30]. As the LP1 sequence is unresolved in structural data, the structure of the E2 glycoprotein was predicted using the online I-TASSER server [31, 32]. The electrostatic potential of the E2 structure (amino acid 1–362) was evaluated with PDB2PQR and APBS [33–35]. Statistical analysis Data are presented as means ± SD or means ± standard error of the mean (SEM). Differences between groups and controls were analyzed using appropriate statistical tests. A P-value of <0.05 was considered significant. Statistical analyses were performed with GraphPad Prism 5. Results Both ECSA and Asian sera have greater neutralizing capacity and binding with the ECSA genotype We employed a sensitive seroneutralization assay to compare the neutralizing capacity of the different sera panels against MY/08/065 (ECSA) and MY/06/37348 (Asian) (S1 and S2 Figs). The heat-inactivated intact sera and DTT-treated sera had similar neutralizing capacity against MY/08/065 for both sera panels (S3 Fig). ECSA sera demonstrated strong neutralizing capacity against homotypic CHIKV compared to heterotypic CHIKV (Fig 1A), with a NT50 against MY/08/065 that was a median 2.67 (range, 1.40–4.61) times greater than the NT50 against MY/06/37348 (Fig 1B). Unexpectedly, Asian sera demonstrated better neutralizing capacity against heterotypic ECSA CHIKV compared to homotypic CHIKV (Fig 1A), with a NT50 against MY/08/065 of a median 1.44 (range, 0.70–3.19) times greater than the NT50 against MY/06/37348 (Fig 1B). The greater neutralizing capacity corresponded to stronger antibody binding to MY/08/065 compared to MY/06/37348 by quantitative ELISA (Fig 1C). Seroneutralization was performed against rescued virus from icDNA of ECSA and Asian genotypes. Both ECSA and Asian sera demonstrated better neutralizing capacity against ICRES1 (ECSA) compared to CAR (Asian) (Fig 1D). Immunoblotting showed stronger reactivity of serum with the whole viral antigen (with a band of about 50kDa, consistent with E1 or E2, a known immunodominant antigen in alphaviruses) and recombinant E2 glycoprotein of similar size derived from ECSA, compared to the Asian genotype. Under non-reducing conditions, ECSA sera had stronger antibody binding to its homotypic CHIKV isolate MY/08/065, ICRES1 and recombinant E2 glycoprotein (rE2) (Fig 1E and 1F). Asian sera bound similarly to both genotypes of viruses (clinical isolates and rescued viruses), and more strongly to rE2 glycoprotein of MY/08/065 (Fig 1E and 1F). Under reducing conditions, both sets of sera retained stronger binding to ECSA CHIKV and rE2 of ECSA CHIKV. Both sets of sera had a similar proportion of total antibodies binding to rE2 (median 50%, range 20–63% for ECSA serum; median 50%, range 16–63% for Asian serum) (Fig 1G), and these percentages suggest that antibodies also target sites other than E2. Taken together, CHIKV serum shows strong neutralizing capacity and binding to CHIKV, particularly of the ECSA genotype, and the epitopes may be presented as part of the conformational E1-E2 glycoprotein and/or as linear determinants in the E2 glycoprotein. 10.1371/journal.pntd.0004960.g001Fig 1 Differential neutralization capacity and antibody binding properties of immune sera against ECSA and Asian CHIKV. (A) Sera collected from ECSA and Asian CHIKV outbreaks have differential neutralizing capacity against MY/08/065 (ECSA) and MY/06/37348 (Asian) isolates of CHIKV. Results are expressed as a percentage of virus control. *P<0.05, **P<0.01, ***P<0.001, two-way ANOVA with the Bonferroni multiple comparisons test. Data are presented as means ± SEM from 23 (ECSA) and 40 (Asian) individual serum samples. (B) Neutralization titers (NT50) of DTT-treated sera were determined by non-linear regression fitting. ***P<0.001, Wilcoxon matched-pairs signed rank test. (C) ECSA and Asian sera were cross-screened against both CHIKV isolates (105 pfu, treated with 1% Triton X-100) in ELISA at different serum dilutions. Data are presented as means ± SEM from 23 (ECSA) and 40 (Asian) individual serum samples. **P<0.01, ***P<0.001, two-way ANOVA with the Bonferroni multiple comparisons test. (D) Seroneutralization was performed against different strains of CHIKV, ICRES1 (ECSA) and CAR (Asian), which were rescued from icDNA CHIKV, at a serum dilution of 1:800. **P<0.01, ***P<0.0001, Wilcoxon matched-pairs signed rank test. (E) Immunoblotting was performed under non-reducing and reducing conditions against rE2 and CHIKV from MY/08/065 and MY/06/37348. Mouse anti-His was used as a control and pooled sera were diluted at 1:1000. (F) Immunoblotting was performed under non-reducing and reducing conditions against ICRES1 and CAR. Mouse anti-E2 was used as a control and pooled sera were diluted at 1:1000. (G) Antibody titers of CHIKV immune sera (IgG) were quantified by end-point titer ELISA using whole virus antigen and recombinant E2 (rE2) derived from MY/08/065. Middle line, median; plus sign, mean; upper and lower boundaries of the box, inter-quartile range; whiskers, range of values. ECSA and Asian sera target epitopes on the E1-E2 glycoprotein To determine if CHIKV immune serum targets E1, recombinant E1 glycoprotein (rE1) was probed in ELISA with serially diluted pooled sera, and signal was detected at low serum dilutions from 1: 100 to 1:800 (Fig 2A). A competitive protein blocking assay was performed, and blocking of ECSA and Asian sera with native rE1 alone did not significantly alter the neutralizing capacity (Fig 2B). However, when the sera was blocked by a mixture of rE1 and rE2, significant increases of infectivity were observed in both panels of sera compared to unblocked sera or sera blocked by either rE1 or rE2 alone. We then hypothesized that antibodies may target conformational epitopes on E1 and E2 glycoproteins together. To test this hypothesis, we constructed 2 chimeras which swapped the ecto-domain regions of the E2 and E1-E2 glycoproteins with those of Semliki Forest virus (SFV). Both sets of sera demonstrated a low degree of cross-neutralization against SFV, another alphavirus which is a member of the same antigenic complex as CHIKV (S4 Fig). At 1:100 serum dilution, loss of neutralizing effect for both sets of sera was observed when CHIKV E2 was replaced with SFV E2. Furthermore, in ECSA serum, loss of neutralization activity was much higher against the chimera with E1-E2 from SFV compared to the chimera with SFV E2 alone (Fig 2C). This provides further evidence that neutralizing antibodies are not solely targeting E2, but are also targeting epitopes spanning E1-E2 glycoproteins. Alternatively, E1 may affect the conformation of E2 and alter its epitopes. 10.1371/journal.pntd.0004960.g002Fig 2 Neutralizing antibodies of immune sera interact with the epitopes on E2 and E1-E2 glycoproteins. (A) CHIKV antibody titer against recombinant E1 glycoprotein (100 ng) was determined in ELISA. The ELISA was performed at different serum dilutions using pooled sera. The dotted line represents the cut-off value (mean + 3SD) derived from healthy controls. (B) Competitive blocking assay was performed at 1:200 dilution in triplicate using 7 pools of ECSA and Asian sera, with similar neutralizing titers in each pool. Data are expressed as percentages of infectivity of an infection control, and are presented as means ± SEM. ***P< 0.001, repeated measures ANOVA with the Bonferroni multiple comparison test. (C) Schematic diagram showing the construction of chimera viruses with replacement of E2 or E1/E2 from SFV into the CHIKV ICRES1 backbone. Seroneutralization was performed against the chimera constructs and the percentage of infectivity was compared to that obtained with ICRES1. Data are represented as means ± SD from 4 independent experiments at a serum dilution of 1:100 (pooled sera). ***P<0.01, Kruskal-Wallis test. G, genomic promoter; SG, subgenomic promoter. The E1-E211K amino acid change enhances neutralization activity of ECSA serum To further determine the importance of conformational epitopes resulting from interactions between E1 and E2, four sets of fusion E1-E2 glycoproteins were constructed. Each hybrid fusion protein contained E1 and E2 sequences from either MY/06/37348 (ECSA) or MY/08/065 (Asian), transiently expressed as secreted native recombinant proteins in insect cells (Fig 3A). The antibody binding capacity of ECSA sera against fusion E1-E2 proteins significantly increased when either the E1 or E2 sequence was changed from that of MY/06/37348 to that of MY/08/065, as shown in immunoblotting (Fig 3B) and quantitative ELISA (Fig 3C). Asian sera had almost equal antibody binding capacity for the 4 fusion glycoproteins, suggesting that Asian serum was not sensitive to sequence changes in E1-E2 glycoproteins. This data shows that the greater binding and neutralization of the ECSA isolate MY/08/065 by ECSA sera (Fig 1A, 1C and 1D) is due to critical conformational epitopes on the E1-E2 heterodimer, which are sequence-dependent. 10.1371/journal.pntd.0004960.g003Fig 3 The E1-E211K amino acid change enhances neutralization activity of ECSA serum. (A) Schematic diagram showing the generation of fusion recombinant E2 (amino acids 1–362) and recombinant E1 (amino acids 1–412) with a 16 residue linker which has glycine/ serine spacers and octa-histidine sequence. The rE2 and rE1 in each fusion protein are from either MY/08/065 (ECSA) or MY/06/37348 (Asian) virus isolates. (B) Immunoblotting was performed under non-reducing condition. Mouse anti-E2 and mouse anti-His monoclonal antibodies were used as controls. (C) The relative binding capacity of ECSA and Asian sera (1:1000 dilution) with the fusion E2-E1 proteins were determined in ELISA as (OD samples/mean OD samples tested with rE2-E1-ECSA) × 100. Data are presented as means ± SD (n = 4). *P<0.05, Mann-Whitney U test. (D) Schematic representation of the E1 glycoprotein with the numbers indicating the amino acid positions of the glycoprotein and its domain proteins. The amino acid differences between E1 glycoprotein of ECSA (MY/08/065, ICRES1) and Asian (MY/06/37348, CAR) were tabulated and mapped (from amino acids 1–412). Amino acid differences within a genotype are underlined. FL, fusion loop; C. tail, cytoplasmic tail. (E) Immunoblotting was performed against fusion E2-E1 glycoprotein under non-reducing conditions, with each named amino acid change from the Asian to the ECSA sequence introduced independently. Mouse anti-His antibody was used as a control. (F) Seroneutralization was performed against different constructs carrying indicated mutations in the CAR-2SG-ZsGreen backbone, which were rescued from the corresponding icDNA clones of CHIKV. Data are represented as means ± SD from 4 independent experiments at a serum dilution of 1:800 (pooled sera). *P<0.05, ***P<0.001, Mann-Whitney U test, relative to CAR. (G) Seroneutralization was performed against the CAR-E1-E211K rescued virus at a serum dilution of 1:800 with 23 individual serum samples. ***P<0.0001, Wilcoxon matched-pairs signed rank test. (H) The amino acid position of K211 which affects neutralization activity is localized on the structural E1-E2 heterodimer complex (based on PDB 3J2W). Between ECSA (MY/08/065 and ICRES1) and Asian (MY/06/37348 and CAR) genotypes of CHIKV in this study, there are 10 amino acids differences in E1 (Fig 3D). Using the fusion rE2-E1-Asian construct as a template, site-directed mutagenesis was performed independently to replace each amino acid of Asian origin with the corresponding ECSA residue, and the proteins were expressed in insect cells. The antibody binding significantly increased with the amino acid changes at A145T, E211K, A226V and M269V, in comparison to hybrid rE2Asian-E1ECSA recombinant proteins (Fig 3E). Recombinant virus carrying E1-211K demonstrated a large increase in neutralizing capacity compared to the parental virus clone (CAR), while the E1-145T change caused a slight decrease in neutralizing capacity (Fig 3F and 3G). The critical 211K amino acid was localized at the surface of E1-E2 heterodimers (Fig 3H). I2T, H5N, G118S and S194G substitutions within linear neutralizing epitopes of E2 glycoprotein enhance the neutralization activity of ECSA and Asian sera To study the linear epitopes in the immunodominant E2 glycoprotein (based on strain MY/08/065, of the ECSA genotype), overlapping synthetic peptides covering amino acids 1–362 were mapped by peptide-ELISA using the ECSA and Asian sera (Fig 4A, 4B and 4C). Both ECSA and Asian sera mapped to the same 9 peptides, and the Asian sera mapped to an additional 3 peptides (Table 1). Between the strains of ECSA (MY/08/065, ICRES1) and Asian (MY/06/37348, CAR) genotypes of CHIKV used in this study, there are 15 amino acid differences in E2 (from amino acids 1–362), of which 4 amino acid differences fall within the identified linear epitopes (Fig 4D). Using the rE2-Asian construct as a backbone, site-directed mutagenesis was performed to replace each amino acid of Asian origin with an ECSA residue, and the proteins were expressed in insect cells. The antibody binding significantly increased with I2T, H5N, G118S, R149K and S194G substitutions in comparison to the original rE2-Asian recombinant protein (Fig 4E and S5 Fig). Recombinant viruses carrying either E2-2T, 5N, 118S or 194G demonstrated increases in neutralizing capacity compared to the parental virus clone (CAR), while the E2-R149K change caused a decrease in neutralizing capacity (Fig 4F). Competitive peptide blocking assay indicated that the anti-CHIKV antibodies interact with the LP1, LP24 and LP38 peptides that cover amino acid sites 2, 5, 118 and 194 on E2 (Fig 4G). These 4 neutralizing linear epitopes are localized on the surface of the E1-E2 heterodimer complex (Fig 4H). 10.1371/journal.pntd.0004960.g004Fig 4 E2-I2T, H5N, G118S and S194G substitutions within linear neutralizing epitopes enhance the neutralization activity of ECSA and Asian sera. (A) Overlapping synthetic peptides covering the E2 glycoprotein of MY/08/065 and its domains from amino acids 1–362 were screened with CHIKV immune sera at 1:1000 dilution. The black solid line represents the mean OD value of healthy controls and the dotted line represents the cut-off value (mean+3SD). The average results from 2 independent experiments are presented. In the event of two adjacent positively-mapping peptides, the peptide with the highest OD reading was taken. Key positive mapping peptides are color-coded. (B) Selected synthetic peptides were re-screened with pooled ECSA sera at lower dilutions of 1:500 and 1:250, in tetraplicate. The black solid line represents the mean OD value of healthy controls and the dotted line represents the cut-off value (mean+3SD). Key positive mapping peptides are color-coded. (C) Schematic diagram of the E2 protein showing the positions of the color-coded mapped epitopes. The numbers refer to the amino acid positions demarcating the E2 domains. N, N-link; C.arch, central arch. (D) Schematic representation of the E2 glycoprotein with the numbers indicating the amino acid positions of the glycoprotein and its domain proteins. The amino acid differences between E2 glycoproteins of ECSA (MY/08/065, ICRES1) and Asian (MY/06/37348, CAR) strains were tabulated and mapped (from amino acids 1–362). Amino acid differences within a genotype are underlined. Amino acid changes which fall within the identified linear epitopes are color-coded. (E) Immunoblotting was performed against recombinant E2 glycoproteins under reducing conditions, with each named amino acid change from the Asian to the ECSA sequence introduced independently. Mouse anti-His antibody was used as a control. (F) Seroneutralization was performed against different constructs with the CAR-2SG-ZsGreen backbone, which were rescued from the corresponding CHIKV icDNAs. Data are represented as means ± SD from 4 independent experiments at a serum dilution of 1:800 (pooled sera). *P<0.05, **P<0.01, ***P<0.001, Mann-Whitney U test. (G) Competitive peptide blocking assay was performed at 1:100 dilution with either pooled ECSA or Asian sera against ICRES1 at an MOI of 1. Data are expressed as percentages of infectivity of an infection control, and are presented as means ± SD from 2 independent experiments. *P<0.05, **P< 0.01, Mann-Whitney U test, relative to unblocked condition. (H) The color-coded mapped neutralizing epitopes are localized on the structural E1-E2 heterodimer complex (based on PDB 3J2W). The epitope sequence of LP1 is only partially localized as the 3D structure is not fully resolved. 10.1371/journal.pntd.0004960.t001Table 1 Sequences of identified B cell epitopes on the E2 glycoprotein of MY/08/065. Domain binding site B cell epitope sequence a Amino acid positions b Peptide annotation N-link STKDNFNVYKATRPY 1–15 LP1 A ATDGTLKIQVSLQIG 41–55 LP9 CTITGTMGHFILARC 91–105 LP19 TDSRKISHSCTHPFH 116–130 LP24 β-ribbon (Arch1) IGREKFHSRPQHGKE 136–150 LP28 B GNVKITVNGQTVRYK 186–200 LP38 VINNCKVDQCHAAVT 216–230 LP44 β-ribbon (Arch 2) NHKKWQYNSPLVPRN 231–245 LP47 HIPFPLANVTCRVPK 256–270 LP52 C VTYGKNQVIMLLYPD 276–290 LP56 LEVTWGNNEPYKYWP 326–340 LP66 Stem GTAHGHPHEIILYYY 346–360 LP70 a Underlined sequences indicate common epitopes recognised by both ECSA and Asian sera. b The first amino acid in E2 is numbered as 1. Sequence variation of a neutralizing linear epitope influences cross-genotype neutralization As naturally-acquired infection of the Asian genotype of CHIKV leads to higher cross-neutralizing efficacy against ECSA CHIKV, we hypothesized that an epitope-based vaccine derived from the Asian genotype might provide a substantial level of cross-protection against ECSA CHIKV. The peptide LP1 (STKDNFNVYKATRPY) is similar to E2EP3, a peptide derived from ECSA virus which has been found to be highly immunogenic in eliciting neutralizing antibodies in an animal model [26]. We generated a variant, LP1A (SIKDHFNVYKATRPY), derived from the sequence of the Asian virus. Rabbit polyclonal antibodies were commercially prepared against LP1A and LP1. Peptide-ELISA was performed using human ECSA and Asian serum with LP1A and LP1 as antigens. Human ECSA serum bound to LP1 but not LP1A (Fig 5A). Rabbit anti-LP1 antibody showed the lowest binding capacity against CAR (Asian), and demonstrated poor neutralizing activity against the CAR virus harboring the LP1A sequence (infectivity 91±10%, Fig 5B). Anti-LP1 binding capacity and neutralization efficacy was partially restored with the mutations I2T and H5N. The anti-LP1 antibody had maximum binding capacity and neutralizing efficacy against CAR-E2-I2T-H5N (Fig 5B), which has the LP1 sequence; a finding in line with the antibody binding of ECSA immune sera against LP1 peptide (Fig 5A). 10.1371/journal.pntd.0004960.g005Fig 5 Sequence variation of a linear neutralizing epitope influences the spectrum of cross-neutralization across genotypes. (A) Synthetic peptides LP1A and LP1 were screened with ECSA immune sera at 1:500 dilution in tetraplicate. The dotted line represents the cut-off value (mean + 3SD). *P<0.05, Mann-Whitney U test. (B) Recombinant viruses were pre-treated with 1% Triton X-100, then coated at 5 x 105 pfu per well, and a binding assay was performed with 1μg/ml of antibody. The seroneutralization assay was performed with infection at an MOI of 50 against 25μg/ml of antibody. Data are represented as means ± SD from 3 independent experiments. ***P<0.001, repeated measures ANOVA with the Bonferroni multiple comparisons test. (C) Synthetic peptides LP1A and LP1 were screened with Asian immune sera at 1:500 dilution in tetraplicate. The dotted line represents the cut-off value (mean + 3SD). *P<0.05, Mann-Whitney U test. (D) Seroneutralization of anti-LP1A against CAR and CAR-E2-I2T-H5N at an antibody concentration of 25 μg/ml. Data are represented as means ± SD from 3 independent experiments.(E) Competitive peptide blocking assay was performed at 1:100 dilution with either LP1A or LP1 in pooled Asian sera against ICRES1 at an MOI of 1. Data are expressed as percentages of infectivity of an infection control, and are presented as means ± SD (n = 6). **P< 0.01, Mann-Whitney U test, relative to unblocked condition. (F) Structural images illustrating the changes of surface electrostatic potential due to differences in amino acid positions 2 and 5 within LP1A (Asian) and LP1 (ECSA) sequences (red arrows). Asian serum could recognize LP1A, although binding was marginally higher to LP1 (Fig 5C), which supports the earlier finding that Asian serum has stronger binding against LP1 with I2T and H5N amino acid changes (Fig 4E). Unexpectedly, rabbit anti-LP1A did not demonstrate significant neutralizing activity against CHIKV with either the LP1A or LP1 sequences (Fig 5D). However, a competitive peptide blocking assay indicated that neutralizing antibodies from Asian sera could still recognize and interact with both LP1A and LP1 peptides (Fig 5E). The electrostatic potential of the E2 surface was computed based on the CAR ecto-domain region to study the charge distribution of these epitopes which affect binding affinity [36]. The I2T change leads to higher electrostatic potential, which is associated with improved binding capacity and neutralization efficacy (Fig 5F). LP47, another linear neutralizing epitope in humans, also failed to induce any functional neutralizing antibodies in rabbits. Discussion CHIKV has become a major public health concern worldwide and causes considerable socio-economic burden. Protective adaptive immunity is mainly provided by specific antibodies, particularly those directed against epitopes on the E2 and E1 glycoproteins [37, 38]. Understanding cross-immunity resulting from infections with different genotypes is particularly important and timely. Many Asian countries now have both endemic Asian and epidemic ECSA strains circulating, and the recent widespread outbreaks in the Americas are due to the Asian genotype rather than the previously epidemic ECSA strains, indicating that viruses from both genotypes are capable of global spread. In this study, we showed differences in cross-genotypic neutralization efficacy of immune sera against ECSA and Asian genotypes of CHIKV. Both ECSA and Asian serum had greater neutralizing capacity against ECSA genotype (MY/08/065 and ICRES1) than Asian genotype (MY/06/37348 and CAR), indicating that neutralizing antibodies regardless of initial infecting genotype preferentially recognized the epitopes presented by the ECSA genotype. The presence of cross-genotype neutralization was clearly shown lasting up to 14 months post-infection. The clinical significance of the differential cross-protective capacity of ECSA and Asian sera remains unclear, as all the immune sera had more than the minimum neutralizing titer (≥10) which appears to correlate with immune protection from symptomatic CHIKV infection in humans [39]. This high degree of cross-neutralization likely contributed to the geographic restriction of CHIKV of different genotypes seen historically, which limited, for example, the spread of ECSA viruses in Asia, at least until CHIKV underwent mutations that facilitated sequential adaptation to the Aedes albopictus vector [40, 41]. Apart from the stronger antigenicity of epitopes of the ECSA genotype, we also showed that neutralizing capacity was also affected by the target and the amount of neutralizing antibodies. Both ECSA and Asian sera contain high levels of neutralizing antibodies to numerous linear epitopes on the E2 glycoprotein as well as conformational epitopes on the E1-E2 heterodimer complex. This supports recent findings that most of the reported CHIKV neutralizing monoclonal antibodies target conformational epitopes on the exposed, topmost outer surfaces of the E2/E1 spike, particularly in domain A and domain B [42–45]. Our findings also suggest that subunit vaccine candidates derived from E1 or E2 glycoproteins alone [46–48] may be insufficient to provide full protection against all genotypes, and that virus-like particle vaccines which present epitopes on E2/E1 in their native configuration may preferentially induce the most highly protective immune response [19, 49, 50]. The loss of neutralization activity against chimeric CHIKV is in line with the finding that total IgG and anti-rE2 antibody titers correlate with the neutralizing titer of Asian serum (S6 Fig), suggesting that most of the neutralizing epitopes are on the E2 glycoprotein. The lack of correlation between anti-rE2 antibodies and neutralizing antibodies seen in ECSA serum could be due to the greater importance of conformational epitopes at E1-E2 sites, but we cannot exclude that it may reflect differences in potency/quality of the circulating antibodies due to the different timings of collection between the Asian and ECSA serum panels (S6 Fig). Correlation between serum neutralization titers and antibody binding titers has been reported in other viral infections such as dengue and influenza [51, 52], and is important for developing serological assays which are accurate correlates of protective immunity following infection or vaccination. Therefore, E2, while appropriate for serological assays to diagnose acute or past infection [53], may not be a suitable candidate for assays to measure protective immunity due to all CHIKV genotypes. Such assays are necessary for vaccine development. Amino acid changes in key epitope regions, such as naturally occurring mutations or antigenic variation between different genotypes could affect surface charge distribution and electrostatic interactions between epitopes and antibodies, affect binding affinity and ultimately alter neutralizing capacity [14]. The E211K mutation in domain II of the E1 glycoprotein is a significant change of a negatively-charged to positively-charged amino acid, and this appears to enhance antibody binding and neutralization efficacy. During the recent Indian outbreak of ECSA CHIKV, the key amino acid change E1-K211E was shown to be under positive selection pressure [54], which may confer a selective advantage for virus dissemination and escape from the action of neutralization in humans. In addition, E211 is highly conserved in strains of the Asian genotype. Peptide-specific rabbit polyclonal antibody prepared against a short linear epitope (GDIQSRTPESKDVY, position 201–214) including 211K did not show neutralization activity (S7 Fig), suggesting that the neutralizing activity of immune sera targeting this amino acid is highly conformation-dependent. As for the E2 glycoprotein, I2T, H5N, G118S and S194G changes increased antibody binding and neutralization efficacy. All these amino acid changes are positioned within linear epitopes, which interacted with neutralizing antibodies. This was supported by a previous report of well-characterized human neutralizing monoclonal antibodies targeting epitopes that cluster around the LP24 and LP38 peptide regions in our study [43]. Notably, the linear epitope LP1 in our study is similar to E2EP3, a well-characterized key neutralizing linear epitope which has been suggested as a serology marker [26, 55], and LP1 demonstrated cross-reactivity with ECSA and Asian serum in our study. However, we found no effect of K252Q in antibody binding capacity in our cohort, although this was reported recently [14], and this could be due to differential immune responses in different populations. Other linear epitopes (LP19, LP47, LP56 and LP70) were identified in this study which had higher binding than LP1, and as all demonstrated binding to both ECSA and Asian sera, they may be potential candidates for diagnostic serological assays. Furthermore, antibodies against LP19 and LP47 demonstrated neutralizing characteristics which warrant further investigation as vaccine candidates (S8 Fig). It was interesting that the Asian serum had greater neutralizing capacity against the heterologous ECSA isolates. The previously reported human CHIKV monoclonal antibodies 5F10 and 8B10 had a broad neutralization activity against isolates of the ECSA and West African genotypes, but were also less potent against an Asian isolate from Indonesia [56]. Monkeys inoculated with a virus-like particle vaccine derived from the West African strain 37997 also developed better neutralizing activity to a heterologous ECSA strain LR2006 OPY-1 than to 37997, possibly due to better presentation of conserved epitopes by LR2006 OPY-1 [49]. ECSA and Asian CHIKV genotypes could have induced different immune mediator profiles; as shown in mice, infection with a Caribbean (Asian) strain was associated with a weaker pro-inflammatory Th1 and natural killer cell response and higher IgG1:IgG2c ratio compared to an ECSA CHIKV strain, resulting in less severe joint pathology [57, 58]. Different CHIKV viruses may also trigger differential regulation of key innate immune responses such as TLR3 [59], which plays an important role in shaping subsequent neutralizing capacity [60]. Further studies are needed to understand how differentially-induced immune mediators modulate the properties of circulating serum antibodies. Two amino acids in LP1 (2T, 5N) of the ECSA virus are critical for binding and neutralization activity, and this further highlights the fact that sequence variation could impact vaccine development. The rabbit polyclonal antibody targeting the linear neutralizing epitope LP1 from the ECSA virus showed reduced cross-neutralization against the Asian genotype, and unexpectedly, rabbit anti-LP1A poorly neutralized the homotypic CAR Asian virus, despite immunization of 4 rabbits. The linear neutralizing epitope LP1A from the Asian virus was not recognized by the ECSA sera. However, clearly there are preexisting antibodies against LP1 and LP1A in the Asian sera. LP47, another linear neutralizing epitope in humans (S8 Fig), which has a sequence that is conserved in both genotypes, did not induce any functional neutralizing antibodies in rabbits despite a similar immunization approach. Future studies will be required to address these apparent underlying differences of neutralizing antibody production from either natural infection or immunization. Nevertheless, our findings indicate that the choice of virus strain for vaccines could impact the spectrum and efficacy of protection across genotypes. For antibody therapy of CHIKV, monoclonal antibodies should retain high potency against a broad diversity of CHIKV isolates [43]. In conclusion, immune serum from humans infected with CHIKV of either ECSA or Asian genotypes showed differences in neutralization and binding capacities. Our findings are relevant to current outbreaks with co-circulating genotypes and provide insights into antibody-mediated immunity resulting from infections with CHIKV of different genotypes. Supporting Information S1 Fig Seroneutralization of ECSA sera panel against ECSA and Asian CHIKV genotypes. Representative acquired immunofluorescence microscopic images of pooled serum at dilutions of 1:100, 1:400, and 1:1600, and virus control against clinical CHIKV isolates MY/08/065 (ECSA) and MY/06/37348 (Asian). Each image contains 9 combined fields within a well (96-well format). Objective magnification: 5X. (PDF) Click here for additional data file. S2 Fig Seroneutralization of Asian sera panel against ECSA and Asian CHIKV genotypes. Representative acquired immunofluorescence microscopic images of pooled serum at dilutions of 1:100, 1:400, and 1:1600, and virus control against clinical CHIKV isolates MY/08/065 (ECSA) and MY/06/37348 (Asian). Each image contains 9 combined fields within a well (96-well format). Objective magnification: 5X. (PDF) Click here for additional data file. S3 Fig Comparison of neutralizing capacity of heat-inactivated intact sera and DTT-treated sera. DTT-treated sera (containing IgG only) have similar neutralizing capacity to intact sera, which have a mixture of IgG and IgM. All sera were assayed up to 1:6400 dilution. The neutralization data was based on experiments performed against ECSA CHIKV (strain MY/08/065). Data are presented as means ± SEM; n = 23 for ECSA sera, n = 40 for Asian sera. (PDF) Click here for additional data file. S4 Fig Seroneutralization of CHIKV immune individuals against Semliki Forest virus (SFV). (A) Seroneutralization was performed against SFV at 1:25 and 1:100 serum dilutions using pooled sera. Data are expressed as percentages of infectivity over infection control, and are presented as means ± SD from 3 independent experiments. **P< 0.01, ***P<0.001, Mann-Whitney U test relative to virus control. (B) Representative acquired immunofluorescence microscopic images of pooled serum at dilutions of 1:25 or 1:100 and virus control against SFV rescued from icDNA SFV6. Objective magnification: 10X. (PDF) Click here for additional data file. S5 Fig Identification of amino acids on E2 which increased the antibody binding capacity. (A) Schematic representation of the E2 glycoprotein with the numbers indicating the amino acid positions of the glycoprotein and its domain proteins. The amino acid differences between E2 glycoproteins of ECSA (MY/08/065, ICRES1) and Asian (MY/06/37348, CAR) strains were tabulated and mapped (from amino acids 1–362). Amino acid differences within a genotype are underlined. Amino acid changes which fall within the identified linear epitopes are color-coded. (B) Immunoblotting was performed against recombinant E2 glycoproteins under reducing conditions, with each named amino acid change from the Asian to the ECSA sequence introduced independently. Mouse anti-His was used as a control. Site-directed mutagenesis was not performed for amino acid positions 312, 317 and 318 as these are predicted not to be exposed on the protein surface. (PDF) Click here for additional data file. S6 Fig High titer of CHIKV-specific antibodies in Asian sera correlated with high antibody protection. The relationships between NT50 and antibody titers against (A) MY/08/065 and (B) recombinant E2 glycoprotein were assessed. Spearman’s rank correlation coefficients (ρ) and P-values are shown. ns, not significant. (PDF) Click here for additional data file. S7 Fig Neutralization of CHIKV with peptide-specific rabbit polyclonal antibodies targeting the E1 glycoprotein. Two antibodies were prepared commercially; anti-E1DII, which targets a linear epitope of E1 (GDIQSRTPESKDVY, position 201–214), and anti-LP1, which targets a linear epitope of E2 (STKDNFNVYKATRPY, position 1–15). Seroneutralization was performed at 25μg/ml against ICRES1. Data are presented as means ± SD from 2 independent experiments, run in triplicate. ** P<0.01, Mann-Whitney U test relative to virus control. (PDF) Click here for additional data file. S8 Fig Functional characterization of high linear epitope responders on the E2 glycoprotein. (A) Schematic diagram of the E2 protein showing the positions of the four mapped epitopes (LP19, yellow; LP47, pink; LP56, orange; LP70, sandy brown) which have higher OD relative to LP1. The numbers refer to the amino acid positions demarcating the E2 domains. N, N-link; C.arch, central arch. (B) Competitive peptide blocking assay was performed at 1:100 dilution with either pooled ECSA or Asian sera against ICRES1 at an MOI of 1. Sera blocked by LP19v and LP47 resulted in increases in infectivity. LP19v is a soluble peptide without cysteine residues at N- and C-terminuses of LP19. Data are expressed as percentages of infectivity of an infection control, and are presented as means ± SD from 2 independent experiments, run in triplicate. **P< 0.01, Mann-Whitney U test, relative to unblocked control. (C) The color-coded mapped neutralizing epitopes (LP19 and LP47) are localized on the structural E1-E2 heterodimer complex (based on PDB 3J2W). (PDF) Click here for additional data file. S1 Table Primers used for mutagenesis of CHIKV E1 and E2 proteins and CHIKV infectious clones. (DOCX) Click here for additional data file. S2 Table Virus rescues after electroporation of DNA-launched icDNA CHIKV. (DOCX) Click here for additional data file. S3 Table Primers used for construction of expression cassettes. (DOCX) Click here for additional data file. S4 Table The fifty-nine overlapping peptides used for the peptide-based ELISA cover the CHIKV E2 glycoprotein sequence from amino acids 1 to 362, based on the CHIKV MY/08/065 sequence (accession no. FN295485). (DOCX) Click here for additional data file. S1 Text Supplementary materials and methods. (DOCX) Click here for additional data file. We would like to acknowledge Age Utt from Institute of Technology, University of Tartu, Estonia for her excellent technical guidance. ==== Refs References 1 Sam IC , Kummerer BM , Chan YF , Roques P , Drosten C , AbuBakar S . Updates on chikungunya epidemiology, clinical disease, and diagnostics . Vector Borne Zoonotic Dis . 2015 ;15 : 223 –230 . 10.1089/vbz.2014.1680 .25897809 2 Powers AM , Brault AC , Tesh RB , Weaver SC . Re-emergence of Chikungunya and O'nyong-nyong viruses: evidence for distinct geographical lineages and distant evolutionary relationships . J Gen Virol . 2000 ;81 : 471 –479 . .10644846 3 Pan American Health Organization. Available: http://www.paho.org/hq/index.php?option=com_topics&view=article&id=343&Itemid=40931&lang=en. Accessed 19 April 2016. 4 Leparc-Goffart I , Nougairede A , Cassadou S , Prat C , de Lamballerie X . Chikungunya in the Americas . Lancet . 2014 ;383 : 514 10.1016/S0140-6736(14)60185-9 .24506907 5 Lam SK , Chua KB , Hooi PS , Rahimah MA , Kumari S , Tharmaratnam M , et al Chikungunya infection—an emerging disease in Malaysia . Southeast Asian J Trop Med Public Health . 2001 ;32 : 447 –451 . .11944696 6 AbuBakar S , Sam IC , Wong PF , MatRahim N , Hooi PS , Roslan N . Reemergence of endemic Chikungunya, Malaysia . Emerg Infect Dis . 2007 ;13 : 147 –149 . 10.3201/eid1301.060617 .17370532 7 Ayu SM , Lai LR , Chan YF , Hatim A , Hairi NN , Ayob A , et al Seroprevalence survey of Chikungunya virus in Bagan Panchor, Malaysia . Am J Trop Med Hyg . 2010 ;83 : 1245 –1248 . 10.4269/ajtmh.2010.10–0279 .21118929 8 Noridah O , Paranthaman V , Nayar SK , Masliza M , Ranjit K , Norizah I , et al Outbreak of chikungunya due to virus of Central/East African genotype in Malaysia . Med J Malaysia . 2007 ;62 : 323 –328 . .18551938 9 Sam IC , Chan YF , Chan SY , Loong SK , Chin HK , Hooi PS , et al Chikungunya virus of Asian and Central/East African genotypes in Malaysia . J Clin Virol . 2009 ;46 : 180 –183 . 10.1016/j.jcv.2009.07.016 .19683467 10 Khan AH , Morita K , Parquet Md Mdel C , Hasebe F , Mathenge EG , Igarashi A . Complete nucleotide sequence of chikungunya virus and evidence for an internal polyadenylation site . J Gen Virol . 2002 ;83 : 3075 –3084 . .12466484 11 Strauss JH , Strauss EG . The alphaviruses: gene expression, replication, and evolution . Microbiol Rev . 1994 ;58 : 491 –562 . .7968923 12 Solignat M , Gay B , Higgs S , Briant L , Devaux C . Replication cycle of chikungunya: a re-emerging arbovirus . Virology . 2009 ;393 : 183 –197 . 10.1016/j.virol.2009.07.024 .19732931 13 Sourisseau M , Schilte C , Casartelli N , Trouillet C , Guivel-Benhassine F , Rudnicka D , et al Characterization of reemerging chikungunya virus . PLoS Pathog . 2007 ;3 : e89 10.1371/journal.ppat.0030089 .17604450 14 Kam YW , Lee WW , Simarmata D , Harjanto S , Teng TS , Tolou H , et al Longitudinal analysis of the human antibody response to Chikungunya virus infection: implications for serodiagnosis and vaccine development . J Virol . 2012 ;86 : 13005 –13015 . 10.1128/JVI.01780-12 .23015702 15 Couderc T , Chretien F , Schilte C , Disson O , Brigitte M , Guivel-Benhassine F , et al A mouse model for Chikungunya: young age and inefficient type-I interferon signaling are risk factors for severe disease . PLoS Pathog . 2008 ;4 : e29 10.1371/journal.ppat.0040029 .18282093 16 Couderc T , Khandoudi N , Grandadam M , Visse C , Gangneux N , Bagot S , et al Prophylaxis and therapy for Chikungunya virus infection . J Infect Dis . 2009 ;200 : 516 –523 . 10.1086/600381 .19572805 17 Poo YS , Rudd PA , Gardner J , Wilson JA , Larcher T , Colle MA , et al Multiple immune factors are involved in controlling acute and chronic chikungunya virus infection . PLoS Negl Trop Dis . 2014 ;8 : e3354 10.1371/journal.pntd.0003354 .25474568 18 Teo TH , Lum FM , Claser C , Lulla V , Lulla A , Merits A , et al A pathogenic role for CD4+ T cells during Chikungunya virus infection in mice . J Immunol . 2013 ;190 : 259 –269 . 10.4049/jimmunol.1202177 .23209328 19 Chang LJ , Dowd KA , Mendoza FH , Saunders JG , Sitar S , Plummer SH , et al Safety and tolerability of chikungunya virus-like particle vaccine in healthy adults: a phase 1 dose-escalation trial . Lancet . 2014 ;384 : 2046 –2052 . 10.1016/S0140-6736(14)61185-5 .25132507 20 Ramsauer K , Schwameis M , Firbas C , Mullner M , Putnak RJ , Thomas SJ , et al Immunogenicity, safety, and tolerability of a recombinant measles-virus-based chikungunya vaccine: a randomised, double-blind, placebo-controlled, active-comparator, first-in-man trial . Lancet Infect Dis . 2015 ;15 : 519 –527 . 10.1016/S1473-3099(15)70043-5 .25739878 21 Nunes MR , Faria NR , de Vasconcelos JM , Golding N , Kraemer MU , de Oliveira LF , et al Emergence and potential for spread of Chikungunya virus in Brazil . BMC Med . 2015 ;13 : 102 10.1186/s12916-015-0348-x .25976325 22 Weaver SC , Lecuit M . Chikungunya virus and the global spread of a mosquito-borne disease . N Engl J Med . 2015 ;372 : 1231 –1239 . 10.1056/NEJMra1406035 .25806915 23 Sam IC , Loong SK , Michael JC , Chua CL , Wan Sulaiman WY , Vythilingam I , et al Genotypic and phenotypic characterization of Chikungunya virus of different genotypes from Malaysia . PLoS One . 2012 ;7 : e50476 10.1371/journal.pone.0050476 .23209750 24 Hallengard D , Kakoulidou M , Lulla A , Kummerer BM , Johansson DX , Mutso M , et al Novel attenuated Chikungunya vaccine candidates elicit protective immunity in C57BL/6 mice . J Virol . 2014 ;88 : 2858 –2866 . 10.1128/JVI.03453-13 .24371047 25 Ferguson MC , Saul S , Fragkoudis R , Weisheit S , Cox J , Patabendige A , et al Ability of the encephalitic arbovirus Semliki Forest virus to cross the blood-brain barrier is determined by the charge of the E2 glycoprotein . J Virol . 2015 ;89 : 7536 –7549 . 10.1128/JVI.03645-14 .25972559 26 Kam YW , Lum FM , Teo TH , Lee WW , Simarmata D , Harjanto S , et al Early neutralizing IgG response to Chikungunya virus in infected patients targets a dominant linear epitope on the E2 glycoprotein . EMBO Mol Med . 2012 ;4 : 330 –343 . 10.1002/emmm.201200213 .22389221 27 Sam IC , Chua CL , Rovie-Ryan JJ , Fu JY , Tong C , Sitam FT , et al Chikungunya virus in macaques, Malaysia . Emerg Infect Dis . 2015 ;21 : 1683 –1685 . 10.3201/eid2109.150439 .26291585 28 Chua CL , Chan YF , Sam IC . Characterisation of mouse monoclonal antibodies targeting linear epitopes on Chikungunya virus E2 glycoprotein . J Virol Methods . 2014 ;195 : 126 –133 . 10.1016/j.jviromet.2013.10.015 .24134938 29 Chua CL , Sam IC , Chan YF . Expression and purification of E2 glycoprotein from insect cells (Sf9) for use in serology . Methods Mol Biol . 2016 ; 1426 : 51 –61 . 10.1007/978-1-4939-3618-2_5 .27233260 30 Pettersen EF , Goddard TD , Huang CC , Couch GS , Greenblatt DM , Meng EC , et al UCSF Chimera—a visualization system for exploratory research and analysis . J Comput Chem . 2004 ;25 : 1605 –1612 . 10.1002/jcc.20084 .15264254 31 Zhang Y . I-TASSER server for protein 3D structure prediction . BMC Bioinformatics . 2008 ;9 : 40 10.1186/1471-2105-9-40 .18215316 32 Roy A , Kucukural A , Zhang Y . I-TASSER: a unified platform for automated protein structure and function prediction . Nat Protoc . 2010 ;5 : 725 –738 . 10.1038/nprot.2010.5 .20360767 33 Dolinsky TJ , Czodrowski P , Li H , Nielsen JE , Jensen JH , Klebe G , et al PDB2PQR: expanding and upgrading automated preparation of biomolecular structures for molecular simulations . Nucleic Acids Res . 2007 ;35 : W522 –525 . 10.1093/nar/gkm276 .17488841 34 Dolinsky TJ , Nielsen JE , McCammon JA , Baker NA . PDB2PQR: an automated pipeline for the setup of Poisson-Boltzmann electrostatics calculations . Nucleic Acids Res . 2004 ;32 : W665 –667 . 10.1093/nar/gkh381 .15215472 35 Baker NA , Sept D , Joseph S , Holst MJ , McCammon JA . Electrostatics of nanosystems: application to microtubules and the ribosome . Proc Natl Acad Sci U S A . 2001 ;98 : 10037 –10041 . 10.1073/pnas.181342398 .11517324 36 McCoy AJ , Chandana Epa V , Colman PM . Electrostatic complementarity at protein/protein interfaces . J Mol Biol . 1997 ;268 : 570 –584 . 10.1006/jmbi.1997.0987 .9159491 37 Voss JE , Vaney MC , Duquerroy S , Vonrhein C , Girard-Blanc C , Crublet E , et al Glycoprotein organization of Chikungunya virus particles revealed by X-ray crystallography . Nature . 2010 ;468 : 709 –712 . 10.1038/nature09555 .21124458 38 Mallilankaraman K , Shedlock DJ , Bao H , Kawalekar OU , Fagone P , Ramanathan AA , et al A DNA vaccine against chikungunya virus is protective in mice and induces neutralizing antibodies in mice and nonhuman primates . PLoS Negl Trop Dis . 2011 ;5 : e928 10.1371/journal.pntd.0000928 .21264351 39 Yoon IK , Alera MT , Lago CB , Tac-An IA , Villa D , Fernandez S , et al High rate of subclinical chikungunya virus infection and association of neutralizing antibody with protection in a prospective cohort in the Philippines . PLoS Negl Trop Dis . 2015 ;9 : e0003764 10.1371/journal.pntd.0003764 .25951202 40 Tsetsarkin KA , Weaver SC . Sequential adaptive mutations enhance efficient vector switching by Chikungunya virus and its epidemic emergence . PLoS Pathog . 2011 ;7 : e1002412 10.1371/journal.ppat.1002412 .22174678 41 Tsetsarkin KA , Chen R , Leal G , Forrester N , Higgs S , Huang J , et al Chikungunya virus emergence is constrained in Asia by lineage-specific adaptive landscapes . Proc Natl Acad Sci U S A . 2011 ;108 : 7872 –7877 . 10.1073/pnas.1018344108 .21518887 42 Fong RH , Banik SS , Mattia K , Barnes T , Tucker D , Liss N , et al Exposure of epitope residues on the outer face of the chikungunya virus envelope trimer determines antibody neutralizing efficacy . J Virol . 2014 ;88 : 14364 –14379 . 10.1128/JVI.01943-14 .25275138 43 Smith SA , Silva LA , Fox JM , Flyak AI , Kose N , Sapparapu G , et al Isolation and characterization of broad and ultrapotent human monoclonal antibodies with therapeutic activity against chikungunya virus . Cell Host Microbe . 2015 ;18 : 86 –95 . 10.1016/j.chom.2015.06.009 .26159721 44 Long F , Fong RH , Austin SK , Chen Z , Klose T , Fokine A , et al Cryo-EM structures elucidate neutralizing mechanisms of anti-chikungunya human monoclonal antibodies with therapeutic activity . Proc Natl Acad Sci U S A . 2015 ;112 : 13898 –13903 . 10.1073/pnas.1515558112 .26504196 45 Fox JM , Long F , Edeling MA , Lin H , van Duijl-Richter MK , Fong RH , et al Broadly neutralizing alphavirus antibodies bind an epitope on E2 and inhibit entry and egress . Cell . 2015 ;163 : 1095 –1107 . 10.1016/j.cell.2015.10.050 .26553503 46 Weber C , Buchner SM , Schnierle BS . A small antigenic determinant of the chikungunya virus E2 protein is sufficient to induce neutralizing antibodies which are partially protective in mice . PLoS Negl Trop Dis . 2015 ;9 : e0003684 10.1371/journal.pntd.0003684 .25905779 47 Khan M , Dhanwani R , Rao PV , Parida M . Subunit vaccine formulations based on recombinant envelope proteins of Chikungunya virus elicit balanced Th1/Th2 response and virus-neutralizing antibodies in mice . Virus Res . 2012 ;167 : 236 –246 . 10.1016/j.virusres.2012.05.004 .22610133 48 Kumar M , Sudeep AB , Arankalle VA . Evaluation of recombinant E2 protein-based and whole-virus inactivated candidate vaccines against chikungunya virus . Vaccine . 2012 ;30 : 6142 –6149 . 10.1016/j.vaccine.2012.07.072 .22884660 49 Akahata W , Yang ZY , Andersen H , Sun S , Holdaway HA , Kong WP , et al A virus-like particle vaccine for epidemic Chikungunya virus protects nonhuman primates against infection . Nat Med . 2010 ;16 : 334 –338 . 10.1038/nm.2105 .20111039 50 Metz SW , Gardner J , Geertsema C , Le TT , Goh L , Vlak JM , et al Effective chikungunya virus-like particle vaccine produced in insect cells . PLoS Negl Trop Dis . 2013 ;7 : e2124 10.1371/journal.pntd.0002124 .23516657 51 Puschnik A , Lau L , Cromwell EA , Balmaseda A , Zompi S , Harris E . Correlation between dengue-specific neutralizing antibodies and serum avidity in primary and secondary dengue virus 3 natural infections in humans . PLoS Negl Trop Dis . 2013 ;7 : e2274 10.1371/journal.pntd.0002274 .23785536 52 Pedersen GK , Hoschler K , Oie Solbak SM , Bredholt G , Pathirana RD , Afsar A , et al Serum IgG titres, but not avidity, correlates with neutralizing antibody response after H5N1 vaccination . Vaccine . 2014 ;32 : 4550 –4557 . 10.1016/j.vaccine.2014.06.009 .24950357 53 Cho B , Jeon BY , Kim J , Noh J , Kim J , Park M , et al Expression and evaluation of Chikungunya virus E1 and E2 envelope proteins for serodiagnosis of Chikungunya virus infection . Yonsei Med J . 2008 ;49 : 828 –835 . 10.3349/ymj.2008.49.5.828 .18972604 54 Sumathy K , Ella KM . Genetic diversity of Chikungunya virus, India 2006–2010: evolutionary dynamics and serotype analyses . J Med Virol . 2012 ;84 : 462 –470 . 10.1002/jmv.23187 .22246833 55 Kam YW , Pok KY , Eng KE , Tan LK , Kaur S , Lee WW , et al Sero-prevalence and cross-reactivity of chikungunya virus specific anti-E2EP3 antibodies in arbovirus-infected patients . PLoS Negl Trop Dis . 2015 ;9 : e3445 10.1371/journal.pntd.0003445 .25568956 56 Warter L , Lee CY , Thiagarajan R , Grandadam M , Lebecque S , Lin RT , et al Chikungunya virus envelope-specific human monoclonal antibodies with broad neutralization potency . J Immunol . 2011 ;186 : 3258 –3264 . 10.4049/jimmunol.1003139 .21278338 57 Gardner J , Anraku I , Le TT , Larcher T , Major L , Roques P , et al Chikungunya virus arthritis in adult wild-type mice . J Virol . 2010 ;84 : 8021 –8032 . 10.1128/JVI.02603-09 .20519386 58 Teo TH , Her Z , Tan JJ , Lum FM , Lee WW , Chan YH , et al Caribbean and La Reunion Chikungunya virus isolates differ in their capacity to induce pro-inflammatory Th1 and NK cell responses and acute joint pathology . J Virol . 2015 10.1128/JVI.00909-15 .25995257 59 Priya R , Dhanwani R , Patro IK , Rao PV , Parida MM . Differential regulation of TLR mediated innate immune response of mouse neuronal cells following infection with novel ECSA genotype of Chikungunya virus with and without E1:A226V mutation . Infect Genet Evol . 2013 ;20 : 396 –406 . 10.1016/j.meegid.2013.09.030 .24126361 60 Her Z , Teng TS , Tan JJ , Teo TH , Kam YW , Lum FM , et al Loss of TLR3 aggravates CHIKV replication and pathology due to an altered virus-specific neutralizing antibody response . EMBO Mol Med . 2015 ;7 : 24 –41 . 10.15252/emmm.201404459 .25452586

PMC005xxxxxx/PMC5003354.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757126910.1371/journal.pone.0162040PONE-D-16-03997Research ArticleBiology and Life SciencesBiochemistryLipidsPhospholipidsBiology and Life SciencesBiochemistryLipidsFatty AcidsBiology and Life SciencesAnatomyBody FluidsMilkBreast MilkMedicine and Health SciencesAnatomyBody FluidsMilkBreast MilkBiology and Life SciencesPhysiologyBody FluidsMilkBreast MilkMedicine and Health SciencesPhysiologyBody FluidsMilkBreast MilkBiology and Life SciencesAnatomyBody FluidsBloodBlood PlasmaMedicine and Health SciencesAnatomyBody FluidsBloodBlood PlasmaBiology and Life SciencesPhysiologyBody FluidsBloodBlood PlasmaMedicine and Health SciencesPhysiologyBody FluidsBloodBlood PlasmaMedicine and Health SciencesHematologyBloodBlood PlasmaPhysical SciencesChemistryChemical CompoundsAcidsLactic AcidBiology and Life SciencesBiochemistryLipidsBiology and Life SciencesBiochemistryMetabolismMetabolitesBiology and Life SciencesBiochemistryLipidsFatty AcidsLinoleic AcidPhospholipid Species in Newborn and 4 Month Old Infants after Consumption of Different Formulas or Breast Milk Phospholipid Species in Newborn and 4 Month Old InfantsUhl Olaf Fleddermann Manja Hellmuth Christian Demmelmair Hans Koletzko Berthold *Ludwig-Maximilians-University of Munich, Dr. von Hauner Children’s Hospital, Medical Centre of LMU Munich, München, GermanyMiyamoto Sayuri EditorUniversidade de Sao Paulo Instituto de Quimica, BRAZILCompeting Interests: The study was financially supported by HiPP GmbH & Co Vertrieb KG (Pfaffenhofen, Germany). This does not alter the authors' adherence to PLOS ONE policies on sharing data and materials. Conceptualization: HD BK. Formal analysis: OU MF CH. Funding acquisition: BK. Investigation: MF. Supervision: HD BK. Writing – original draft: OU. Writing – review & editing: HD BK. * E-mail: office.koletzko@med.lmu.de29 8 2016 2016 11 8 e016204028 1 2016 16 8 2016 © 2016 Uhl et al2016Uhl et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Introduction Arachidonic acid (AA) and docosahexaenoic acid (DHA) are important long-chain polyunsaturated fatty acids for neuronal and cognitive development and are ingredients of infant formulae that are recommended but there is no evidence based minimal supplementation level available. The aim of this analysis was to investigate the effect of supplemented AA and DHA on phospholipid metabolism. Methods Plasma samples of a randomized, double-blind infant feeding trial were used for the analyses of phospholipid species by flow-injection mass spectrometry. Healthy term infants consumed isoenergetic formulae (intervention formula with equal amounts of AA and DHA—IF, control formula without additional AA and DHA—CF) from the first month of life until the age of 120 days. A group of breast milk (BM) -fed infants was followed as a reference. Results The plasma profile detected in newborns was different from 4 month old infants, irrespective of study group. Most relevant changes were seen in higher level of LPC16:1, LPC20:4, PC32:1, PC34:1 and PC36:4 and lower level of LPC18:0, LPC18:2, PC32:2, PC36:2 and several ether-linked phosphatidylcholines in newborns. The sum of all AA and DHA species at 4 month old infants in the CF group showed level of 40% (AA) and 51% (DHA) of newborns. The supplemented amount of DHA resulted in phospholipid level comparable to BM infants, but AA phospholipids were lower than in BM infants. Interestingly, relative contribution of DHA was higher in ether-linked phosphatidylcholines in CF fed infants, but IF and BM fed infants showed higher overall ether-linked phosphatidylcholines levels. Conclusion In conclusion, we have shown that infant plasma phospholipid profile changes remarkably from newborn over time and is dependent on the dietary fatty acid composition. A supplementation of an infant formula with AA and DHA resulted in increased related phospholipid species. http://dx.doi.org/10.13039/501100000781European Research CouncilERC-2012-AdG no.322605Koletzko Berthold Hipp GmbH and Co Vertrieb KGThe study was financially supported by HiPP GmbH & Co Vertrieb KG (Pfaffenhofen, Germany). Additionally, the research leading to these results has received funding from the European Research Council Advanced Grant ERC-2012-AdG – no.322605 META-GROWTH. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction After birth, breast milk consumption is assumed to be the best nutrition to support optimal infantile development. Thus, infant formulae should be designed according to human milk composition to meet infant’s needs. Besides protein and carbohydrate content, lipid composition is essential, e.g. an essential fatty acid ratio of linoleic acid to alpha-linolenic acid ratio in the range of 5–15 is recommended [1, 2]. Arachidonic acid (AA) and docosahexaenoic acid (DHA) are the most important essential long-chain polyunsaturated fatty acids (LC-PUFA) which are recommended ingredients for infant formula, but there are no evidence based minimal supplementation level available. A series of studies have shown that LC-PUFA supplementation of infant formula improved the fatty acids status in different matrices such as plasma and red blood cells in infants [3]. Phospholipids are membrane lipids and known to be a good biomarker for LC-PUFA status, due to a high content of these essential fatty acids [4]. Thus, the analysis of plasma phospholipids is used to assess the bioavailability of supplemented LC-PUFA. For the development of new infant formulae, fatty acid analyses of plasma or red blood cell phospholipids are usually used to determine the fatty acid status [5]. Phospholipids are molecules with a lipophilic fatty acid side chain and a hydrophilic phosphate containing head group. Hundreds of diverse molecular species of phospholipids that differ in kind of head group, chain length, number and positions of double bonds and chemical bindings of fatty acids to the glycerol or sphingosine backbone [6] are known. There are several lipid molecules that contain fatty acid chains like AA or DHA. Thus, it seems to be important to know how supplemented LC-PUFA is incorporated in the lipidome and which metabolites reflect LC-PUFA best in infant plasma. To investigate this question, we chose a previously published randomized, double-blind intervention trial which investigated a new designed low protein intervention formula (IF) supplemented with equal amounts of AA and DHA (7.2 mg/ 100 mL) in comparison to a control formula (CF) without AA and DHA [7]. Fully breastfed infants were followed as a non-randomized reference group. The supplemented AA and DHA in the IF resulted in higher percentages of these fatty acids in plasma glycerophospholipids in comparison to CF [7]. The aim of this analysis was to investigate the effect of supplemented AA and DHA and the change of different plasma phospholipid species in the first month of life. We focused our analysis on choline containing phospholipid species, namely lyso-phosphatidylcholines (LPC), diacyl-phosphatidylcholines (PC), ether-linked phosphatidylcholines (PCe) and sphingomyelines (SM), which covered 95% of phospholipids [8]. Methods Study We analysed plasma samples of infants from the BeMIM study (Belgrade-Munich Infant Milk trial), a randomized, double-blind, controlled intervention study. Details on the study design, participating subjects, as well as details about infant formula composition have been previously published [7]. Briefly, healthy term infants were recruited until the age of 28 days and randomised to receive either an intervention formula (IF) with AA and DHA (from egg and fish oils) or a control formula (CF) without AA and DHA. Furthermore, IF had a lower protein content (1.89 g/100 kcal) and a higher fat content (5.3 g/ 100 kcal), compared to the CF (2.2 g protein and 4.9 g fat per 100 kcal). IF was enriched in alpha-lactalbumin and supplemented with free L-phenylalanine and L-tryptophan to meet required contents [2]. A reference group of exclusively breastfed infants was included (BM). Plasma samples were taken at 120 days of age and from a subgroup of subjects plasma samples were taken at recruitment before the study started (newborn, 0–5 days after birth). Newborn and subjects did not fully coincide after the intervention due to drop outs and thus a longitudinal analysis was unrewarding. The study was approved by the Clinical Center Serbia Ethical Committee. Written informed consent was obtained from all participating families. Analyses of this publication have been performed as secondary analyses of a randomized controlled trial and got approval from the Ethical Committee of Ludwig-Maximilians University of Munich, Germany. The original study was registered at Clinical Trials.gov (NCT01094080). The fatty acid compositions of both infant formulas are shown in Table 1. 10.1371/journal.pone.0162040.t001Table 1 Fatty acid content (g/100 mL) of the studied infant formulas. Intervention Formula Control Formula  C12:0 0.2 0  C14:0 0.1 0  C16:0 0.8 0.9  C16:1 0 0  C18:0 0.1 0.1  C18:1 1.5 1.4  C18:2 0.7 0.7  C18:3 0.1 0.1  C20:4 (mg/100 mL) 7.2 0  C22:6 (mg/100 mL) 7.2 0 Analysis of glycerophospholipids The plasma samples were analysed as described in S1 File and in previous studies [9–11]. Briefly, 10 μl plasma were diluted with methanol, containing internal standards, to correct for any effects appearing during sample preparation and ionization. After centrifugation, supernatants were used for flow-injection mass spectrometry analysis. The liquid chromatographic system (Agilent, Waldbronn, Germany) was coupled to a triple quadrupole mass spectrometer (QTRAP4000, Sciex, Darmstadt, Germany) with an electrospray ionization source. Mass spectrometric analysis was run in Multiple Reaction Monitoring mode. Quantification of metabolites has been done by comparison of signal-to-internal standard-ratios between samples and commercial available lyophilized aliquots of control plasma (Recipe, Germany) as one point calibration. The entire analytical process was post-processed by Analyst 1.5.1 and the isotopomer correction for up to M+4 was applied by R (programming language, version 3.0.1). The full analytical analysis comprised LPC, PC, PCe, SM, acylcarnitines and sum of hexoses, but data analyses focus only on phospholipids. As a point to note, the analytical technique applied here is not capable of determining the position of the double bonds and the distribution of carbon atoms between fatty acid side chains. The polar lipids are mentioned as X:Y. In this nomenclature, X is the length of the carbon chain, Y is the number of double bonds. There will be no differentiation between ether-linked and vinyl ether-linked phosphatidylcholines. However in PC, the ether-linked species are dominating and thus the interpretation of fatty acids follows this assumption. Furthermore, it needs to be addressed that there might be more than one species covered by one mass transition and the most concentrated species (according to lipidmaps.org [12] and human metabolome data base [13]) was used as species name. Data analysis A number of 484 plasma samples were measured randomly in batches composed of 81 samples, 6 quality control samples and 9 standards. Aliquots of a pooled plasma sample were used as quality control samples for intra- and inter-batch variation. Coefficients of variation were used as the criteria of quality. Metabolites are presented as median and interquartile ranges (IQR) in μmol/L. A principal component analyses (PCA) was carried out to identify possible separation of subjects according to grouping. Mann-Whitney-U tests were performed to investigate differences between groups. Significance was accepted at p <0.05. Due to multiple testing, the level of significance for p-values was corrected according to Bonferroni to p-values <5.49E-04. All statistical tests were performed with SPSS software version 21 (IBM, NY, USA) or Excel 2010. Results A total of 484 plasma samples (231 newborn, 79 CF, 83 IF and 91 BM) of the BeMIM trial have been determined for 425 phospholipids. Eleven LPC, 26 PC, 27 PCe and 27 SM were determined with intra-batch and inter-batch coefficient of variation less than 30% and considered for further statistical analysis. Details on the anthropometric data have been previously published [7]. Briefly, the mean age of the 231 newborn was 3 days (range 2 to 5) with a mean birthweight of 3437 g (SD 373 g). Changes from birth to 4 month of age The plasma profile of newborn was strictly different from the plasma profile of 4 month old infants, irrespective of study group as illustrated in the plot of principal component 1 and 2 of the PCA analyses (Fig 1). The separation was found approximately along the angle bisector and thus both principal components were responsible for the separation. Loadings of the principal components did not identify a specific class of phospholipids, responsible for the separation. Mann-Whitney-U test identified 75 metabolites with different concentrations in the newborn compared to BM infants at 4 month of age (Table 2). This corresponds to 82% of all detected species. Most relevant different species were LPC16:1, LPC20:4, PC32:1, PC34:1 and PC36:4 with higher concentrations and LPC18:0, LPC18:2, PC32:2, PC36:2 and several PCe with lower concentrations in the newborn. The sum of all PCe species was rising from newborn to 4 month old infants with median (IQR) values of 85 (32) μmol/L for newborn, 92 (22) μmol/L for CF, 109 (24) μmol/L for IF and 127 (35) μmol/L for BM. PC36:4 was the highest phospholipid species in newborn, while PC34:2 was highest at 4 month of age. Strong effects were seen in species with AA and DHA, especially to the group of CF, which did not receive any additional AA and DHA over the study period. Nearly all species with these two fatty acids were affected and thus the sum was calculated to assess the total effect (Fig 2). Highest values for AA were found at the time of birth and all groups showed significant lower level at 4 month. Lowest values for AA as well as for DHA species were detected in the CF group. The DHA content at 4 month of age was not different between the IF and the BM group compared to the newborn. 10.1371/journal.pone.0162040.g001Fig 1 Principal component analyses of the phospholipid profile of newborn (n = 231) and at 4 month of age fed an intervention formula (IF, n = 83), a control formula (CF, n = 79) or breast milk (BM, n = 91). 10.1371/journal.pone.0162040.g002Fig 2 Sum of all species containing docosahexaenoic acid (black bar) and arachidonic acid (white bars) in newborn (n = 231) and at 4 month feeding a control formula (CF, n = 79), intervention formula (IF, n = 83) or breast milk (BM, n = 91). Significant differences to newborn are marked with an asterisk. 10.1371/journal.pone.0162040.t002Table 2 Median (M) and interquartile ranges (IQR) of plasma lipids (in μmol/L) of newborn (NB, n = 231) and at 4 month of age fed an intervention formula (IF, n = 83), a control formula (CF, n = 79) or breast milk (BM, n = 91). Mann-Whitney-U tests with Bonferroni correction were performed to compare between the groups. Significant (Sig.) changes (p <0.05) are marked in gray. Metabolite NB CF IF BM NB vs BM CF vs IF IF vs BM M IQR M IQR M IQR M IQR LPC14:0 0.87 0.36 0.56 0.35 0.90 0.41 1.12 0.46 1.10E-11 3.67E-10 1.68E-05 LPC16:0 89.24 26.88 69.18 10.62 72.01 13.96 70.11 13.23 2.71E-14 3.43E-02 5.83E-02 LPC16:1 3.32 1.26 1.23 0.29 1.26 0.45 1.58 0.54 7.67E-38 9.67E-01 2.68E-07 LPC18:0 21.52 6.47 23.78 5.04 26.51 6.18 28.82 7.31 6.25E-27 2.37E-05 5.15E-04 LPC18:1 22.84 8.09 20.26 4.66 19.65 4.38 18.79 6.35 5.15E-09 4.79E-02 6.59E-01 LPC18:2 21.32 10.93 41.57 8.54 44.14 16.21 44.31 12.85 3.33E-33 1.01E-01 6.33E-01 LPC20:3 1.95 0.97 1.32 0.51 1.22 0.44 1.67 0.74 6.51E-04 1.49E-01 8.14E-10 LPC20:4 14.93 4.29 3.80 1.65 5.03 2.00 6.82 2.85 5.44E-38 4.47E-07 3.90E-12 LPC22:6 2.96 1.13 0.95 0.66 2.22 1.02 2.00 0.65 1.88E-22 8.31E-20 3.09E-02 LPCe16:0 0.24 0.13 0.26 0.16 0.29 0.15 0.47 0.22 6.88E-31 1.34E-01 1.22E-11 LPCe18:0 0.49 0.20 0.51 0.20 0.54 0.28 1.01 0.49 5.31E-35 1.01E-01 4.08E-14 PC30:0 3.16 1.38 1.32 0.78 2.26 0.72 3.02 1.25 2.33E-01 1.13E-10 1.04E-09 PC32:0 28.19 10.91 13.17 3.82 14.58 3.25 15.87 4.97 1.46E-33 4.96E-03 5.92E-03 PC32:1 24.49 12.26 7.14 2.05 7.36 2.72 8.69 3.36 4.27E-42 4.07E-01 3.01E-04 PC32:2 0.79 0.64 1.33 0.58 1.79 0.75 1.69 0.97 8.42E-28 8.34E-08 7.01E-01 PC34:1 249.96 121.42 188.03 44.08 182.17 44.89 149.28 39.94 5.63E-34 2.50E-01 1.31E-07 PC34:2 246.66 128.86 440.37 86.72 468.43 115.00 413.68 113.86 1.19E-22 1.65E-01 1.18E-04 PC34:3 4.70 2.66 7.55 2.24 8.93 2.44 6.77 2.47 6.45E-14 9.36E-06 1.04E-10 PC34:4 0.46 0.21 0.41 0.24 0.61 0.27 0.71 0.31 5.86E-17 1.02E-08 3.35E-03 PC34:5 0.08 0.06 0.06 0.04 0.11 0.06 0.08 0.05 3.63E-01 8.19E-08 9.48E-05 PC36:0 1.90 1.01 1.24 0.86 2.59 0.80 2.32 0.72 3.61E-05 3.01E-19 3.58E-04 PC36:1 40.54 18.94 42.50 9.20 43.85 11.02 41.40 11.92 9.12E-01 4.81E-01 5.83E-02 PC36:2 129.68 69.80 267.34 50.27 284.88 72.88 289.96 88.42 1.76E-36 5.90E-02 5.60E-01 PC36:3 75.49 38.56 107.04 23.38 95.96 26.04 107.78 37.21 1.82E-13 1.17E-03 2.40E-03 PC36:4 284.46 108.48 103.66 31.36 130.38 34.52 162.10 55.63 8.04E-34 6.93E-08 2.24E-07 PC36:5 4.89 2.64 3.74 1.86 6.03 2.49 4.71 1.48 1.37E-01 8.36E-15 4.27E-08 PC36:6 0.20 0.11 0.17 0.12 0.38 0.18 0.31 0.11 1.42E-16 3.33E-16 4.31E-05 PC38:1 0.64 0.62 0.94 0.54 1.33 0.53 1.39 0.44 1.51E-23 5.43E-08 4.95E-01 PC38:3 20.81 10.54 22.64 6.36 22.46 7.22 26.03 10.32 1.85E-08 1.96E-01 1.79E-09 PC38:4 170.97 61.08 69.42 28.41 89.31 24.20 136.35 52.48 2.73E-13 7.78E-07 3.58E-15 PC38:5 33.74 13.94 28.93 8.75 30.59 8.54 35.08 12.30 3.20E-01 1.23E-01 1.79E-05 PC38:6 74.86 32.02 35.39 25.96 83.73 34.15 68.30 22.33 2.97E-04 2.47E-20 2.69E-06 PC40:4 2.89 1.35 2.51 0.72 2.30 0.65 3.20 1.22 1.42E-03 7.76E-03 4.90E-15 PC40:5 6.89 3.57 7.01 2.08 7.21 2.38 8.46 3.16 5.60E-07 9.75E-01 1.13E-06 PC40:6 22.92 10.53 12.51 9.33 28.19 11.06 25.11 8.05 1.62E-02 5.52E-19 6.15E-02 PC42:5 0.23 0.11 0.26 0.12 0.27 0.11 0.35 0.12 7.63E-19 3.22E-01 1.14E-08 PC44:12 0.73 0.33 0.94 0.26 1.14 0.33 1.04 0.29 1.11E-15 3.65E-08 2.55E-03 PCe32:0 3.34 1.30 3.07 0.78 2.97 0.72 3.11 0.76 1.92E-02 8.32E-01 2.33E-01 PCe32:1 3.15 1.19 2.07 0.50 2.39 0.65 2.40 0.70 5.08E-16 1.38E-05 6.73E-01 PCe32:2 0.50 0.20 0.25 0.12 0.30 0.12 0.30 0.16 7.48E-20 3.35E-04 1.04E-01 PCe34:0 1.03 0.42 0.63 0.31 0.67 0.42 1.25 0.60 1.15E-03 3.81E-01 5.39E-14 PCe34:1 5.43 2.49 7.06 1.83 6.87 1.79 6.86 1.72 2.03E-08 5.06E-01 6.52E-01 PCe34:2 3.89 1.52 7.08 1.78 6.90 1.80 6.32 1.77 4.00E-27 2.14E-01 3.69E-02 PCe34:3 2.34 1.32 4.53 1.07 5.08 1.52 4.17 1.38 5.34E-28 4.08E-02 2.11E-03 PCe34:4 0.08 0.05 0.09 0.04 0.13 0.06 0.12 0.05 6.39E-09 2.32E-07 9.13E-04 PCe36:0 0.71 0.30 0.53 0.20 0.60 0.16 0.65 0.22 2.21E-03 1.01E-03 1.46E-01 PCe36:1 1.64 0.73 1.76 0.59 1.93 0.71 2.87 1.16 9.72E-24 5.22E-02 5.03E-11 PCe36:2 2.86 1.37 6.26 1.57 6.56 1.66 8.90 4.09 3.90E-42 5.30E-02 5.35E-10 PCe36:3 1.88 0.78 4.64 1.22 4.40 1.24 4.51 1.29 1.48E-39 7.94E-02 6.61E-01 PCe36:4 12.50 4.97 10.09 2.72 11.72 3.01 13.59 4.41 1.74E-02 9.22E-06 2.03E-06 PCe36:5 10.01 4.39 6.61 2.26 8.99 2.57 11.24 3.14 2.93E-02 8.65E-11 2.90E-07 PCe36:6 0.08 0.13 0.11 0.16 0.19 0.14 0.19 0.13 1.40E-13 4.02E-06 9.12E-01 PCe38:0 0.73 0.37 0.52 0.35 1.07 0.44 1.06 0.44 1.78E-15 2.71E-19 6.90E-01 PCe38:4 9.43 3.38 7.42 2.23 8.75 2.45 13.60 6.24 7.46E-15 1.95E-05 1.69E-15 PCe38:5 9.82 4.04 13.63 3.98 16.25 3.43 20.14 6.28 5.68E-35 2.45E-07 6.93E-09 PCe38:6 4.52 1.94 3.71 1.98 6.40 1.36 7.09 1.56 1.82E-19 3.59E-17 4.51E-03 PCe40:0 2.93 1.29 2.39 1.31 4.10 1.27 3.88 1.18 1.00E-09 4.39E-16 3.06E-02 PCe40:1 0.90 0.33 0.67 0.38 1.22 0.40 1.14 0.39 4.12E-09 2.90E-16 1.02E-01 PCe40:4 1.39 0.57 1.22 0.37 1.33 0.40 1.84 0.74 1.82E-11 2.94E-03 3.03E-13 PCe40:5 1.70 0.63 2.17 0.77 2.57 0.83 3.53 1.23 1.50E-37 1.89E-05 4.83E-12 PCe40:6 1.95 0.73 2.19 1.03 3.58 0.75 4.15 1.32 1.94E-37 1.28E-15 4.53E-07 PCe42:2 0.18 0.11 0.21 0.10 0.25 0.09 0.26 0.08 7.90E-14 3.09E-05 7.69E-01 PCe42:5 0.68 0.31 0.87 0.35 1.09 0.29 1.17 0.41 2.56E-28 1.45E-07 3.67E-02 PCe42:6 0.70 0.29 0.73 0.32 0.98 0.27 1.10 0.38 7.20E-23 5.63E-10 1.25E-03 SM12:0 0.32 0.15 0.39 0.38 0.88 0.29 0.86 0.45 2.73E-37 2.13E-16 6.27E-01 SM14:0 6.87 2.48 15.82 4.80 15.23 3.08 14.86 4.36 1.03E-38 4.25E-01 6.81E-02 SM14:1 0.61 0.23 0.35 0.22 0.61 0.22 0.78 0.42 1.10E-11 5.36E-13 3.30E-07 SM15:0 3.83 1.60 5.85 1.56 5.68 1.26 5.95 1.44 3.96E-22 5.46E-02 1.75E-02 SM16:0 103.65 32.88 102.98 20.99 118.18 17.23 125.20 28.13 2.00E-08 1.04E-06 2.21E-02 SM16:1 17.21 5.52 11.47 2.59 12.70 2.23 14.61 4.04 6.21E-08 3.05E-05 7.94E-06 SM17:0 2.20 0.88 1.83 0.50 1.86 0.56 3.13 1.30 1.60E-13 8.78E-02 5.60E-16 SM18:0 33.43 11.19 21.16 5.56 24.28 5.03 27.88 8.52 1.02E-10 4.83E-04 1.10E-06 SM18:1 18.74 6.71 7.02 2.17 8.62 2.33 12.78 5.20 1.16E-25 7.92E-07 3.70E-13 SM18:2 0.55 0.26 0.48 0.18 0.60 0.23 1.07 0.56 7.32E-24 2.40E-05 1.85E-14 SM19:0 1.40 0.60 1.56 0.75 1.58 0.53 1.89 0.54 4.09E-13 2.70E-01 2.25E-04 SM20:0 22.49 9.77 33.96 10.91 34.92 9.33 30.67 9.99 1.98E-10 4.68E-01 1.14E-04 SM20:1 10.26 4.12 11.30 2.40 12.27 3.42 11.36 2.67 5.62E-02 9.00E-03 5.44E-03 SM21:0 1.75 0.75 5.52 2.01 4.95 1.36 4.27 1.58 1.86E-39 1.42E-02 6.24E-06 SM22:0 14.30 5.00 13.01 6.59 14.58 5.10 11.13 6.03 4.19E-11 1.38E-01 1.13E-06 SM22:1 17.68 7.51 25.20 6.47 28.00 6.51 26.01 6.62 3.65E-21 2.58E-02 1.81E-01 SM22:3 5.84 1.97 2.44 0.76 2.85 0.84 3.23 0.93 4.17E-33 1.23E-07 1.24E-02 SM22:4 0.43 0.18 0.24 0.10 0.36 0.15 0.36 0.12 1.45E-05 5.63E-10 4.41E-01 SM23:0 3.57 1.76 6.91 1.88 6.93 1.53 6.48 1.84 3.46E-30 4.40E-01 2.16E-02 SM24:0 14.78 5.14 10.65 2.55 12.22 2.29 12.00 3.59 1.18E-10 2.88E-06 4.97E-01 SM24:1 35.04 12.74 29.50 8.33 37.82 7.29 35.11 8.96 8.47E-01 1.48E-11 1.86E-01 SM24:2 22.31 8.13 14.63 4.33 17.10 3.85 21.45 7.84 3.49E-02 6.67E-08 3.60E-08 SM24:3 11.96 4.13 5.10 1.95 6.17 1.99 9.71 4.55 2.44E-11 5.92E-05 1.82E-15 SM24:5 3.95 1.55 1.82 1.41 4.06 1.48 3.38 1.16 1.16E-06 3.27E-20 6.93E-05 SM25:1 0.87 0.43 0.54 0.24 0.70 0.26 1.15 0.60 8.53E-08 3.03E-07 7.86E-13 SM13:0 0.13 0.06 0.13 0.08 0.16 0.09 0.23 0.13 7.62E-24 1.86E-05 7.81E-07 SM17:1 0.31 0.15 0.20 0.10 0.23 0.10 0.46 0.22 3.27E-12 1.62E-02 2.17E-16 Group comparison To find out what effect supplemented AA and DHA on different plasma phospholipid species might have at 4 month of age, a Mann-Whitney-U test between the two formula groups was performed. A total of 51 species turned out to be significant different, which corresponds to 56% of detected metabolites (Table 2). Phospholipid species containing 6 double bonds presumably represent the presence of DHA, which are in detail LPC22:6, PC36:6, PC38:6, PC40:6, PCe36:6, PCe38:6, PCe40:6, and PCe42:6. The concentrations of all 8 species were significant lower in CF. In IF fed infants, all di-acylated PC species were between 2.25 and 2.37 fold higher. PCe species were enriched between 1.34 and 1.75 fold. Phospholipids containing AA were presumably LPC20:4, PC34:4, PC36:4, PC38:4, PC40:4, PCe34:4, PCe36:4, PCe38:4, and PCe40:4. All but PC40:4 and PCe40:4 were higher in the IF group. To evaluate compositional shifts, the relative distribution of all DHA and AA containing species was calculated, separately. In the CF, a higher compositional content of DHA was found for the species PCe38:6, PCe40:6 and PCe42:6 and a lower content was found for PC38:6. The composition of AA species was affected with higher values of PC34:4 and lower values of PC40:4. Discussion The phospholipid profile of infants a few days after birth was quite different to the phospholipid profiles at 4 months of age, irrespective of the study group. We found remarkable high concentrations of LPC(20:4) and PC(36:4) in newborn. PC(36:4) is composed to the main extend of PC(16:0/20:4) [14], which represents the phospholipid species with the highest content of AA. AA is known to be preferably transported through the placenta, but the process of transport is not jet completely understood [15]. In a previous study we found a high correlation of cord blood AA to placental PC(16:0/20:4) content and we speculated that placental PC(16:0/20:4) might be the source of AA for the fetal development and that this molecule plays a major role in the transport mechanism [16]. In the current study we could now specify the AA content in the newborn metabolism to 9 different AA containing species with a high proportion in the molecular species of PC(16:0/20:4). Furthermore, the high concentration of PC(36:4) in newborn might fit to the theory that fatty acids are released as free fatty acids from the placenta to the fetal circulation and then are further metabolised to phospholipids to be protect from the placental reabsorption [17]. Our finding of high LPC(20:4) concentration are not easy understandable, since lyso-phospholipids are thought to be produced from phospholipases A1 or A2. In phospholipids the AA is located at the sn-2 position and thus a phospholipase A1 would be necessary to produce LPC(20:4), albeit the phospholipases A1 is rarely distributed. Thus, it seems to be unrealistic that a phospholipase is responsible for the LPC(20:4). A further possible candidate might be the lecithin-acyl-transferase (LCAT), which transfers a fatty acid from phospholipids to cholesterol and releases a lyso-phospholipid. The sn-1 activity of LCAT was found to be dependent on the fatty acid chain length of the fatty acid on position sn-1 with 50% sn-1 activity for PC(16:0/20:4) [18]. Thus, this might be a good candidate for LPC(20:4) source. However, a new developed transporter for lyso-phospholipids was discussed to be responsible for the transport of LPC(22:6) through the placenta [19]. One could speculate about this process being also responsible for the transport of AA to the fetal metabolism. In case of no further nutritional supply of LC-PUFA, their concentration decrease after birth and remain constant at lower level after a period of less than 3 months as shown in a study on the 13C-content of plasma fatty acids over time [20]. At the age of 4 months, the lower concentrations of phospholipid species with AA or DHA represents the endogenous synthesis of AA and DHA from linoleic acid (18:2n-6) and alpha-linolenic acid (18:3n-3) of CF feeding without additional AA and DHA supply, respectively [21]. The comparison between CF and newborn allowed assessing roughly the contribution of placental transfer to AA and DHA content in comparison to endogenous production. Fig 2 shows the level of all AA and DHA species in newborn and CF. In CF, level were significant lower with values of 40% AA and 51% DHA at 4 month of age compared to newborn. Thus, we conclude that the contribution of placental transfer during pregnancy of AA and DHA must be more than 50%. During pregnancy, the placental contribution is presumably even higher, since the endogenous production is not fully developed. Breast milk is very complex and can show very high variation in the content. Thus, effects can always be multifactorial and causal changes might be difficult. In vivo studies of AA and DHA can be affected by their precursors such as dihomo-γ-linolenic acid and eicosapentaenoic acid. In this study, the two formula groups have the same fatty acid composition up to additional AA and DHA in the intervention group and effects can be limited to these fatty acids. The comparison between the IF and CF group allowed to assess the effect of dietary LC-PUFA to the endogenously produced LC-PUFA in very similar feeding groups. In IF group, all AA and DHA containing species but PC40:4 were enriched with smaller differences in species containing AA, but higher increases in DHA containing species. The composition of DHA containing species was slightly different between the CF and the IF with accumulation of DHA in PCe in the CF. Thus, either endogenously produced DHA is preferably incorporated into PCe or PCe are preferred when DHA is low abundant to prevent from oxidation. Tracer-studies might be able to answer this question in future studies. The phospholipid profile in newborn showed very low level (~50%) of LPC(18:2), PC(32:2), PC(34:2), PC(36:2) and SM(18:2) in comparison to 4 month old infants. All species presumably contain linoleic acid (18:2) which is the precursor of endogenously synthesised AA. AA and DHA can be converted from linoleic acid and alpha-linolenic acid (18:3), respectively by elongation and desaturation of the carbon chain [22]. Phospholipid species containing alpha-linolenic acid, which are PC(34:3), PC(36:3) and PC(38:3), were also lower (~70%) in newborn than in infants at 4 month of age, although with smaller deviation. The low abundance of phospholipid species containing 18:2 and 18:3 in newborn indicates a low placental supply, while AA and DHA were much higher abundant. Since the conversion rate of precursors (18:2 and 18:3) to AA and DHA is low in adults, the rate might be even lower in foetuses and infants [23]. High demands during late pregnancy for cognitive and visual development is ensured by the direct placental supply of high level of AA and DHA. The total amount of PCe was increasing from newborn to 4 month old infants, regardless of the study group. Regarding to literature, PCe might be important for the healthy neuronal development and could serve as reservoir for LC-PUFA [24]. Our results showed a dietary effect on the PCe content and we speculate about positive long-term effects for cognitive functions. PCe contain high amounts of LC-PUFA and thus the quantity of PCe might be explained by the consumption of LC-PUFA. The CF did not obtain any additional AA or DHA, while the IF was supplemented with AA and DHA and breast milk presumably contain highest values of different LC-PUFA [25]. A suggested preferred incorporation into PCe was not detected. As shown in Table 3, IF fed infants showed lower percentages of DHA within PCe species than infants fed CF. Moreover, PCe species without AA or DHA, such as PCe38:0, PCe40:1 or PCe40:0 were also higher in the group of IF, in comparison to the CF group. Thus, the synthesis of all PCe was promoted by the consumption of the IF. Nevertheless, infants of the BM group showed the highest values of all PCe species, which emphasized the outstanding role of breast milk for infants in the context of the suggested healthy properties of PCe. 10.1371/journal.pone.0162040.t003Table 3 Median values and interquartile ranges (IQR) in % of total arachidonic acid and docosahexaenoic acid species in infant plasma at 4 month of age. Significant changes between groups are calculated by Mann-Whitney-U tests after Bonferroni correction and marked in gray. Control formula (n = 79) Intervention formula (n = 83) p-value   Median IQR Median IQR Arachidonic acid containing species       LPC20:4 1.90% 0.49% 1.99% 0.43% 6.96E-02 PC34:4 0.21% 0.07% 0.24% 0.09% 9.95E-05 PC36:4 51.41% 2.11% 51.55% 2.64% 2.26E-01 PC38:4 35.75% 2.47% 36.20% 2.58% 4.05E-01 PC40:4 1.30% 0.23% 0.94% 0.24% 9.21E-19 PCe34:4 0.05% 0.02% 0.05% 0.02% 2.74E-02 PCe36:4 4.99% 0.87% 4.69% 0.72% 2.13E-03 PCe38:4 3.64% 0.69% 3.53% 0.69% 2.55E-02 PCe40:4 0.61% 0.15% 0.53% 0.13% 7.63E-04 Docosahexaenoic acid containing species   LPC22:6 1.62% 0.49% 1.70% 0.57% 2.49E-02 PC36:6 0.30% 0.16% 0.30% 0.12% 6.47E-01 PC38:6 63.65% 2.57% 66.08% 3.15% 5.88E-14 PC40:6 22.09% 1.72% 22.29% 1.93% 8.92E-01 PCe36:6 0.16% 0.23% 0.15% 0.11% 3.47E-01 PCe38:6 6.70% 1.27% 5.04% 1.22% 9.29E-15 PCe40:6 3.84% 1.10% 2.81% 0.85% 5.33E-16 PCe42:6 1.21% 0.41% 0.81% 0.33% 1.43E-17 For an optimal development of infants, breast feeding is assumed to be the best choice of nutrition during the first 6 months of life. To optimise infant formula fatty acid content, the analyses of molecular phospholipid species may help to investigate metabolic relations between nutrition and physiological outcomes. The plasma phospholipid profile of infants fed formula was separated from the BM infants by PCA. Thus, different diets are represented by the plasma phospholipid profile. The IF infants showed a phospholipid profile between BM and CF. This result was confirmed by the group comparison which showed that less species were different between the IF and BM infants, than between CF and BM. Our results showed that dietary availability of LC-PUFA has a huge impact on the plasma phospholipids pattern of infants. Several studies have shown that supplementation of infant formula did improve the LC-PUFA status in different compartments. Regarding literature of mature breast milk fatty acid composition of triacylglycerols, the concentration of AA (0.41 +/- 0.05%) is at least twice as high as DHA (0.18 +/- 0.02%) [25]. In our study, the IF was added by equal amounts (72 mg/ 100 mL) of AA and DHA. Plasma concentrations of phospholipids containing DHA were within the same region for infants receiving the IF than BM and thus the IF provided appropriate DHA to match breast milk (Fig 2). However, a concentration of 72 mg/100 mL AA seems not to be enough, compared to BM infants (Fig 2). To realize a fatty acid composition of infant formulas most closely to the breastmilk, a higher concentration of AA than 72 mg/ 100 ml should be considered. A limitation of the study might be the missing of further interesting lipid fractions, such as phosphatidylethanolamines, cholesterol ester or triacylglycerols. Since the used method is a screening method for a serious of phospholipid species, there are limitations in quantification and identification. Identification occurred based on mass spectrometric detection and in some cases there might be more than one species behind the mass transition. Identification occurred to our best knowledge for the highest concentrated species. A further limitation might be the correction of PCe and SM by internal standard of PC(28:0), which might not compensate for all class specific effects during sample preparation and ionisation. However, we belief that PC(28:0) might be suitable for compensating the overwhelming class of effects because of very similar chemical constitution of PC, PCe and SM. Quantification was applied including isotopomer correction, which can result in higher variation of low concentrated species. Isotopomer correction was made until M+4 isotopomers to ensure that no false positive results were obtained. Conclusion In conclusion, we have shown that infant plasma phospholipid profile changes from newborn to 4 month old infants and is dependent on the dietary fatty acid composition. Infants of CF group, which did not obtain additional LC-PUFA showed values of 40% (AA) and 51% (DHA) in comparison to newborn. The supplementation of an infant formula with AA and DHA resulted in markedly increased related phospholipid species. DHA was enriched in PCe of CF fed infants, probably to save residual DHA from oxidation, while PCe synthesis may be enhanced by PUFA supply in breast fed and IF infants. Supporting Information S1 File Method description. (PDF) Click here for additional data file. Abbreviations PCeether-linked phosphatidylcholines AAarachidonic acid BMbreastmilk CFcontrol formula PCdiacyl-phosphatidylcholines DHAdocosahexaenoic acid IQRinterquartile ranges IFintervention formula LC-PUFAlong-chain polyunsaturated fatty acids LPClyso-phosphatidylcholines PCAprincipal component analyses SMsphingomyelines ==== Refs References 1 Koletzko B , Baker S , Cleghorn G , Neto UF , Gopalan S , Hernell O , et al Global standard for the composition of infant formula: recommendations of an ESPGHAN coordinated international expert group . Journal of pediatric gastroenterology and nutrition . 2005 ;41 :584 –99 . 16254515 2 The Commission of the European Communities . Commission Directive 2006/141/EC of 22 December 2006 on infant formulae and amending Directive 1999/21/EC . Official Journal of the European Union . 2006 ;L401 :1 –33 . 3 Makrides M , Uauy R . LCPUFAs as conditionally essential nutrients for very low birth weight and low birth weight infants: metabolic, functional, and clinical outcomes-how much is enough? Clinics in perinatology . 2014 ;41 :451 –61 . 10.1016/j.clp.2014.02.012 24873843 4 Fekete K , Marosvolgyi T , Jakobik V , Decsi T . Methods of assessment of n-3 long-chain polyunsaturated fatty acid status in humans: a systematic review . The American journal of clinical nutrition . 2009 ;89 :2070S –84S . 10.3945/ajcn.2009.27230I 19420097 5 Kuratko CN , Salem N Jr. Biomarkers of DHA status . Prostaglandins, leukotrienes, and essential fatty acids . 2009 ;81 :111 –8 . 10.1016/j.plefa.2009.05.007 19545987 6 Dowhan W . Molecular basis for membrane phospholipid diversity: why are there so many lipids? Annual review of biochemistry . 1997 ;66 :199 –232 . 9242906 7 Fleddermann M , Demmelmair H , Grote V , Nikolic T , Trisic B , Koletzko B . Infant formula composition affects energetic efficiency for growth: the BeMIM study, a randomized controlled trial . Clinical nutrition . 2014 ;33 :588 –95 . 10.1016/j.clnu.2013.12.007 24411489 8 Hodson L , Skeaff CM , Fielding BA . Fatty acid composition of adipose tissue and blood in humans and its use as a biomarker of dietary intake . Progress in lipid research . 2008 ;47 :348 –80 . 10.1016/j.plipres.2008.03.003 18435934 9 Rauschert S , Uhl O , Koletzko B , Kirchberg F , Mori TA , Huang RC , et al Lipidomics reveals associations of phospholipids with obesity and insulin resistance in young adults . J Clin Endocrinol Metab . 2015 :jc20153525. 10 Lindsay KL , Hellmuth C , Uhl O , Buss C , Wadhwa PD , Koletzko B , et al Longitudinal Metabolomic Profiling of Amino Acids and Lipids across Healthy Pregnancy . PLoS One . 2015 ;10 :e0145794 10.1371/journal.pone.0145794 26716698 11 Larque E , Demmelmair H , Gil-Sanchez A , Prieto-Sanchez MT , Blanco JE , Pagan A , et al Placental transfer of fatty acids and fetal implications . The American journal of clinical nutrition . 2011 ;94 :1908S –13S . 10.3945/ajcn.110.001230 21562082 12 Fahy E , Subramaniam S , Murphy RC , Nishijima M , Raetz CR , Shimizu T , et al Update of the LIPID MAPS comprehensive classification system for lipids . Journal of lipid research . 2009 ;50 Suppl :S9 –14 . 10.1194/jlr.R800095-JLR200 19098281 13 Wishart DS , Jewison T , Guo AC , Wilson M , Knox C , Liu Y , et al HMDB 3.0—The Human Metabolome Database in 2013 . Nucleic acids research . 2013 ;41 :D801 –7 . 10.1093/nar/gks1065 23161693 14 Uhl O , Glaser C , Demmelmair H , Koletzko B . Reversed phase LC/MS/MS method for targeted quantification of glycerophospholipid molecular species in plasma . Journal of chromatography B, Analytical technologies in the biomedical and life sciences . 2011 ;879 :3556 –64 . 10.1016/j.jchromb.2011.09.043 22014895 15 Koletzko B , Larque E , Demmelmair H . Placental transfer of long-chain polyunsaturated fatty acids (LC-PUFA) . Journal of perinatal medicine . 2007 ;35 Suppl 1 :S5 –11 . 17302540 16 Uhl O , Demmelmair H , Segura MT , Florido J , Rueda R , Campoy C , et al Effects of obesity and gestational diabetes mellitus on placental phospholipids . Diabetes research and clinical practice . 2015 ;109 :364 –71 . 10.1016/j.diabres.2015.05.032 26021978 17 Haggarty P . Fatty acid supply to the human fetus . Annual review of nutrition . 2010 ;30 :237 –55 . 10.1146/annurev.nutr.012809.104742 20438366 18 Subbaiah PV , Liu M , Paltauf F . Role of sn-2 acyl group of phosphatidylcholine in determining the positional specificity of lecithin-cholesterol acyltransferase . Biochemistry . 1994 ;33 :13259 –66 . 7947733 19 Prieto-Sanchez MT , Ruiz-Palacios M , Blanco-Carnero JE , Pagan A , Hellmuth C , Uhl O , et al Placental MFSD2a transporter is related to decreased DHA in cord blood of women with treated gestational diabetes . Clinical nutrition . 2016 . 20 Carnielli VP , Simonato M , Verlato G , Luijendijk I , De Curtis M , Sauer PJ , et al Synthesis of long-chain polyunsaturated fatty acids in preterm newborns fed formula with long-chain polyunsaturated fatty acids . The American journal of clinical nutrition . 2007 ;86 :1323 –30 . 17991642 21 Glaser C , Heinrich J , Koletzko B . Role of FADS1 and FADS2 polymorphisms in polyunsaturated fatty acid metabolism . Metabolism: clinical and experimental . 2010 ;59 :993 –9 .20045144 22 Gibson RA , Muhlhausler B , Makrides M . Conversion of linoleic acid and alpha-linolenic acid to long-chain polyunsaturated fatty acids (LCPUFAs), with a focus on pregnancy, lactation and the first 2 years of life . Maternal & child nutrition . 2011 ;7 Suppl 2 :17 –26 .21366864 23 Koletzko B , Lien E , Agostoni C , Bohles H , Campoy C , Cetin I , et al The roles of long-chain polyunsaturated fatty acids in pregnancy, lactation and infancy: review of current knowledge and consensus recommendations . Journal of perinatal medicine . 2008 ;36 :5 –14 . 10.1515/JPM.2008.001 18184094 24 Braverman NE , Moser AB . Functions of plasmalogen lipids in health and disease . Biochimica et biophysica acta . 2012 ;1822 :1442 –52 . 10.1016/j.bbadis.2012.05.008 22627108 25 Sala-Vila A , Castellote AI , Rodriguez-Palmero M , Campoy C , Lopez-Sabater MC . Lipid composition in human breast milk from Granada (Spain): changes during lactation . Nutrition . 2005 ;21 :467 –73 . 15811767

PMC005xxxxxx/PMC5003355.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757135710.1371/journal.pone.0161828PONE-D-16-09216Research ArticleComputer and Information SciencesNetwork AnalysisProtein Interaction NetworksBiology and Life SciencesBiochemistryProteomicsProtein Interaction NetworksBiology and Life SciencesBiochemistryBioenergeticsEnergy-Producing OrganellesMitochondriaBiology and Life SciencesCell BiologyCellular Structures and OrganellesEnergy-Producing OrganellesMitochondriaBiology and life sciencesBiochemistryProteinsDNA-binding proteinsResearch and Analysis MethodsDatabase and Informatics MethodsBiological DatabasesProteomic DatabasesBiology and Life SciencesBiochemistryProteomicsProteomic DatabasesBiology and Life SciencesCell BiologyCellular Structures and OrganellesVesiclesBiology and Life SciencesCell BiologyCellular Structures and OrganellesCell MembranesMembrane ProteinsBiology and Life SciencesCell BiologyCell PhysiologyCell BindingMedicine and Health SciencesOncologyCancers and NeoplasmsBlastomasGlioblastoma MultiformeMedicine and Health SciencesOncologyCancers and NeoplasmsNeurological TumorsGlioblastoma MultiformeMedicine and Health SciencesNeurologyNeurological TumorsGlioblastoma MultiformeProtein Co-Expression Analysis as a Strategy to Complement a Standard Quantitative Proteomics Approach: Case of a Glioblastoma Multiforme Study Protein Co-Expression Analysis of Glioblastoma MultiformeKanonidis Evangelos I. 1Roy Marcia M. 2Deighton Ruth F. 3http://orcid.org/0000-0003-0498-8063Le Bihan Thierry 1*1 SynthSys and School of Biological Sciences, Waddington building, University of Edinburgh, Edinburgh, United Kingdom, EH9 3BF2 Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh United Kingdom, EH16 4SB3 Edinburgh Medical School: Deanery of Biomedical Sciences, University of Edinburgh, Edinburgh, United Kingdom, EH8 9AGRabilloud Thierry EditorCentre National de la Recherche Scientifique, FRANCECompeting Interests: The authors have declared that no competing interests exist. Conceptualization: EIK TLB MMR. Data curation: TLB MMR. Formal analysis: EIK TLB MMR. Funding acquisition: TLB. Investigation: EIK TLB. Methodology: EIK TLB MMR. Project administration: TLB. Resources: TLB. Software: EIK TLB. Supervision: TLB. Validation: EIK TLB MMR RD. Visualization: TLB EIK MMR. Writing – original draft: TLB EIK MMR. Writing – review & editing: TLB EIK MMR RD. * E-mail: thierry.lebihan@ed.ac.uk29 8 2016 2016 11 8 e01618283 3 2016 14 8 2016 © 2016 Kanonidis et al2016Kanonidis et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Although correlation network studies from co-expression analysis are increasingly popular, they are rarely applied to proteomics datasets. Protein co-expression analysis provides a complementary view of underlying trends, which can be overlooked by conventional data analysis. The core of the present study is based on Weighted Gene Co-expression Network Analysis applied to a glioblastoma multiforme proteomic dataset. Using this method, we have identified three main modules which are associated with three different membrane associated groups; mitochondrial, endoplasmic reticulum, and a vesicle fraction. The three networks based on protein co-expression were assessed against a publicly available database (STRING) and show a statistically significant overlap. Each of the three main modules were de-clustered into smaller networks using different strategies based on the identification of highly connected networks, hierarchical clustering and enrichment of Gene Ontology functional terms. Most of the highly connected proteins found in the endoplasmic reticulum module were associated with redox activity while a core of the unfolded protein response was identified in addition to proteins involved in oxidative stress pathways. The proteins composing the electron transfer chain were found differently affected with proteins from mitochondrial Complex I being more down-regulated than proteins from Complex III. Finally, the two pyruvate kinases isoforms show major differences in their co-expressed protein networks suggesting roles in different cellular locations. http://dx.doi.org/10.13039/501100004963Seventh Framework Programmeno 602470http://orcid.org/0000-0003-0498-8063Le Bihan Thierry This project has received funding from the European Union’s Seventh Framework Programme for research, technological development and demonstration under grant agreement no 602470. Data AvailabilityAll data are in supplementary information and in Pride: PRD000620.Data Availability All data are in supplementary information and in Pride: PRD000620. ==== Body Introduction Large-scale quantitative proteomic analysis acquired under different conditions has been used to gain deeper insight into protein function and regulation [1, 2]. One widely used approach consists of comparing the level of expression of a given protein between different conditions and to determine whether or not the difference between the various groups is meaningful based on statistical analysis [3]. The following step, which consists of assigning a biological function context to the proteomics data or identifying key molecular targets, remains a challenging task. Correlation within gene expression (i.e. co-expression analysis) has been used to extract biologically meaningful information from different data sets [4, 5], but has rarely been used on proteomics data with the exception of the work of Gibbs et al 2013 [6]. Here, we have used different topologically-based strategies to divide the main list of identified proteins into different modules by first using a Weighted Gene Co-expression Network Analysis (WGCNA) developed by the Horvath group [7, 8]. These modules were, in turn, separated and broken down into clusters and sub-clusters using MCODE [9] and hierarchical clustering was applied to the protein expression patterns. As these approaches rely solely on expression profiles without priory functional knowledge, we then employed several knowledge-based tools to both verify and assign biological relevance to the observed sub-clusters of data. We compared the protein-protein interaction networks generated de novo using WGCNA against predicted networks for the same subset of proteins using STRING [10, 11] which clearly shows a significant overlap between the WGCNA analysis of the proteomics data and STRING. In this study, we present a protein co-expression analysis of the dataset for glioblastoma multiforme previously acquired and published by Deighton et al. [12]. Our new findings support these previous observations. In addition to the previous findings from this study, we have highlighted three major modules of co-expressed proteins that are associated with specific membrane structures; the mitochondria, the endoplasmic reticulum (ER), and vesicle membranes. We show that within these modules, we can generate protein networks, which are similar to protein interaction networks predicted by data-mining from the literature without using an immunoprecipitation approach or native gel separation. In addition to a major disruption of the Electron Transfer Chain (ETC) observed in the tumour samples, we show that the proteins composing each of the main ETC complexes (Complex I to IV) are mostly co-expressed but that each of the complexes are affected differently. In the ER, the unfolded protein response as well as the oxidative stress pathway are up-regulated. Furthermore, two isoforms of Pyruvate kinase (PKM) (isoform M1 and isoform M2) were differentially co-expressed with a high PKM2/PKM1 ratio supporting aerobic glycolysis (a hallmark feature of cancer) at the expense of oxidative phosphorylation (most likely inefficient due to the disruption of the ETC). While the M2 isoform seems poorly co-expressed with other proteins, the M1 isoform is part of a more defined network which is involved in ion transport, cellular response to insulin stimulus, glutamate secretion as well as syntaxin binding. In this study, we show that the use of a weighted protein co-expression analysis provides a level of information about protein interaction networks which is not possible to obtain using a standard data analysis approaches. Methods The data used were the quantitative proteomics data from a glioblastoma multiforme study conducted by Deighton et al. [12]. All protein identities are publically available through PRIDE (http://www.ebi.ac.uk/pride) PRD000620 and the label-free quantitation output presented in S1 Table. In that study, 6 controls and 6 tumour samples were used. A mitochondrial extraction was performed, the samples were trypsinised, followed by a shotgun proteomics analysis. The quantitative analysis was performed using Progenesis (Non Linear Dynamics, UK). The MS data for this present study were searched against a human RefSeq database (34 284 sequences) using Mascot (version 2.4.1), Matrix Sciences), with a significance threshold p < 0.05 in addition to peptide ion score cut-off of 20. Each analysed protein needed at least 2 identified peptides. Conversion from RefSeq to gene symbol was performed using the biological DataBase network (bioDBnet) [13]. Label-free intensity data were ArcsinH transformed prior to analysis as log transform of 0 is not ideal. A simple trait matrix was built as follows; the parameter “state” was a single number defined as “1” for disease and “0” for control. The R WGCNA package [7] was used to perform the analysis of the data set. The Topological Overlap Matrix (TOM) was created using a cut height of 0.25 and a minimum module size of 30. The analysis produced five modules, identified with different colours (‘brown’, ‘turquoise’, ‘blue’, ‘grey’ and ‘yellow’) with the ‘grey’ module containing all proteins that were not sorted to any of the other modules shown in Table 1. A hard threshold approach was used for comparison purposes where: aij = corr(proti, protj)β The correlation aij between the ArcsinH intensity of the protein proti and protj is measured. The factor β is a thresholding parameter, for hard thresholding, we used a β of 1 only for validation purposes (for FDR evaluation by comparing the same dataset against a randomised one). For the remainder of the study, we used a β of 10, justified from S1 Fig which corresponds to the lowest value showing a good scale-free topology. 10.1371/journal.pone.0161828.t001Table 1 GO term enrichment assignments for the five main clusters. Module GO GO Term Description P-value FDR   Category       q-value Blue Process GO:0022900 electron transport chain 1.04E-26 2.98E-23   Function GO:0008137 NADH dehydrogenase (ubiquinone) activity 3.74E-15 4.85E-12   Component GO:0044455 mitochondrial membrane part 2.93E-25 2.10E-22 Brown Process GO:0006397 mRNA processing 4.28E-08 2.44E-04   Function GO:0003676 nucleic acid binding 7.16E-11 9.29E-08   Component GO:0005783 endoplasmic reticulum 1.06E-06 7.62E-04 Turquoise Process GO:0007268 synaptic transmission 4.89E-08 2.79E-04   Function GO:0030276 clathrin binding 5.95E-05 3.86E-02   Component GO:0097458 neuron part 4.80E-14 3.45E-11 Yellow Process   NO ENRICHMENT FOUND       Function   NO ENRICHMENT FOUND       Component GO:0005739 mitochondrion 4.37E-10 3.14E-07 Grey Process   NO ENRICHMENT FOUND       Function   NO ENRICHMENT FOUND       Component   NO ENRICHMENT FOUND     Selected groups of proteins were exported to the online Gene Ontology enRIchment anaLysis and visuaLizAtion tool (GOrilla) [14] with the gene names of each individual module used as a target set and those of the remaining modules used as a background set for the enrichment. The network data were exported to Cytoscape v3.2.1 [15], with the corresponding WGCNA function [8] where they were visualised. “Hub” clusters were defined using MCODE v1.4.1 [9]. Default suggested parameters were used. Protein interaction networks for each module were generated from co-expression similarity using WGCNA. The same set of proteins was then clustered into a network using STRING v10 using specific confidence parameters presented in table [10, 11]. The two generated networks were then compared using “Network Analysis Tool” (NeAT) [16] with default parameters, randomisation was based on the Erdos-Renyi method. The WGCNA networks were used as the ‘Query’ networks and those from STRING as the ‘Reference’ networks. For both network types, different cut-off points were tested and are described in Table 2. 10.1371/journal.pone.0161828.t002Table 2 Different cut-off combinations for comparing between networks generated using WGCNA and STRING prediction.   WGCNA STRING     Module name1 Cut-off2 Cut-off3 P-value4 Jaccard5 Brown 0.3 0.4 1.40E-32 0.0871     0.7 5.50E-20 0.081   0.2 0.4 6.10E-63 0.1029     0.7 4.00E-53 0.0693   0.1 0.4 3.50E-47 0.0681     0.7 1.40E-40 0.0377 Turquoise 0.3 0.4 1.20E-12 0.048     0.7 2.10E-11 0.0289   0.2 0.4 1.70E-15 0.0455     0.7 1.60E-13 0.024   0.1 0.4 2.30E-14 0.0427     0.7 4.00E-11 0.0206 Blue 0.3 0.4 0.00E+00 0.2209     0.7 0.00E+00 0.1767   0.2 0.4 6.00E-289 0.1631     0.7 2.40E-221 0.1137   0.1 0.4 1.20E-154 0.1206     0.7 5.30E-141 0.078 1) Different Modules extracted using WGCNA. 2) Threshold values for pair-wise protein co-expression (Pearson correlation^10). 3) Confidence score cut off to generate protein-protein network in STRING (http://string-db.org/). 4) p-value calculated for the overlap of the 2 protein networks (from WGCNA and from STRING) using NeAT. 5) Jaccard similarity coefficient: size of the intersection divided by the size of the union of the sample sets Hub proteins were associated with proteins having a number of interactions which was two-fold greater than the standard deviation above the average number of interactions found in a specific module (i.e. z-score above 2). A hierarchical clustering of protein intensity was applied on the largest clusters generated from MCODE for each of the module networks using R version 3.1 GPLOT package and Ward’s method. Sub-clusters were then generated and their nature analysed using ToppCluster [17] for comparative analysis. Additionally, the clusters were analysed using the database of differentially expressed proteins in human cancers, dbDEPC 2.0 [18]. Fig 1 illustrates the overall data analysis platform used for this work and the number of proteins associated to each of the modules. 10.1371/journal.pone.0161828.g001Fig 1 Workflow illustrating the analysis performed. On the left, the step-wise protein list fragmentation is illustrated. On the right, the different bioinformatics tools used are described. The number of proteins per group are presented and coloured using the WGCNA colour coding. The names associated to the sub-clusters are illustrated on the right side of each sub-Clusters. Results Defining the different modules It has been previously described that peptide and protein interaction networks possess a scale-free network topology similar to those found in gene co-expression networks [6, 19]. The glioblastoma dataset from Deighton et al. [12] is composed of a set of 799 proteins identified with at least two peptides (presented in S1 Table). As shown in S1 Fig, a power of β = 10 has been extracted from the original data and used for further analysis. The overall data analysis strategy used in this work is presented in Fig 1. A series of data analysis tools, based on either network topology characteristics or literature knowledge was used to cluster groups of proteins. One possible concern with the use of the Deighton et al (10) dataset for network analysis is its rather small size (a 2 group comparison with 6 replicates only) where normally datasets of at least 25–30 samples are commonly used for co-expression analysis (4). We have estimated the false discovery rate (FDR) by using a permutation approach as described elsewhere [20, 21]. Hierarchical clustering of the pair wise correlation coefficient was evaluated first using a thresholding parameter of β = 1, shown in Fig 2A each row and column are represents proteins, the colour purple is associated with clusters having a high correlation coefficient, and white is associated to a high anti-correlation coefficient. Protein intensities were also randomly permutated and the same clustering method was used once again (shown in Fig 2B). As expected, a significant decrease in the level of correlation is observed. The plot of the distribution of the correlation coefficients for both datasets (the direct dataset and the randomised one) is shown in Fig 2C. The randomised dataset is centred around 0 (blue) while the normal dataset (turquoise) exhibits two distributions roughly centred on 0.5 and -0.5 associate to either correlated or anti-correlated pairs or proteins, respectively. We have evaluated the FDR for different correlation coefficients and a FDR of 5% was calculated for a correlation coefficient of 0.754 and above and values of -0.728 and less for meaningful anti-correlation (Fig 2D). A similar calculation was performed using a β = 10 (Fig 2E), a FDR of 5% was found for value an aij of 0.0993 and above (Fig 2F). 10.1371/journal.pone.0161828.g002Fig 2 Evaluation of the confidence in the protein pair-wise measured correlation coefficient. Fig 2A hierarchical clustering of the protein pair-wise correlation coefficient for a β = 1; in Fig 2B, correlation coefficient evaluated after intensity randomisation for a β = 1. Fig 2C Distribution of the correlation coefficient from direct correlation in turquoise (extracted from Fig 2A) or after intensity randomisation in blue (extracted from Fig 2B). Fig 2D is the ratio false positive hits versus measurements obtained in the dataset. A correlation coefficient of 0.754 and above indicates positive correlation while -0.728 and less for negative correlation corresponds to a ratio of false positive below 5%. In Fig 2E the same measurements as in 2C in the case of a soft threshold β = 10. In Fig 2F, a false positive rate equal to or below 5% corresponds to a value of 0.0993 and above. The initial analysis was based on the WGCNA package for R [7, 8]. Fig 3 shows the Topological Overlap Matrix (TOM) plot applied to the dataset from Deighton et al. [12]. Each row and column represents proteins. The colours red and yellow indicate the high and low weighted correlation values, respectively, and are assigned by the TOM-based dissimilarity between each protein co-expression level. Each of the five modules, represented by the coloured bar on the top and left of the matrix (blue, yellow, brown, turquoise, and grey), are associated to set of proteins sharing a high value for co-expression level. Only 18 proteins did not cluster into a module and were allocated to the grey module. The turquoise module is the largest one, containing 272 proteins, followed by the blue module with 256 proteins, the brown module with 177 proteins, and the yellow module with 76 proteins (illustrated in Figs 1 and 3). 10.1371/journal.pone.0161828.g003Fig 3 Clustering of the proteomic label-free analysis of glioblastoma multiforme. The data shows five major clusters. The clustering heatmap was created using a soft thresholding of β = 10 on the entire proteomics dataset. Data clustering and module membership generation are described in Materials and Methods. The scale ranges from yellow to red, with yellow demonstrating low topological overlap and red representing high topological overlap. Similar to gene expression patterns, proteins within a given module are co-expressed with higher correlation than with proteins from different modules. We then asked if those proteins which are part of the same cluster share some similarities in terms of biological function. To address this, we performed a Gene Ontology (GO) enrichment analysis for each module. The proteins from each cluster were analysed using Gorilla. Table 1 shows the results of the GO enrichment analysis. Both, the yellow and the grey modules do not show any major functional enrichment. On the other hand, the three other modules clearly show significant GO terminology enrichment. They represent primarily three different components, the blue module being associated with the mitochondrial membrane part (q-value of 2.10e-22), the brown module being associated with the ER (q-value of 7.62e-4) and the turquoise module being associated with the neuronal part/membrane vesicles (q-value of 3.45e-11), suggesting that the proteins from a given cellular location have similar function and display a higher degree of co-expression. The five main protein modules were then correlated to phenotype data (trait matrix) to highlight possible trends. This step was performed in order to identify possible links between the clusters of proteins and higher-level information. The phenotype information is presented in S2 Table. Fig 4 is a heatmap showing the correlation between the five modules and three different traits. The traits used for correlation were general traits (age and gender of patient taken from Deighton et al. [12] and state (extracted from S2 Table). The numbers within the heatmap squares show Pearson correlation coefficients quantifying the correlation between the modules and the phenotype traits. The numbers in brackets are the respective p-values and corrected p-values, respectively. 10.1371/journal.pone.0161828.g004Fig 4 Data-trait correlation between the first principal component (Eigengene) of each module (y-axis) and the clinical traits (x-axis). All positive correlations are shown in red and the negative correlations are shown in blue. The correlation coefficients between cells are shown and p-values are displayed within brackets below the correlation coefficient itself. The modules with the lowest and highest significant p-values are the brown, the turquoise, and the blue module. In the two modules having the lowest functional information (yellow and grey), no significant correlations were found. No significant correlation was found with Age and Gender for any of the modules. The brown module (ER) is anti-correlated to both the blue (mitochondrial membrane) and turquoise modules (membrane vesicles). The brown module shows a high level of correlation with the state (i.e. control = 0, tumour = 1) (r = 0.93). These results indicate that the proteins within the brown module are mostly up-regulated in glioblastoma tumour samples. The turquoise module shows strong anti-correlation with state whereas the blue module shows a similar, but less pronounced anti-correlation with the state. Validating the protein networks generated from WGCNA High values of co-expression between two proteins may be predictive of protein- protein interactions. In order to assess the validity of the interactions generated with the presented analysis, the networks that were generated using WGCNA were compared with networks of known interactions obtained from STRING for the same protein dataset. The statistical comparison between the two different approaches was performed using the Network Analysis Tool NeAT (see Materials and Methods section for detailed description). Table 2 shows the results of the comparison between WGCNA and STRING outputs for different threshold values. The Jaccard coefficient was used to determine the similarity between two sample sets. A combination of different parameter thresholds for both the WGCNA analysis and STRING was tested in order to optimise the best overlap of the two independent approaches to predict protein-protein interactions (illustrated in Table 2). The parameter threshold for WGCNA is the minimal threshold Pearson’s correlation coefficient of co-expressed paired proteins. The STRING score is defined as the confidence in the interaction between two protein nodes. Different combinations of parameters have been used and their effect on network overlap (Jaccard coefficient) and prediction quality (shown by p-value) is shown in Table 2. In addition, we have evaluated the similarity between the STRING output and a randomised pairing (using the same node but having randomised the same number of edges predicted by WGCNA for a given cut-off). The chosen combination of cut-off was based on several factors including minimal p-value and Jaccard score as shown in Table 2 and the highest difference in p-value obtained between the WGCNA and a randomised similar data set against STRING. The best threshold combination appears to be 0.3 (associated to a FDR of 0.5% and less) for WGCNA and 0.4 for the confidence score generated by STRING (expressed as 0.3/0.4 pair in the text), which gives the higher Jaccard value for the blue and turquoise modules. On the other hand, for the brown module the optimal threshold combination seems to be 0.2 for WGCNA(which is associated to a FDR of 1% and less) and 0.4 for the confidence score in STRING (0.2/0.4 pair). However, more pronounced differences between WGCNA prediction and a random dataset were observed with a cut-off of 0.2/0.4 for the brown and turquoise modules. We have calculated a p-value of 6e-289 for the blue module against STRING, whereas a randomised dataset under the same condition had a p-value of 1.2e-13. In the brown module, we measured a p-value of 6.1e-63, whilst a random dataset generated a p-value of 9.1e-9. In the turquoise module we observed 1.2e-12 whilst a randomised dataset generated a p-value of 3.3e-7. Although low p-values were observed with randomised datasets using STRING, they were largely different from what was predicted with the real dataset. Those highly significant p-values for the random datasets are a consequence of a high ratio of number of edges versus nodes. The higher this ratio, the less of an effect the edge position randomisation has on the predicted network. In general, the number of interactions predicted by WGCNA was significantly higher compared to what has been reported in STRING. The resulting outcomes are densely interconnected protein networks. In order to reduce the dimensions of the three main large modules identified and extract more subtle information regarding their nature, we used other topological based tools. MCODE [9], a tool that identifies highly interconnected nodes within a complex network, was used to isolate smaller groups of proteins (which will be referred to as a “cluster”) within each of the three main modules (blue, brown and turquoise module) and identify key highly connected proteins (i.e. Hubs). For each module, a major dense cluster was identified and several minor clusters were also generated (Figs 5, 6 and 7). 10.1371/journal.pone.0161828.g005Fig 5 Visualisation of the brown module using a network generated in Cytoscape. The global network is shown in Fig 5A. The main large cluster identified by the MCODE application, being coloured in green and the most interconnected (‘hub’) proteins shown in purple. This main cluster extracted from the brown module is shown in Fig 5B and other secondary clusters identified by MCODE are also shown (Fig 5C, 5E and 5F). In Fig 5B, 5C, 5E and 5F, proteins highlighted in red are associated with a defined GO term assigned by GOrilla. The main cluster as shown in Fig 5B was analysed using a hierarchical clustering approach based on protein intensity across the tumour (Tu) and the control (Ct) samples and is shown in Fig 5D. Five sub-clusters were identified and further analysed using Toppcluster. 10.1371/journal.pone.0161828.g006Fig 6 Visualisation of the blue module using a network generated in Cytoscape. The global network is shown in Fig 6A with the main large cluster, identified by the MCODE application, being coloured in green. The main cluster extracted from this module is shown in Fig 6B and other secondary clusters identified by MCODE are also shown in Fig 6D and 6E. In Fig 6B, 6D and 6E, proteins highlighted in red are parts of a defined GO term according to GOrilla. The main cluster as shown in Fig 6B was analysed using a hierarchical clustering approach based on protein intensity across the tumour (Tu) and the control (Ct) samples and is shown in Fig 6C. Four sub-clusters were identified. Distribution of the proteins from the five complexes across the different sub-clusters is shown in Fig 6F. 10.1371/journal.pone.0161828.g007Fig 7 Visualisation of the turquoise module using a network generated in Cytoscape. The global network is shown in Fig 7A which contains a large (Fig 7B) and a small (Fig 7C) network. In Fig 7A, the main large cluster, identified by the MCODE application, is coloured in green and the most interconnected (‘hub’) proteins are visualized in purple. The main cluster extracted from this module is shown in Fig 7B and a smaller secondary cluster identified by MCODE is shown in Fig 7D. In Fig 7B, proteins highlighted in red are part of a described biological function according to GOrilla, no specific enrichment has been found for cluster 7D. The main cluster as shown in Fig 7B was analysed using a hierarchical clustering approach based on protein intensity across the tumour (Tu) and the control (Ct) samples and is shown in Fig 7C. Partitioning the modules; cluster and sub-clusteranalysis Figs 5–7 show network representations of each module. Figs 5A, 6A and 7A are the global networks for the brown, blue and turquoise modules, respectively. Coloured in purple are the highly connected hub proteins for each module (the blue module has no identified protein hub). In the global network, proteins that belong to the first cluster generated by MCODE (as presented in Figs 5B, 6B and 7B, respectively) are coloured in green in Figs 5A, 6A and 7A. The three main clusters in Figs 5B, 6B and 7B were still densely interconnected, with an overlap of 59 out of 177 proteins for the brown module (33%), 129 out of 256 proteins for the blue module (50%) and 94 out of 272 proteins for the turquoise module (35%). As those main clusters represent an important part of each module, they are mostly an enriched version in terms of function and protein localisation of each module. In addition to those main clusters, several smaller clusters were as well identified and are described below. Some smaller clusters for the three modules (see Figs 5C, 5E, 5F, 6D and 6E) showed significant GO term enrichment based on GOrilla (node coloured in red). The main terms describing the most significant enrichment varied in most cases, but were mainly found to be well described by cellular component and biological process GO terms. In order to identify subtle variation within each main cluster, we applied hierarchical clustering on the protein intensity for each of the main clusters (i.e. the large clusters in Figs 5B, 6B and 7B), which highlighted some possible sub-clustering. These heatmaps are shown in Fig 5D for the brown module, Fig 6C for the blue module and Fig 7C for the turquoise module. Each of those clusters and sub-clusters were analysed using the comparative tool Toppcluster. Description of the ER (Brown) Clusters and Sub-clusters From the initial 177 proteins composing this module, 154 proteins had at least one WGCNA co-expression parameter above 0.2 the threshold value we used for that clusters/sub-clusters. From those 154 proteins, 87 proteins are sub-grouped into four clusters. The 67 proteins that did not associate with any cluster were also not assigned any major biological function. A group of seven proteins were identified as “hub” proteins, i.e. proteins which are highly interconnected (shown in purple in Fig 5A and 5B). Those proteins are CAT, PDIA6, CALU, SCP2, TMX1, MYH9, and VIM, of which PDIA6 and TMX1 are involved in disulphide isomerase activity and SCP2, CAT and VIM in peroxisome signalling. In Fig 5B, the proteins highlighted in red are associated with cell compartment GO terminology “ER”. In Fig 5E, the term used to describe the proteins in red was the GO Cellular Component “cell cortex part” (ANK1, SLC2A1, SLC4A1, SPTA1, SPTB) primarily involved with cytoskeletal protein binding, while the subgroup ANK1, SPTA1, SPTB is also related to biological processes associated with the tetrapyrrole and porphyrin-containing compound biosynthetic process. RHD is the only protein not associated with the cell cortex part, but is linked to the plasma membrane along with the other proteins in this cluster. For Fig 5C, the “cell cortex” and “cortical cytoskeleton” are over-represented in the cellular compartment GO terminology (EZR, FLNA, MAPRE1), the proteins EZR, FLNA, PFN1, TLN1 are involved in maintenance of protein location while the large group of proteins containing EEF2, EIF4A1, EZR, FLNA, HSPB1, KPNB1, MAPRE1, PFN1, RPL4, SERPINH1 share the molecular function “poly(A) RNA binding”. The term used to describe the proteins in red was the GO Function term “nucleic acid binding”. One can notice that two proteins, were not characterised by the prevalent GO term (i.e., protein TAGLN2 and TLN1 both in blue in Fig 5C). These two proteins are associated with actin binding. However, TAGLN2 is a poorly characterised protein without a determined function. Fig 5F shows proteins involved in poly(A) RNA binding (APEX1, FUS, HMGB2, HNRNPA2B1, HNRNPA3, PARP1), and most of the proteins found in this cluster are primarily located in the nucleoplasm (Cellular compartment). They are: APEX1, FUS, H2AFY, HMGB2, HNRNPA2B1, HNRNPA3 and PARP1. The term used to describe the proteins in red was the GO Function “DNA binding.” The main brown cluster illustrated in Fig 5B contains proteins enriched in the ER part, with proteins involved in ER stress. Some specific domain enrichments were found, such as Thioredoxin-like fold (EEF1G, P4HB, PDIA3, PDIA4, PDIA6, PRDX4, TMX1) and ER targets (CALR, HSP90B1, HSPA5, P4HB, PDIA4, PDIA6, PRKCSH). The three main pathways represented in these data are: mRNA processing (HNRNPA1, HNRNPH1, HNRNPK, HNRNPU, NONO, PTBP1, SFPQ, TMED10) Protein processing in ER CALR, CANX, CKAP4, DDOST, HSP90B1, HSPA5, P4HB, PDIA3, PDIA4, PDIA6, PRKCSH, RPN1, STT3A Calnexin/calreticulin cycle (CALR, CANX, PDIA3, PRKCSH) The main cluster shown in Fig 5B was separated into five sub-clusters, Br1a to Br1e, of 31,6,8,2 and 12 proteins, respectively (shown in Fig 5D). Mainly the two sub-clusters Br1a and Br1e generate functional information. The ER parts are found in the Br1a and Br1e sub-cluster with CALU, CKAP4, DDOST, PDIA4, PRKCSH, RPN1, STT3A, TMED10, TMX1 for Br1a. Proteins associated with mRNA processing were found in the sub-cluster Br1a (HNRNPA1, HNRNPH1, HNRNPK, HNRNPU, NONO, PTBP1, SFPQ) and nucleoplasm (HNRNPA1, HNRNPH1, HNRNPK, HNRNPU, LMNB1, NONO, PTBP1, SFPQ, XRCC5, XRCC6). Interestingly, the pair XRCC5 and XRCC6 were identified, which play a major role in the non-homologous end joining (NHEJ) pathway [22]. Proteins associated to the cytoplasmic membrane-bound vesicles are unique to the Br1e sub-cluster (CALR, CANX, HSP90B1, HSPA5, P4HB, PDIA3 and PPIB). In addition, unique proteins associated to calcium ion binding such as ANXA1, CALR, CANX, HSP90B1, HSPA5 are found in the sub-cluster Br1e which also contains unique proteins involved in response to ER (CALR, HSP90B1, HSPA5, P4HB, PDIA3). Proteins found in this last subgroup (CALR, HSP90B1 and especially HSPA5) are well known to be involved in the activation of signalling protein activity and unfolded protein response (UPR). Description of the Mitochondrial (Blue) Clusters and Sub-clusters From the initial 256 proteins composing this module, 214 proteins have a WGCNA co-expression parameter above 0.2. From those 214 proteins, a group of 147 proteins can be sub-divided into three clusters. The 67 proteins that are not part of any major cluster are not significantly co-expressed, however, they did share some biological function such as fibrinogen Complex FGA, FGB, FGG, FN1 and are parts of the Integrin signalling linked to the MAP kinase pathway by recruiting Grb2 to the FADK1/SRC activation complex. In contrast to the majority of protein in this module, this small subset of proteins is up-regulated in the tumour samples. Fig 6A shows the main module Blue while Fig 6B is associated to the main blue cluster generated by MCODE. The main cluster in Fig 6B is largely composed of proteins involved in the ‘Electron Transport Chain’. The small cluster (Fig 6E) (HPX, ORM1, SERPINA1, TF) is associated to the cellular component “extracellular space”. The larger network (Fig 6D) has no significant functional enrichment according to GOrilla, although STRING significantly associates (p-value of 2.059e-5) all of its proteins to the extracellular region, except for GLS, CKMT1B and GDAP1L1. This module also contains three proteins having a thioredoxin fold domain (GDAP1L1, PRDX1, and PRDX6). The cluster in Fig 6D contains several proteins which have been associated with a variety of different cancer types including breast cancer which are WDR1, PRDX1, PRDX6 [23], and HSP90AB1 [24], hepatocellular carcinoma HSP90AB1, PRDX1, PRDX6 [25], gastric cancer WDR1, HSPB90AB1[26], cervical cancer PRDX1, HSP90AB1 [27], thyroid cancer HSP90AB1, PRDX6 [28], prostate cancer HSP90AB [29], and colorectal cancer WDR1 [30]. The main blue cluster from Fig 6B can be separated into four sub-clusters (shown in Fig 6C), Bl1a to Bl1d, consisting of 52, 33, 16, and 28 proteins, respectively. The smallest cluster (Bl1c) had little biological information deduced. The blue cluster 1 (Fig 6B) is mainly composed of proteins associated to the mitochondrial respiratory chain which, in turn, comprises Complexes I to V. These different complexes are co-expressed slightly differently and are therefore distributed across the four sub-clusters (Fig 6F). The Complex I proteins are mainly found in sub-clusters Bl1a (18 proteins out of 26 identified in this study), proteins from Complex III are mainly found in Bl1b (5 out of 7 proteins identified in this study), Complex IV is found across Bl1b and Bl1d while Complex V is distributed between sub-clusters Bl1a and Bl1d. Only two proteins from Complex II were identified (SDHA and SDHB) that were not found to be part of the same sub-cluster. Description of the Neuronal (Turquoise) Clusters and Sub-clusters From the initial 272 proteins composing this module, 198 proteins have a WGCNA co-expression parameter above 0.3. From those 198 proteins, a group of 118 proteins is involved in two clusters. The 80 proteins not part of any major clusters although not significantly co-expressed shared some biological function such as fatty acid beta oxidation (ACAA2, ACADS, ACADVL, DECR1, ECI2, HADHA, HADHB), gluconeogenesis (ALDOA, ENO2, PGAM1, SLC25A1, SLC25A13), and glucose metabolism (ALDOA, ENO2, PGAM1, PKM1, SLC25A1, SLC25A13). The main turquoise module (Fig 7A) generated both a large and a small network while using a threshold of 0.3 for the WGCNA coefficient. The large main network (Fig 7A) is composed of proteins involved in different “membrane vesicles” structures whilst the small network is mostly related to the myelin sheath (CNP, MBP, PLP1, SIRT2). The overall module containing the “neuronal part” is associated with proteins assigned the terms endocytic vesicles and cytoplasmic membrane-bounded vesicles with some ATPase and GTPase activity; furthermore, a subgroup of proteins is associated to glial cell differentiation (CNP, GAP43, MBP, PLP1, TPPP). A group of proteins which are highly connected (i.e. Hub proteins; VSNL1, YWHAG, ATP6V1E1, ATP6V0A1, GNAZ, SYT1, DNM1, ATP6V1A, STXBP1) were identified. Three ATPase H+ transporting lysosomal units were found to be quite interconnected and are involved in several different functions (e.g. ATP hydrolysis coupled proton transport and ferric ion transport). Three proteins combined with SYT1, DNM1 and STXBP1 are part of the synaptic vesicle cycle. In Fig 7C, the main cluster in 7B has been divided into five sub-clusters of 24, 23, 9, 31, and 7 proteins, respectively, (Tu1a to Tu1e). According to Toppcluster, mainly three sub-clusters show biological enrichments which are Tu1a, Tu1b and Tu1d. The sub-cluster Tu1a is rich in proteins involved in ion/cation transport (ANK2, ATP1B1, ATP6V0A1, ATP6V1A, ATP6V1B2, CAMK2A, CNTN1, NSF, SNAP25, STX1A, STX1B, SYT1, THY1, and YWHAZ). In addition, proteins from cluster Tu1a have molecular functions associated to SNARE binding (NSF, SNAP25, STX1A, STX1B, and SYT1). A group of three proteins from Tu1a is involved in regulation of mitochondrial membrane permeability (CAMK2A, YWHAG, and YWHAZ). The sub-cluster Tu1b is rich in proteins involved in pathways associated to coated vesicle membrane and clathrin-coated vesicle (AP2A1, AP2M1, DNAJC5, SNAP91, and VAMP2), while Tu1d is mainly composed of proteins involved in synaptic vesicle endocytosis and synaptic vesicle recycling (AMPH, RAB3A, SH3GL2, SNCA, SYNJ1, and SYP). The data presented in Fig 7D showed no strong functional enrichment after being analysed by GOrilla, although according to Toppcluster and STRING the following proteins are associated to protein targeting to ER as biological process: RPL18, RPL7A, RPLP0, and RPN2. Additionally, STRING identified several proteins as parts of membrane-bound vesicle from this cluster (CAMKV, CAMK2G, PALM, RPLP0, RAP2A, RPL7A, GSTK1, TUBB2A, AP2A2, DPYSL2, FTH1, PFKP, DPP6, and AK2). Pyruvate kinase isoforms co-expression network The two isoforms of PKM (PKM1 and PKM2) were identified (Fig 8). While PKM1 was found to be down-regulated and associated to the turquoise module, the PKM2 isoform was up-regulated and associated to the brown module. The direct co-expressed proteins for each pyruvate kinase protein isoform are illustrated in Fig 8. Twenty-nine proteins were found to be co-expressed with PKM1 while only three showed co-expression with PKM2 in this study. 10.1371/journal.pone.0161828.g008Fig 8 Pyruvate kinase isoform M1 (Left) and M2 (Right) and their respective co-expression networks (direct interactors only). Nodes in dark blue are the 2 PKM proteins, in pink are proteins defined as HUB proteins from the turquoise Module. The proteins in green are proteins associated to the larger sub-cluster presented in Fig 7. Regarding PKM1, the largest group of proteins exhibiting direct co-expression are those related to the synaptic vesicle cycle (KEGG Pathway): ATP6V0A1, ATP6V0D1, ATP6V1A, ATP6V1B2, ATP6V1E1, CPLX2, DNM1, NSF, RAB3A, SNAP25, STXBP1, SYT1, and VAMP2. Several other metabolite-associated groups of proteins were identified, such as proteins related to cellular response to insulin stimulus (YWHAG, VAMP2, ATP6V0A1, ATP6V0D1, ATP6V1A, ATP6V1B2, ATP6V1E1, and GOT1), glutamate secretion (VAMP2, SYT1, SNAP25, STXBP1, and RAB3A), and to syntaxin binding (CPLX2 NAPB NSF SNAP25 STXBP1, and VAMP2).Some of the highly correlated expression profile proteins with PKM1 include guanine nucleotide binding protein (GNAO1) and syntaxin binding protein 1 (STXBP1) for which no known direct interaction has been reported yet. In a similar manner, the protein cell adhesion molecule 3, CADM3 involve in the calcium-independent cell-cell adhesion molecules is as well highly correlated with PKM1 but no known relation between CADM3 and PKM1 has been previously reported. Both cases merit to be explored by studying the role of both PKM in specific tissues, in this cases, PKM1 role in the synaptic vesicle. For the PKM2 cluster, no significant term enrichment was found. Discussion In the present study, we used a combination of different analytical methods to characterise protein co-expression measured from a quantitative proteomics analysis. The main method (WGCNA) was applied to the Deighton et al dataset [12] and allowed for the subgrouping of all proteins into five main modules, of these modules, three are associated with membrane-based organelles. The soft threshold power used in this study (β = 10) is in the same order of magnitude as that used in previous work [31]. This approach has resulted in identification and sub-grouping of proteins by their distinct features into three different cellular locations. Although the dataset is of a modest size (a total of 12 experiments), we have shown that it was possible to extract valid and meaningful information. We evaluated the FDR for different correlation coefficient thresholds using the same dataset but with the position of each intensity for a given protein being randomised. The threshold values selected to generate the different networks in Figs 5–7 have a FDR of between 0.5% and 1% which is quite conservative. The 3 major modules clearly show significant enrichments thus supporting the validity of the approach even on small datasets. The networks generated using WGCNA have been compared to the knowledge-based method STRING and shows the overlap between the two independent methods to be significant. Thus, we have shown that the use of WGCNA to generate protein networks de novo without the need for an immunoprecipitation-based approach. These networks could not have been generated with the initial type of analysis used in Deighton et al [12] The over-represented GO term to describe functional enrichment was mainly the cellular component with the blue module’s proteins being significantly localised in the mitochondria. The brown module was enriched in ER proteins and the turquoise module enriched in various types of vesicle membranes. In addition, these abundance of the proteins in these modules correlated to traits which included the relative increase, or decrease of expression in cancer tissue (Fig 4). The brown module (ER) proteins correlate with proteins that are up-regulated in tumour samples, while the blue module (mitochondrial membrane part) and turquoise module (membrane vesicles) both correlate with proteins that are down- regulated in tumour samples. The original proteomic analysis reported by Deighton et al. [12] was based on a mitochondrial fraction enrichment. However, in this current study we have identified several proteins from other membrane-based organelles such as the ER and vesicular membranes. Although these membrane fractions share similar physical properties to the mitochondrial fractions and could introduce complexity to the samples, their identified interaction networks reveal the broader of the many effects of glioblastoma multiforme. In addition, the concomitant enrichment of ER in the mitochondrial fraction might be a result of those two organelles being interconnected through mitochondria-associated membranes (MAM) [32] a finding that may provide a deeper understanding of intra-cellular organelle coordination during tumorigenesis. Despite the use of a soft threshold β = 10 to generate the different networks, these networks were significantly denser than what was predicted by STRING. Although the overlap between STRING and WGCNA was found to range between 5 and 22%, the calculated p-values clearly support that the observed networks were not simply due to chance (p-values between 1.2e-12 to 6e-289). One observation, also reported in Deighton et al. [12] is that the electron transfer chain (ETC) is significantly down-regulated in cancer cells (part of the blue module). Proteins from the major complexes of the ETC were identified in this study and were found to be mostly down-regulated. This observation was supported by electron microscopy showing that the inner membrane of the mitochondria is severely disrupted [12]. However, in this manuscript we have found that the different complexes were marginally co-expressed in different sub-clusters especially for Complex I (70% of Complex I proteins were found in sub-cluster Bl1a) and Complex III (70% of Complex III proteins found in sub-cluster Bl1b) which suggests that these two complexes are not affected in the same way, with Complex I proteins being slightly more down-regulated than the proteins from Complex III. A similar observation on the different effects on complexes of the ETC has been made on mitochondrial fractions isolated from a transgenic mouse model [33]. A disruption of the electron transfer chain and oxidative phosphorylation could potentially lead to elevated ROS generation [34]. Several proteins involved in the oxidative damage response were also found to be up-regulated such as catalase, superoxide dismutase 2, peroxiredoxin 1, 4 and 6. Several key proteins involved in the “ER stress response” or the “unfolded protein response” (UPR) were found to be up-regulated. The disruption of the ETC and the up-regulation of several proteins involved in oxidative stress support a link with cellular events such as protein oxidation and protein folding. Oxidative stress and ROS generation are important components of the ER stress response. The major enzymatic components of ROS production during UPR induction are protein disulfide isomerase (PDIA4 was found up-regulated in this study); ER proteins involved in stress response were found significantly co-expressed (CALR, HSP90B1, HSPA5, P4HB, and PDIA3) specifically in the sub-cluster Br1e. Most of these proteins were also found up-regulated during oxygen and glucose deprivation for 18h [35] which supports an integrated cellular survival response. Furthermore, mitochondrial HSP90 has been reported to play an important role in controlling core metabolic processes by stabilising Complex II of the ETC and allowing cellular respiration to continue under compromised conditions, contributing to tumorigenesis [36]. Cells under normal conditions have a basal level of ROS, which is intrinsic to signalling mechanisms. However, an increase of ROS levels is observed upon exposure to specific stress such as cytotoxic reagents, irradiation, and environmental pollutants and during some specific enzymatic reactions such as: mitochondrial respiratory chain reactions, activity of glucose oxidase, amino acid oxidase, xanthine oxidase, and NADP/NADPH oxidase). Triggering of the unfolded protein response (UPR) consequential to the exposure to oxidative stress is most likely a mechanism to preserve both cell function and survival. On the other hand, continuous oxidative stress and protein misfolding induce apoptotic pathways and play crucial roles in the pathogenesis of multiple human diseases including diabetes, atherosclerosis, and neurodegenerative diseases. HSPA5 (also known as GRP78, Bip) is a chaperone protein whose expression is significantly enhanced under various conditions including glucose deprivation, oxidative stress, treatment with Ca2+ ionophores, and hypoxia [37]. Higher levels of HSPA5 are essential for sustaining cell viability under specific kinds of stress. The up-regulation of stress proteins in tumour cells has been shown to inhibit programmed cell death and to contribute to drug resistance [37]. Therefore, HSPA5 has some potential as a novel therapeutic target for both anti-tumor and anti-angiogenesis activity [38]. Similar to the blue module, the turquoise module is mainly composed of proteins which are down-regulated under tumour-forming conditions and are mainly enriched in “vesicle membrane” fractions. The main cluster in Fig 7B contains most of the proteins having known biological functions. Surprisingly, the YWHAZ protein was found to be down-regulated in our study, whilst Nishimura et al. [39] observed that YWHAZ-overexpression plays a major role in tumour cell proliferation. One of the highly interconnected protein members of the hub proteins was YWHAG, which is a 14-3-3 adapter protein involved in the regulation of a broad spectrum of signalling pathways. YWHAG binds to a large number of partners, usually by recognition of a phosphoserine or phosphothreonine motif. Binding generally results in the modulation of the activity of the binding partner by protein kinase C inhibitor activity. A protein kinase C (PRKG) was also found co-expressed in the turquoise module. Both, 14-3-3 protein YWHAG and YWHAZ in combination with CAMK2A were found in the same sub-cluster Tu1a and are involved in the regulation of mitochondrial membrane permeability. The soft threshold method used in this study (β = 10) significantly reduced the importance of module interconnection. However, a few interesting proteins were identified in the ER which are more strongly co-expressed with proteins in the mitochondria including PDIA6 and HSPA5/GRP78 which are known to play a crucial role on apoptosis inhibition [38, 40]. Regarding the blue module, a few proteins were found to be highly co-expressed with other proteins outside the module suggesting a co-ordination role far beyond their immediate environment. One of the identified proteins is CKMT1B, which was also found to be down-regulated in squamous cell carcinomas and in clinical samples [41]. A component of the turquoise module is the isoform 1 of pyruvate kinase (PKM1), which was highly co-expressed with more proteins than its counterpart PKM2 from the brown module (Fig 8). It has often been described in the literature that the PKM protein expression switches from PKM1 to the PKM2 isoform during tumourigenesis [42, 43]. We observed a change in isoform ratio where the PKM1 isoform is down-regulated with a ratio tumour/control = 0.26 associated to the turquoise module. While the PKM2 isoform is up-regulated (ratio tumour/control = 2.14 and clustered in the brown module). The observed changes in this current study, although meaningful, do not support a complete shift from one isoform to the other one as described by Bluemlein et al.[44]. The two isoforms of PKM are differentially expressed (M1 and M2) with the different co-expression network proteins of each isoform supporting an increase in aerobic glycoysis at the expense of oxidative phosphorylation (rendered inefficient due to the disruption of the ETC). Decreasing the PKM2/PKM1 ratio has recently been described as a therapeutic strategy in patients with glioblastoma multiforme [45]. As shown in Fig 8, co-expression of PKM2 was limited to only three other proteins (ANXA5, PFN1, and RPS11). Conversely, PKM1 was found co-expressed with more than 30 other proteins from the turquoise module which are mostly involved in ion transport, cellular response to insulin stimulus, glutamate secretion as well as syntaxin binding; a common theme among these proteins is related to the synaptic vesicle cycle with 12 out of the 32 proteins being directly involved in this pathway. Although a broad range of functions is associated to the different proteins co-expressed with PKM1, our findings support that pyruvate kinases are possibly bound to synaptic vesicles with substrates that may be supporting vesicular glutamate uptake [46]. In addition, several of the highly PKM1 co-expressed proteins reported in this study were newly identified. Guanine nucleotide binding protein (GNAO1) and syntaxin binding protein 1 (STXBP1) and the protein cell adhesion molecule 3, CADM3 involved in the calcium-independent cell-cell adhesion molecules has not been previously reported and merit to be explored by more tissue targeted analysis of both PKM. It is intriguing that the 2 PKM isoforms show expression patterns which are not co-localised; PKM2 found mostly co-expressed with proteins from the mitochondrial fraction while PKM1 found co-expressed with proteins related with vesicular membrane. In summary, protein co-expression analysis of the mitochondrial protein fraction revealed novel protein networks with several intrinsically linked functions and uncovered functional modulesHere we have shown and validated with several different strategies that a weighted protein co-expression analysis complements more conventional approaches based on differentially expressed proteins from different groups and can serve as a valuable method for revealing new trends and information clustering which are impossible to capture otherwise. Supporting Information S1 Fig Soft threshold parameters and resulting scale-free topology. (DOC) Click here for additional data file. S1 Table Complete proteomics dataset used in this study (proteins identified with at least 2+ peptides). (XLS) Click here for additional data file. S2 Table Table of the trait matrix used for the correlation with the modules. (XLS) Click here for additional data file. This project has received funding from the European Union’s Seventh Framework Programme for research, technological development and demonstration under grant agreement no 602470. ==== Refs References 1 Denervaud N , Becker J , Delgado-Gonzalo R , Damay P , Rajkumar AS , Unser M , et al A chemostat array enables the spatio-temporal analysis of the yeast proteome . Proc Natl Acad Sci U S A . 2013 ;110 (39 ):15842 –7 . 10.1073/pnas.1308265110 24019481 2 Wegener KM , Singh AK , Jacobs JM , Elvitigala T , Welsh EA , Keren N , et al Global proteomics reveal an atypical strategy for carbon/nitrogen assimilation by a cyanobacterium under diverse environmental perturbations . Mol Cell Proteomics . 2010 ;9 (12 ):2678 –89 . 10.1074/mcp.M110.000109 20858728 3 Levin Y . The role of statistical power analysis in quantitative proteomics . Proteomics . 2011 ;11 (12 ):2565 –7 . 10.1002/pmic.201100033 .21591257 4 Horvath S , Zhang B , Carlson M , Lu KV , Zhu S , Felciano RM , et al Analysis of oncogenic signaling networks in glioblastoma identifies ASPM as a molecular target . Proc Natl Acad Sci U S A . 2006 ;103 (46 ):17402 –7 . 10.1073/pnas.0608396103 17090670 5 Pancaldi V , Schubert F , Bahler J . Meta-analysis of genome regulation and expression variability across hundreds of environmental and genetic perturbations in fission yeast . Mol Biosyst . 2010 ;6 (3 ):543 –52 . 10.1039/b913876p .20174682 6 Gibbs DL , Baratt A , Baric RS , Kawaoka Y , Smith RD , Orwoll ES , et al Protein co-expression network analysis (ProCoNA) . J Clin Bioinforma . 2013 ;3 (1 ):11 10.1186/2043-9113-3-11 23724967 7 Langfelder P , Horvath S . WGCNA: an R package for weighted correlation network analysis . BMC Bioinformatics . 2008 ;9 :559 10.1186/1471-2105-9-559 19114008 8 Langfelder PaH, S. Package ‘WGCNA’. http://cranr-projectorg/web/packages/WGCNA/indexhtml 2014. 9 Bader GD , Hogue CW . An automated method for finding molecular complexes in large protein interaction networks . BMC Bioinformatics . 2003 ;4 :2 12525261 10 Snel B , Lehmann G , Bork P , Huynen MA . STRING: a web-server to retrieve and display the repeatedly occurring neighbourhood of a gene . Nucleic Acids Res . 2000 ;28 (18 ):3442 –4 . 10982861 11 Szklarczyk D , Franceschini A , Wyder S , Forslund K , Heller D , Huerta-Cepas J , et al STRING v10: protein-protein interaction networks, integrated over the tree of life . Nucleic Acids Res . 2015 ;43 (Database issue):D447 –52 . 10.1093/nar/gku1003 25352553 12 Deighton RF , Le Bihan T , Martin SF , Gerth AM , McCulloch M , Edgar JM , et al Interactions among mitochondrial proteins altered in glioblastoma . Journal of neuro-oncology . 2014 ;118 (2 ):247 –56 . 10.1007/s11060-014-1430-5 24728830 13 Mudunuri U , Che A , Yi M , Stephens RM . bioDBnet: the biological database network . Bioinformatics . 2009 ;25 (4 ):555 –6 . 10.1093/bioinformatics/btn654 19129209 14 Eden E , Navon R , Steinfeld I , Lipson D , Yakhini Z . GOrilla: a tool for discovery and visualization of enriched GO terms in ranked gene lists . BMC Bioinformatics . 2009 ;10 :48 10.1186/1471-2105-10-48 19192299 15 Shannon P , Markiel A , Ozier O , Baliga NS , Wang JT , Ramage D , et al Cytoscape: a software environment for integrated models of biomolecular interaction networks . Genome Res . 2003 ;13 (11 ):2498 –504 . 10.1101/gr.1239303 14597658 16 Brohee S . Using the NeAT toolbox to compare networks to networks, clusters to clusters, and network to clusters . Methods Mol Biol . 2012 ;804 :327 –42 . 10.1007/978-1-61779-361-5_18 .22144162 17 Kaimal V , Bardes EE , Tabar SC , Jegga AG , Aronow BJ . ToppCluster: a multiple gene list feature analyzer for comparative enrichment clustering and network-based dissection of biological systems . Nucleic Acids Res . 2010 ;38 (Web Server issue):W96 –102 . 10.1093/nar/gkq418 20484371 18 He Y , Zhang M , Ju Y , Yu Z , Lv D , Sun H , et al dbDEPC 2.0: updated database of differentially expressed proteins in human cancers . Nucleic Acids Res . 2012 ;40 (Database issue):D964 –71 . Epub 2011/11/19. 10.1093/nar/gkr936 22096234 19 Barabasi AL , Bonabeau E . Scale-free networks . Sci Am . 2003 ;288 (5 ):60 –9 . .12701331 20 Wang L , Zheng W , Zhao H , Deng M . Statistical analysis reveals co-expression patterns of many pairs of genes in yeast are jointly regulated by interacting loci . PLoS Genet . 2013 ;9 (3 ):e1003414 10.1371/journal.pgen.1003414 23555313 21 Storey JD , Akey JM , Kruglyak L . Multiple locus linkage analysis of genomewide expression in yeast . PLoS Biol . 2005 ;3 (8 ):e267 10.1371/journal.pbio.0030267 16035920 22 Wang Z , Lin H , Hua F , Hu ZW . Repairing DNA damage by XRCC6/KU70 reverses TLR4-deficiency-worsened HCC development via restoring senescence and autophagic flux . Autophagy . 2013 ;9 (6 ):925 –7 . 10.4161/auto.24229 23518600 23 Alldridge L , Metodieva G , Greenwood C , Al-Janabi K , Thwaites L , Sauven P , et al Proteome profiling of breast tumors by gel electrophoresis and nanoscale electrospray ionization mass spectrometry . J Proteome Res . 2008 ;7 (4 ):1458 –69 . 10.1021/pr7007829 .18257521 24 Patwardhan AJ , Strittmatter EF , Camp DG , 2nd, Smith RD, Pallavicini MG. Quantitative proteome analysis of breast cancer cell lines using 18O-labeling and an accurate mass and time tag strategy . Proteomics . 2006 ;6 (9 ):2903 –15 . Epub 2006/04/06. 10.1002/pmic.200500582 .16596714 25 Sun Y , Mi W , Cai J , Ying W , Liu F , Lu H , et al Quantitative proteomic signature of liver cancer cells: tissue transglutaminase 2 could be a novel protein candidate of human hepatocellular carcinoma . J Proteome Res . 2008 ;7 (9 ):3847 –59 . 10.1021/pr800153s .18646787 26 Chen YR , Juan HF , Huang HC , Huang HH , Lee YJ , Liao MY , et al Quantitative proteomic and genomic profiling reveals metastasis-related protein expression patterns in gastric cancer cells . J Proteome Res . 2006 ;5 (10 ):2727 –42 . 10.1021/pr060212g .17022644 27 Gu Y , Wu SL , Meyer JL , Hancock WS , Burg LJ , Linder J , et al Proteomic analysis of high-grade dysplastic cervical cells obtained from ThinPrep slides using laser capture microdissection and mass spectrometry . J Proteome Res . 2007 ;6 (11 ):4256 –68 . 10.1021/pr070319j .17902640 28 Krause K , Karger S , Schierhorn A , Poncin S , Many MC , Fuhrer D . Proteomic profiling of cold thyroid nodules . Endocrinology . 2007 ;148 (4 ):1754 –63 . 10.1210/en.2006-0752 .17194741 29 Garbis SD , Tyritzis SI , Roumeliotis T , Zerefos P , Giannopoulou EG , Vlahou A , et al Search for potential markers for prostate cancer diagnosis, prognosis and treatment in clinical tissue specimens using amine-specific isobaric tagging (iTRAQ) with two-dimensional liquid chromatography and tandem mass spectrometry . J Proteome Res . 2008 ;7 (8 ):3146 –58 . 10.1021/pr800060r .18553995 30 Friedman DB , Hill S , Keller JW , Merchant NB , Levy SE , Coffey RJ , et al Proteome analysis of human colon cancer by two-dimensional difference gel electrophoresis and mass spectrometry . Proteomics . 2004 ;4 (3 ):793 –811 . 10.1002/pmic.200300635 .14997500 31 Pradeep C , Prerna D , Lukiw WJ . An Exploratory Analysis of Conservation of Co-Expressed Genes across Alzheimer’s disease Progression . J Comput Sci Syst Biol 2013 ;6 :215 –27 . 32 Raturi A , Simmen T . Where the endoplasmic reticulum and the mitochondrion tie the knot: the mitochondria-associated membrane (MAM) . Biochim Biophys Acta . 2013 ;1833 (1 ):213 –24 . 10.1016/j.bbamcr.2012.04.013 .22575682 33 James R , Searcy JL , Le Bihan T , Martin SF , Gliddon CM , Povey J , et al Proteomic analysis of mitochondria in APOE transgenic mice and in response to an ischemic challenge . Journal of cerebral blood flow and metabolism: official journal of the International Society of Cerebral Blood Flow and Metabolism . 2012 ;32 (1 ):164 –76 . 10.1038/jcbfm.2011.120 21878944 34 Chen Y , McMillan-Ward E , Kong J , Israels SJ , Gibson SB . Mitochondrial electron-transport-chain inhibitors of complexes I and II induce autophagic cell death mediated by reactive oxygen species . J Cell Sci . 2007 ;120 (Pt 23 ):4155 –66 . 10.1242/jcs.011163 .18032788 35 Herrmann AG , Deighton RF , Le Bihan T , McCulloch MC , Searcy JL , Kerr LE , et al Adaptive changes in the neuronal proteome: mitochondrial energy production, endoplasmic reticulum stress, and ribosomal dysfunction in the cellular response to metabolic stress . Journal of cerebral blood flow and metabolism: official journal of the International Society of Cerebral Blood Flow and Metabolism . 2013 ;33 (5 ):673 –83 . 10.1038/jcbfm.2012.204 23321784 36 Chae YC , Angelin A , Lisanti S , Kossenkov AV , Speicher KD , Wang H , et al Landscape of the mitochondrial Hsp90 metabolome in tumours . Nat Commun . 2013 ;4 :2139 10.1038/ncomms3139 23842546 37 Song MS , Park YK , Lee JH , Park K . Induction of glucose-regulated protein 78 by chronic hypoxia in human gastric tumor cells through a protein kinase C-epsilon/ERK/AP-1 signaling cascade . Cancer Res . 2001 ;61 (22 ):8322 –30 . .11719466 38 Dong D , Ni M , Li J , Xiong S , Ye W , Virrey JJ , et al Critical role of the stress chaperone GRP78/BiP in tumor proliferation, survival, and tumor angiogenesis in transgene-induced mammary tumor development . Cancer Res . 2008 ;68 (2 ):498 –505 . 10.1158/0008-5472.CAN-07-2950 .18199545 39 Nishimura Y , Komatsu S , Ichikawa D , Nagata H , Hirajima S , Takeshita H , et al Overexpression of YWHAZ relates to tumor cell proliferation and malignant outcome of gastric carcinoma . Br J Cancer . 2013 ;108 (6 ):1324 –31 . Epub 2013/02/21. 10.1038/bjc.2013.65 23422756 40 Tufo G , Jones AW , Wang Z , Hamelin J , Tajeddine N , Esposti DD , et al The protein disulfide isomerases PDIA4 and PDIA6 mediate resistance to cisplatin-induced cell death in lung adenocarcinoma . Cell Death Differ . 2014 ;21 (5 ):685 –95 . 10.1038/cdd.2013.193 24464223 41 Onda T , Uzawa K , Endo Y , Bukawa H , Yokoe H , Shibahara T , et al Ubiquitous mitochondrial creatine kinase downregulated in oral squamous cell carcinoma . Br J Cancer . 2006 ;94 (5 ):698 –709 . 10.1038/sj.bjc.6602986 16479256 42 Christofk HR , Vander Heiden MG , Harris MH , Ramanathan A , Gerszten RE , Wei R , et al The M2 splice isoform of pyruvate kinase is important for cancer metabolism and tumour growth . Nature . 2008 ;452 (7184 ):230 –3 . Epub 2008/03/14. 10.1038/nature06734 .18337823 43 Clower CV , Chatterjee D , Wang Z , Cantley LC , Vander Heiden MG , Krainer AR . The alternative splicing repressors hnRNP A1/A2 and PTB influence pyruvate kinase isoform expression and cell metabolism . Proc Natl Acad Sci U S A . 2010 ;107 (5 ):1894 –9 . 10.1073/pnas.0914845107 20133837 44 Bluemlein K , Gruning NM , Feichtinger RG , Lehrach H , Kofler B , Ralser M . No evidence for a shift in pyruvate kinase PKM1 to PKM2 expression during tumorigenesis . Oncotarget . 2011 ;2 (5 ):393 –400 . Epub 2011/07/27. 21789790 45 Han D , Wei W , Chen X , Zhang Y , Wang Y , Zhang J , et al NF-kappaB/RelA-PKM2 mediates inhibition of glycolysis by fenofibrate in glioblastoma cells . Oncotarget . 2015 ;6 (28 ):26119 –28 . 10.18632/oncotarget.4444 .26172294 46 Ishida A , Noda Y , Ueda T . Synaptic vesicle-bound pyruvate kinase can support vesicular glutamate uptake . Neurochem Res . 2009 ;34 (5 ):807 –18 . 10.1007/s11064-008-9833-3 18751889

PMC005xxxxxx/PMC5003356.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757122710.1371/journal.pone.0161930PONE-D-16-14364Research ArticleBiology and life sciencesBiochemistryProteinsDNA-binding proteinsBiology and Life SciencesCell BiologyCell PhysiologyCell FusionBiology and Life SciencesBiochemistryProteinsProtein InteractionsBiology and Life SciencesImmunologyImmune System ProteinsMedicine and Health SciencesImmunologyImmune System ProteinsBiology and Life SciencesBiochemistryProteinsImmune System ProteinsBiology and Life SciencesBiochemistryProteinsConjugated ProteinsBiology and life sciencesBiochemistryProteinsRNA-binding proteinsBiology and Life SciencesBiochemistryProteinsProtein DomainsBiology and life sciencesBiochemistryProteinsDNA-binding proteinsNucleasesRibonucleasesBiology and Life SciencesBiochemistryEnzymologyEnzymesHydrolasesNucleasesRibonucleasesBiology and Life SciencesBiochemistryProteinsEnzymesHydrolasesNucleasesRibonucleasesDevelopment of a Quantitative BRET Affinity Assay for Nucleic Acid-Protein Interactions Nucleic Acid-Protein BRET Affinity AssayVickers Timothy A. *Crooke Stanley T. Department of Core Antisense Research, Ionis Pharmaceuticals, Inc., 2855 Gazelle Court, Carlsbad, CA, 92010, United States of AmericaCaputi Massimo EditorFlorida Atlantic University, UNITED STATESCompeting Interests: This study was funded by Ionis Pharmaceuticals, Inc. The funder provided support in the form of salaries for authors TAV and STC but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. There are no patents, products in development or marketed products to declare. This does not alter our adherence to all the PLOS ONE policies on sharing data and materials. Conceptualization: TAV STC. Data curation: TAV. Formal analysis: TAV. Funding acquisition: TAV STC. Investigation: TAV. Methodology: TAV. Project administration: STC. Resources: TAV. Supervision: STC. Validation: TAV. Writing – original draft: TAV. Writing – review & editing: STC TAV. * E-mail: tvickers@ionisph.com29 8 2016 2016 11 8 e01619308 4 2016 15 8 2016 © 2016 Vickers, Crooke2016Vickers, CrookeThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Protein-nucleic acid interactions play a crucial role in the regulation of diverse biological processes. Elucidating the roles that protein-nucleic acid complexes play in the regulation of transcription, translation, DNA replication, repair and recombination, and RNA processing continues to be a crucial aspect of understanding of cell biology and the mechanisms of disease. In addition, proteins have been demonstrated to interact with antisense oligonucleotide therapeutics in a sequence and chemistry dependent manner, influencing ASO potency and distribution in cells and in vivo. While many assays have been developed to measure protein-nucleic acid interactions, many suffer from lack of throughput and sensitivity, or challenges with protein purification and scalability. In this report we present a new BRET assay for the analysis of DNA-protein interactions which makes use of an extremely bright luciferase as a tag for the binding protein, along with a long-wavelength fluorophore conjugated to the nucleic acid. The resulting assay is high throughput, sensitive, does not require protein purification, and even allows for quantitative characterization of these interactions within the biologically relevant context of whole cells. This work was supported by Ionis Pharmaceuticals, Inc. The funder provided support in the form of salaries for authors TAV and STC but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Proteins interact with DNA and RNA through electrostatic interactions, hydrogen bonding, hydrophobic interactions, and base stacking [1–4]. These forces contribute in varying degrees to proteins binding in a structure and sequence specific or non-sequence specific manner [5]. Understanding how proteins interact with nucleic acids and identifying the nucleic acid sequence (and possibly structure) required to assemble these complexes are vital to understanding the role these complexes play in regulating cellular processes. Furthermore, the broad and increasing use of oligonucleotides of various types as research tools and platforms for drug discovery and development demand a more thorough understanding of how these pharmacological agents interact with various proteins and how chemical modifications, sequence and structure influence interactions with proteins. We have recently demonstrated that approximately 50 intracellular proteins interact with phosphorothioate modified antisense oligonucleotides (PS ASOs) [6]. To fully characterize interactions with all of these proteins and various mutants is currently a daunting task due to the need to generate and purify proteins to evaluate binding interactions. In addition, antibodies must frequently be generated to facilitate isolation and evaluation of proteins. Moreover, it would be ideal to be able to compare binding to the purified protein with binding to the protein of interest in the native complexes present in the cell or cell homogenates. In this paper we report the development of a new assay that is rapid, high throughput, and provides information on protein-nucleic acid interactions that is not provided by other assays. The assay is based on a bioluminescence resonance energy transfer (BRET) assay utilizing the recently introduced Nanoluc luciferase (Nluc) [7], which has resulted in a significant improvement in the performance of BRET assays for protein/protein interactions (NanoBRET) [8–10]. Nluc is an engineered protein, which uses a novel coelenterazine derivative (furimazine) as its substrate. Its small size (19 kDa) and high physical stability result in minimal influence when tethered to other proteins. Although it is only about half the size of Rluc, it produces sustained luminescence with much greater intensity, allowing very small quantities to be accurately quantified. Moreover, its bioluminescence spectrum is narrower than Rluc, permitting better spectral discrimination with acceptor fluorophores. We took advantage of these unique biochemical properties of NLuc to develop a BRET assay which relies on the transfer of light energy from an Nluc tagged binding protein acting as the BRET donor, to a fluorescently tagged ASO acting as the BRET acceptor. This rapid and high throughput BRET assay works with immunopurified proteins, cell homogenates, or even in intact cells, and can generate KD’s and relative KD’s for the Protein/ASO interaction. The assay also provides information on relative binding distances and orientation and does not require protein purification or denaturation, thereby supporting evaluation of native fusion protein/ASO interactions. In addition, mutant proteins can be rapidly generated and expressed to obtain information on binding domains and protein structure. In addition to interactions of proteins with modified ASOs, the assay can also be used to measure interactions between natural nucleic acids and their binding partners. We present data demonstrating the utility of the BRET assay for binding of dsDNA by transcription factors, and for measuring affinity of structured RNAs to RNA binding proteins. Materials and Methods Preparation of antisense oligonucleotides Synthesis and purification of phosphorothioate/2’-MOE, 2’-F or S-cEt oligonucleotides was performed as described previously [11]. Standard phosphodiester deoxy and ribo oligonucleotides were obtained from Integrated DNA technologies (Coralville, Iowa). Oligonucleotides used in this work are detailed in S1 Table. Construction, expression, and purification of fusion proteins NanoLuc fusion protein construction, expression, and purification were performed using the vectors pFN31K Nluc CMV-neo for amino-terminal clones and pFC32K Nluc CMV-neo for carboxy-terminal clones (Promega). Briefly, the gene of interest (GOI) was amplified using PCR primers complementary to full length cDNAs obtained from Origene. For cloning into pFN31K Nluc CMV-neo, the forward PCR primer was comprised of sequence complimentary to the sequence of the GOI following the AUG start preceded by an XhoI site for cloning in frame with NLuc, whereas the reverse primer was complementary to the sequence preceding the stop codon of the GOI followed by an EcoRI site. The PCR amplified product was then digested with XhoI and EcoRI then ligated into the pFN31K Nluc CMV-neo vector prepared with the same enzymes using standard techniques. For cloning into pFC32K Nluc CMV-neo, the forward PCR primer was comprised of sequence complimentary to the GOI sequence, including the AUG start codon preceded by a Kozak sequence and an NheI site (gctagcAGCCACC), whereas the reverse primer was complementary to the sequence, including the stop codon of the GOI followed by an XhoI site. The PCR amplified product was then digested with NheI and XhoI and ligated into the pFC32K Nluc CMV-neo vector prepared with the same enzymes. Sequences of PCR cloning primers can be found in S2 Table. Deletion mutants for P54nrb, La, and RNAse H1 were generated by site directed mutagenesis of the fusion constructs described above using a QuikChange Lightning Site Directed Mutagenesis (SDM) Kit (Agilent Technologies) essentially as detailed previously [12]. Sequences of SDM primers can be found in S3 Table. NLuc fusion proteins were expressed by transfecting into 6 x 105 HeLa cells using Effectene transfection reagent according to the manufacturer’s protocol (Qiagen). Following a 24 hour incubation, cells were removed from the plate by trypsinization, pelleted, washed 1X with PBS, then resuspended in 250 μl Pierce IP Lysis Buffer (Thermo Scientific). Lysates were incubated 1 hour at 4°C while rotating, then debris pelleted by centrifugation at 15,000 rpm for 5 min. Fusion proteins were immunoprecipitated overnight at 4°C while rotating using 1–2 μg of antibody specific to the GOI or NLuc (Promega) (S1 Fig). The following day, 20 μl of Pierce Protein G Magnetic Beads (Thermo Scientific) were added and the incubation continued for 2 hours. Beads were washed 4X with IP Lysis Buffer and finally suspended in 250 μl 2X binding buffer (0.2 M Potassium acetate, 40 mM Tris, pH 7.5, 2 mM EDTA, .02% NP-40). Western blots were carried out as detailed previously [13]. BRET binding affinity assay BRET assays were performed in white 96 well plates. Alexa-linked ASOs at the indicated concentrations were incubated at room temperature for 15 min in 1X binding buffer with 106 RLU/well of immunoprecipitated NLuc fusion protein or whole cell lysate. Following the incubation, NanoGlo substrate (Promega) was added at 0.1 μl/well. Readings were performed for 0.3 sec using a Glomax Discover system using450 nm/8 nm band pass for the donor filter, and 600 nm long pass for the acceptor filter. BRET was calculated as the ratio of the emission at 600/450 (fluorescent excitation emission/RLU). For competitive binding assays the Alexa-linked ASO was added at approximately the KD for the particular fusion protein and the unconjugated competing ASO added at the indicated concentrations in 50 μl water. 106 RLU/ well of immunoprecipitated fusion protein or whole cell lysate well was then added in 50 μl 2X binding buffer for a final volume of 100 μl. After a 15 min incubation at room temperature, substrate addition and BRET readings were carried out as detailed above. Permeabilized cell assay NLuc fusion plasmids were transfected into 6 x 105 HeLa cells using Effectene transfection reagent as detailed above. After 4 hours transfection, reagent was removed, cells trypsinized and then plated in 96 well white clear bottom plates at 5000/well. After an overnight incubation, growth media was removed, and cells washed 1X with PBS. Cells were then incubated in OptiMEM media + 25 μg/ml digitonin for 5 min, then Alexa-linked ASOs at indicated concentrations, followed by a 15 min incubation at 37°C. Substrate addition and BRET readings were then carried out as detailed above. Fluorescent polarization assay The 20nt 3’Alexa-linked cEt gap-mer ASO was combined at 0.3 nM with unconjugated competing ASO at the indicated concentrations in 50 μl water. 50 μl of 10 nM purified HIS-tagged P54nrb protein in 2X binding buffer was then added. Following a 15 min incubation, fluorescence polarization was read using an Infinite M1000 PRO plate reader (TECAN) with an excitation wavelength of 590 ±5 nm and an emission wavelength of 620± 20 nm. Results It was recently reported that the Drosophila behavior/human splicing (DBHS) family proteins P54nrb (NONO), PSF (SFPQ), and PSPC1 can bind to PS ASOs, resulting in inhibition of ASO directed RNAseH1-mediated activity [14]. In these experiments, association of proteins with ASOs was determined by affinity selection of proteins bound to PS-ASO/RNA duplex. Interactions between P54nrb/PSF and single-stranded PS-ASOs are mainly influenced by the PS modification, however different 2’ moieties can further influence the binding of PS-ASOs to P54nrb [15]. To investigate interactions of modified ASOs with proteins in a more quantitative manner, we sought to develop a binding assay that was rapid, robust, and highly quantitative. We developed a BRET affinity assay in which NLuc was fused in frame to P54nrb at either the amino- (NLuc-P54nrb) or carboxy-terminus (P54nrb-NLuc) of the protein under the control of the CMV immediate early enhancer/promoter as detailed in Materials and Methods (Fig 1A). Transfection of HeLa cells with the P54nrb fusion constructs resulted in significant levels of the fusion protein as determined by western blot analysis (S1 Fig). Whole cell lysates were generated and the fusion protein immunoprecipitated using an antibody directed to P54nrb. To normalize for differences in expression levels, 106 RLU/well of the immunoprecipitated fusion protein were incubated in a 96-well plate with PS-modified, 5–10–5 “gap-mer” ASOs containing 10 deoxyribonucleotides in the middle flanked at both ends by five 2′-constrained ethyl (cEt) and conjugated at either the 5’ or 3’ end with Alexa 594 (S1 Table) at concentrations ranging from 10 pM to 10 μM. NLuc substrate was then added and BRET ratios determined as detailed in Materials and Methods. The KD for ASO binding was determined to be 9–14 nM for P54nrb-NLuc and 6–12 nM for NLuc-P54nrb. Interestingly, the amplitude of the BRET signal was greater with NLuc-P54nrb than P54nrb-NLuc. BRET efficiency is a measure of distance between the donor and the acceptor, which varies according to an inverse sixth power of the distance between the two molecules [16]. Therefore, the greater amplitude of signal with the NLuc-P54nrb construct may be indicative of the ASO binding nearer the amino terminus than the carboxy terminus of P54nrb. Such an interaction would be consistent with the ASO binding either of the RNA recognition motifs found near the amino terminus of the P54nrb protein [17]. These results were confirmed by evaluating ASO binding to La, a conserved component of eukaryotic ribonucleoprotein complexes that binds the 3' poly (U)-rich elements of nascent RNA polymerase III (pol III) transcripts to assist folding and maturation [18]. In addition, La has previously been shown to interact with PS ASOs [6]. Similar to P54nrb, the amplitude of the BRET signal was significantly greater for NLuc-La than the La-Nluc construct (S2A Fig), consistent with the structural organization of the La RNA binding domains nearer the amino terminus of the protein [19]. 10.1371/journal.pone.0161930.g001Fig 1 A) General cloning strategy for NLuc fusion proteins. P54nrb cDNA was amplified by PCR to include an N-terminal XhoI site and a C-terminal EcoRI site. The resulting cDNA was cloned in frame using the same sites in the plasmid pFN31K Nluc. For cloning into pFC32K Nluc the cDNA was amplified by PCR to include an N-terminal NheI site and a C-terminal XhoI site. Primer sequences can be found in S2 Table. Expression of all clones is driven from the CMV promoter. The resulting fusion plasmids were expressed in HeLa cells. B) ASO/BRET affinity for p54nrb. P54nrb/NLuc fusion proteins were immunopurified as detailed in Materials and Methods and subsequently incubated with Alexa 594 conjugated 5-10-5 cEt gap-mer ASO at concentrations ranging from 10 pM to 10 μM. BRET ratios were determined for P54nrb-NLuc (green, black) or NLuc-P54nrb (red, blue) with either a 3’ conjugated (766636) or 5’ conjugated (766635) ASO. Concentration response curves and KD’s (nM) for two independent experiments are shown. C) ASO/BRET binding affinity varies with chemistry of the 2’ modification. ASO/BRET assay was performed with P54nrb-NLuc fusion and 5’ conjugated 5-10-5 ASOs at concentrations ranging from 10 pM to 10 μM. 2’-F, red; cEt, blue; MOE, green. D) Relative affinities for 2’F (red), cET (blue), and MOE (green) gap-mer ASO as determined by competitive ASO binding to the NLuc/P54nrb fusion protein in the BRET assay. 10 nM 3’ Alexa conjugated cEt ASO (766636) was competed with unconjugated 5–10–5 2′-F (red), MOE (green), or cEt (blue) gap-mer ASO at concentrations from 0.1 to 1000 nM. Relative KD’s are shown. Data in panels C and D are mean ± SEM from 3–4 independent experiments. As it was previously observed that the affinity of ASOs for P54nrb is highly influenced by the 2′ modification included in the “wings” of the ASO gap-mer [14], we evaluated the effect of 2’ modifications on ASO affinity for P54nrb by BRET. Approximately 106 RLU/well of the immunoprecipitated P54nrb-NLuc fusion protein were incubated in a 96-well plate with 20 nucleotide 5-10-5 cEt, 2′-fluoro (2′-F), or 2′-O-methoxyethyl (MOE) gap-mer ASOs conjugated at the 5’ end with Alexa 594 at concentrations ranging from 10 pM to 10 μM. NLuc substrate was then added and BRET ratios determined. As observed previously, 2’ modifications had a significant effect on the affinity of the ASO for P54nrb, with the 2’-F binding with a KD of 2.1 nM, the cEt 9.3 nM, and the MOE 82.9 nM (Fig 1C). This is the same rank order of affinity previously observed by affinity selection. 2’modifications had a less significant effect on ASO affinity for NLuc-La protein, with the 2’-F binding with a KD of 1.1 nM, the cEt 5.0 nM and the MOE 9.3 nM (S2B Fig). We next sought to broaden the utility of the assay by eliminating the requirement that the ASO be conjugated to an ALEXA fluorophore. A competitive NLuc-P54nrb binding assay was performed in which 10 nM Alexa conjugated 2’F ASO was competed with unconjugated 5–10–5 2′-F, MOE, or cEt gap-mer ASO at concentrations from 0.1 to 1000 nM. As with direct binding, competitive binding curves revealed the same rank order in binding affinity based upon 2’ modification of the competing ASO: 2’F >cEt >MOE (Fig 1D). However, while the relative KD’s were similar to those observed by direct binding for the 2’F and cEt, the KD for the MOE was higher in the competitive binding assay. This may result from differences in binding of the MOE ASO to the N-terminally vs C-terminally conjugated P54nrb fusion protein. KD’s obtained by competitive binding of NLuc-La were very similar to those obtained by direct binding and also followed the same rank order based upon the 2’ modification of the ASO (S2C Fig). To further investigate the utility of BRET for ASO/protein affinity measurements, multiple NLuc fusion proteins were cloned then evaluated by direct binding with the same 2’F, MOE and cEt ASOs (S1 Table). The binding affinity for the 2’-F ASO was determined to be highly dependent on the conjugated protein, with KDs varying over 4 orders of magnitude (Fig 2). The affinity of the cEt and MOE ASOs also varied widely depending on the protein bound. A summary of affinity for all proteins can be found in Table 1. It is possible that a portion of this high variability in KDs may be due differences in sequence specificity of the proteins evaluated. However, it is interesting to note that the chemistry of the 2’ modification also had a variable effect on protein binding. For many proteins, affinity increased with increasing hydrophobicity of the 2’ modification (2’F>cEt>MOE), with almost 2 orders of magnitude difference in affinity between 2’ modifications for certain proteins. There also seems to be little correlation between the number or type of nucleic acid binding domains present in the protein and ASO affinity. 10.1371/journal.pone.0161930.g002Fig 2 ASO/BRET binding affinities for various NLuc fusion proteins. NLuc protein fusions were constructed, expressed, and immunopurified as detailed in Materials and Methods. Binding affinities were determined by incubating 106 RLU of immunopurified fusion protein with a 3’ ALEXA 594 conjugated 5-10-5 2’F gap-mer ASO (766638) at concentrations from 1 pM to 1 μM. Data are plotted as the percent of the maximal BRET ratio to control for differences in BRET amplitude for the various proteins. KD’s were determined using GraphPad Prism software. Similar experiments were performed for cEt (766636) and MOE (766634) gap-mer ASOs. The KD’s can be found in Table 1. 10.1371/journal.pone.0161930.t001Table 1 Summary of NLuc fusion protein constructs and ASO/BRET binding affinities. Gene ID of full length protein is given. For certain proteins the full length protein was not cloned and the domains included in the NLuc fusion are indicated in parentheses. Known nucleic acid binding domains: RNA binding domain (RBD), DNA binding domain (DBD), Hybrid binding domain (HBD). KD’s (nM) for 5-10-5 2’F (766638), cEt (766636) and MOE (766634) gap-mer ASOs were determined using GraphPad Prism software. Protein NLuc Domains size (kD) KD Fl KD cEt KD MOE LRPPRC C 1-RBD 41.35 0.19 0.16 0.77 FUS C 1-RBD 52.7 0.12 0.6 1.8 PC4 N 1-DBD 14.4 0.21 1.1 6.1 RPL5 N   20 0.6 1.3 3.7 NCL (RBD 1–4) N/C 4-RBD 39 0.002 1.7 0.009 SFPQ C 2-RBD 76 0.72 2.7 3.7 Ku70 C 2-DBD 69.9 3 4 15 RNAseH1 N/C 1-HBD 32 2 5 2 La N/C 2-RBD 46.8 1.1 5.0 9.3 P54nrb N/C 2-RBD 54 2.1 9.3 82.9 RPL11 N   34.4 9.8 17.4 15.7 HSP90 (mid) C   47 98 43 167 Staufen C 3-RBD 55   100   TCP1-B N   57 189 113 398 ACTB N   42 28 295 252 NMP1 N/C 1-DBD 28.4 >1000 >1000 >1000 ANXA2 C 1-RBD 38 >1000 >1000 >1000 To evaluate the specificity of the assay, the 3’-linked cEt gap-mer ASO was evaluated for binding to NLuc-RNAse H1. The same ASO was also evaluated for binding to NLuc alone, and NLuc-barnase (pFN31K Nluc), a protein not known to interact with ASOs. As has been previously observed [20], the PS ASO bound RNAse H1 with high affinity (Fig 3A, red line). In contrast, no BRET signal was observed for the NLuc-barnase (blue line) or Nluc (green line) control proteins. Similar results were observed with 2’-F or MOE PS gap-mer ASOs (data not shown). Together these data show that all of the ASO binding can be attributed to the fused protein and that NLuc does not contribute to binding. 10.1371/journal.pone.0161930.g003Fig 3 ASO binding to RNase H1 is highly specific. A) NLuc-RNAseH1 fusion (red) was constructed as described in Materials and Methods. The NLuc/Barnase fusion (blue) was produced from the plasmid pFN31K Nluc CMV-neo (Promega), whereas the NLuc only plasmid (green) was generated by deletion of the barnase coding region from the same plasmid. Proteins were expressed then immunopurified using and antibody specific for NLuc. Binding affinities were determined by incubating 106 RLU of immunopurified protein with a 3’ ALEXA 594 conjugated 5-10-5 2’F (766638, solid lines) or MOE gap-mer (766634, dashed lines) ASOs at concentrations from 0.1 nM to 1 μM. B) The RNAse H1 hybrid binding domain (HBD) was deleted from NLuc/RNAseH1 by SDM. NLuc/RNAseH1 (solid lines) and NLuc/RNAseH1-HBD (dashed line) were expressed then immunopurified as above. Binding affinities were determined by incubating 106 RLU of immunopurified protein with a 3’ ALEXA 594 conjugated PO ASO with or without complementary PO DNA or RNA. BRET rations were plotted and KD’s (nM) determined for single stranded DNA (D, red), DNA/DNA duplex (D/D, green), or DNA/RNA heteroduplex (D/R, blue). Data shown are mean ± SEM from 3 independent experiments. In contrast to PS ASOs, PO ASOs have a much lower affinity for RNAse H1, unless they are part of DNA/RNA heteroduplex [20]. Furthermore, it has been demonstrated that the hybrid binding domain (HBD), located at the amino terminus of human RNase H1, mediates binding to the heteroduplex [21]. A 20-mer PO DNA linked with Alexa 594, was evaluated for binding to RNAse H1 in the BRET assay as single strand (D), or as a duplex with a complementary PO DNA (D/D) or PO RNA (D/R). Consistent with data obtained using purified RNAse H1 and a competitive cleavage assay [20], very little binding was observed to the DNA/DNA duplex (Fig 3B, green). While the ssDNA showed some affinity for RNAse H1 (red), the binding affinity of the DNA/RNA heteroduplex (blue) was more than 10 fold greater. Furthermore, deletion of the RNAse H1 HBD in the NLuc fusion, effectively ablated ASO binding (blue dashed lines). To explore the domains involved in ASO- protein interactions, several deletion mutants were produced by site directed mutagenesis (SDM) for La protein. It is known that specific recognition of 3’ poly (U)-rich elements is mediated by the N-terminal domain (NTD) of La, which is comprised a La motif and an RNA recognition motif (RRM1) [18]. However, while it has also previously been shown that La also interacts with PS ASOs [6], little is known concerning the La domains which mediate interactions with ASOs. We deleted the La motif (dM), RRM1 (dR1), RRM2 (dR2), or both RRM1 and RRM2 (dR1/R2), then evaluated binding to the immunoprecipitated protein with the cEt ASO. An approximately 4 fold decrease in affinity was observed for the for the La motif (dM) or RRM2 (dR2) mutants as compared to the full length protein; while deletion of RRM1 (dR1) resulted in an almost 10 fold decrease in affinity and deletion of both RRM1 and RRM2 (dR1/R2) resulted in a ~50 fold decrease (Fig 4A). Similar results were observed for binding of 2’-F and MOE gap-mer ASOs to the same deletion mutants (data not shown). These data suggest that unlike the interaction of La with poly (U)-rich elements, PS ASOs can interact with either RRM1 or RRM2, with a slight preference for RRM1. Similar experiments were performed to identify the P54nrb domains which mediate ASO binding. Deletion of the P54nrb RRM1 and RRM2 domains resulted in a 2–3 fold decrease in affinity as compared to the full length protein (Fig 4B). However, deletion of both RRM1 and RRM2 strongly reduced ASO binding to P54nrb by more than 50 fold. These data suggest that PS ASOs do indeed interact with P54nrb via the RRM domains, however, in contrast to La, there is no strong preference for RRM1 over RRM2. 10.1371/journal.pone.0161930.g004Fig 4 ASOs interaction with protein domains is specific and chemistry dependent. A) cEt gap-mer ASO binds to La RNA binding domains 1 and 2. Deletion mutants were generated from Nluc-La by SDM, then expressed and immunopurified using an antibody to La. Binding affinities were determined by incubating 106 RLU of immunopurified protein with a 3’ ALEXA 594 conjugated 5-10-5 cEt gap-mer ASO (766636) at concentrations ranging from 100 pM to 10 μM. Data are plotted as the percent of the maximal BRET ratio to control for differences in BRET amplitude for the full length La (red), ∆La motif (violet), ∆RRM1 (green), ∆RRM2 (blue), or ∆RRM1/2 (black). B) cEt gap-mer ASO interacts with P54nrb at RRM1 and RRM2. Deletion mutants were generated from P54nrb/Nluc by SDM, then expressed and immunopurified using an antibody to P54nrb (Millipore). Binding affinities were determined by incubating 106 RLU of immunopurified proteins with a 3’ ALEXA 594 conjugated 5-10-5 2’F gap-mer ASO (766638) at concentrations ranging from 100 pM to 1 μM. Concentration curves were plotted for BRET ratios using GraphPad PRISM software for full length P54nrb (red), ∆RRM1 (green), ∆RRM2 (blue), or ∆RRM1/2 (black). Data shown are mean ± SEM from 3–4 independent experiments. We next explored the effect of sequence on ASO affinity in the BRET assay. A series of 12, 3-10-3 cEt gap-mer ASOs (Table 2) targeting mouse trace amine associated receptor 5 (TAAR5), were evaluated for affinity to P54nrb by competitive binding in the BRET assay. ASO affinity to P54nrb was found to vary by over 2 orders of magnitude for this set of ASOs (Fig 5A). Binding of the same set of ASOs to purified P54nrb was also evaluated by fluorescence polarization (FP) (Fig 5B). While the variation in affinity between the tightest binding ASO and the weakest binding ASO was less for the FP assay than the BRET assay, the relative affinity was similar with a significant degree of correlation between the two methods (Fig 5C). Other proteins evaluated in the BRET assay with the same series of ASOs also demonstrated a great deal of variation in affinity, unique to each protein (S3 Fig). For P54nrb and RNAse H1, ASOs with the highest affinity all shared a GGG motif. No common sequence motifs were apparent for NCL or LRPPRC. 10.1371/journal.pone.0161930.g005Fig 5 ASO binding affinity to P54nrb is highly sequence dependent. A) Competitive ASO/BRET binding of 3-10-3 cEt ASOs to P54nrb. Binding affinities were determined by incubating 106 RLU of immunopurified NLuc/P54nrb fusion protein with 10 nM 3’ ALEXA 594 conjugated 5-10-5 2’F gap-mer ASO (766638) along with unconjugated TAAR5 3-10-3 cEt ASOs at concentrations ranging from 1 nM to 3 μM. Concentration curves are plotted for BRET ratios in the presence of each unconjugated ASO. TAAR5 ASO sequences and KD’s can be found in Table 2. B) P54nrb competitive fluorescent polarization assay. Competitive FP was performed with the same set of TAAR5 ASOs as detailed in Materials and Methods. C) Correlation between KD’s for ASO binding obtained by BRET and FP. 10.1371/journal.pone.0161930.t002Table 2 TAAR5 ASO protein affinity. KD’s (nM) for competitive binding of ASOs to P54nrb obtained by BRET and FP. IonisNo Sequence KD (BRET) KD (FP) 660908 GGGAGGAGGACAGCTC 0.51 ± 0 4.601 660956 AACATGTCTGCCAGGG 0.71 ± 0.27 32.95 660923 CCAGCGGGTGGACTGT 1.22 ± 0.37 38.28 661121 CCCTCCCTCCCGCTAG 3.39 ± 0.84 19.53 660992 ACCCTGCCACGATGTA 15.19 ± 0.73 95.06 661115 TCAGTCATGGTATAAA 15.22 ± 1.49 86.93 660899 CTGGGAACTGGTCACC 15.36 ± 3.63 118.8 661103 TGAGAAGATCTCCCGG 15.67 ± 3.01 33.57 661010 CTTCTAGCCACTGGCT 42.69 ± 4.17 325.6 660935 CCACTGCGCAGGCCAG 74.83 ± 13.02 336.9 661073 GTTAAGAAGGCTGTCC 95.54 ± 12.16 259.7 660965 TCCACAGAGCGGACTG 190.6 ± 47.12 309.9 In an attempt to extend the utility of the assay, BRET was performed with crude lysates or in intact cells as detailed in Materials and Methods. The 2’F, MOE, or cEt ASO bound the immunopurified LRPPRC fusion protein with KD’s of 0.35, 0.59 and 1.15 nM respectively (Fig 6A). Affinities for the 2’F and MOE ASO in the crude lysate were similar (KD’s of 0.22 and 0.29 nM), whereas the KD for the cET was slightly lower (0.19 nM) than observed with the purified fusion protein (Fig 6B). Live cells were permeabilized with digitonin (DIG) in the presence of ALEXA-conjugated ASO, followed by addition of NLuc substrate. In these intact, permeablized cells, the BRET affinity was reduced by approximately 10–15 fold relative to the lysate or IP-ed fusion protein (Fig 6C). For other proteins, the difference in ASO affinity between intact cells and immunopurified protein was even more significant (S4 Fig). In general, binding was found to be more significantly reduced in lysates and DIG permeabilized cells for those proteins with lower ASO affinity. This may be due to competition with higher affinity and/or more abundant proteins in the cells and lysate relative to the immunopurified protein, although the actual concentration of the ASO in the permeabilzed cell may be lower than that added to the media. It should also be noted that in the absence of efficient permeabilizing agents (S4C Fig), ASO binding was not detectable. Delivery of ASO to the cells by cationic lipid also did not result in significant BRET signal (data not shown). 10.1371/journal.pone.0161930.g006Fig 6 ASO/BRET in lysates and permeabilized cells. LRPPRC/NLuc was constructed as detailed in Materials and Methods. The fusion protein was expressed by transient transfection in Hela cells. Cell lysate was prepared and the fusion protein immunopreciptated from ½ of the lysate with an antibody to NLuc. For ASO/BRET using the IP’ed protein (A) and cell lysate (B), binding affinities were determined by incubating 106 RLU of LRPPRC/NLuc fusion protein with 3’ ALEXA 594 conjugated 5-10-5 gap-mer ASO at concentrations between 3 pM to 300 nM. Concentration curves were plotted for BRET ratios using GraphPad PRISM software for 2’F (766638, red), MOE (766634, green), or cEt (766636 blue) gap-mer ASOs. For Nano BRET in intact cells (C), the HeLa cells expressing the LRPPRC/NLuc fusion were seeded in 96-well plates at 5000 cells/well. 24 hours after the initiation of transfection, cells were permeabilzed with 25 ng/mL digitonin in OptiMEM media plus 5-10-5 gap-mer ASO at concentrations between 3 pM to 300 nM. After a 15 minute incubation NanoGlo substrate was added and BRET ratios determined as above. Data shown are mean ± SEM from 3 independent experiments. Staufen1 (STAU1)-mediated mRNA decay (SMD) is an mRNA degradation process in mammalian cells that is mediated by the binding of STAU1 to a STAU1-binding site (SBS) within the 3'-untranslated region (3'-UTR) of target mRNAs [22]. STAU1 is a member of the family of double-stranded RNA (dsRNA)-binding proteins involved in the transport and/or localization of mRNAs to different subcellular compartments and/or organelles. These proteins are characterized by the presence of multiple dsRNA-binding domains which are required to bind RNAs having double-stranded secondary structures [23]. An amino-terminal NLuc/STAU1 fusion was evaluated for binding specificity to dsRNA in the BRET assay. A 20 nucleotide, 3’-Alexa-linked RNA was hybridized to a complementary 40-mer RNA (RNA/RNA) or 40-mer DNA of the same sequence (RNA/DNA) (S1 Table). Binding of the RNA/RNA or RNA/DNA duplex, as well as the 3’-Alexa-linked ssRNA and ssDNA 20-mers, was then evaluated by BRET. The affinity of the ds RNA duplex was determined to be approximately 3 nM (red line), in close agreement with the KD previously obtained using purified protein and a filter binding assay [24]. In contrast, STAU1 affinity for ssRNA (blue), ssDNA (black), or RNA/DNA heteroduplex (green) was reduced by ~30–50 fold relative to the affinity for dsRNA, with KD’s ranging from 130–207 nM. The interaction of proteins with DNA is central to the control of many cellular processes including DNA replication, recombination and repair, transcription, and viral assembly. One important technique for studying gene regulation and determining protein:DNA interactions is the electrophoretic mobility shift assay (EMSA) [25]. One advantage of EMSA over other binding assays is that the source of the DNA-binding protein may be a crude nuclear or whole cell extract. Furthermore, EMSA can be used qualitatively to identify sequence-specific DNA-binding proteins (such as transcription factors) in crude lysates and, in conjunction with mutagenesis, to identify the important binding sequences within a given gene’s upstream regulatory region. However, EMSA is relatively low throughput. The AP-1 transcription factor is composed of a mixture of homo- and hetero-dimers formed between Jun and Fos proteins [26]. An amino-terminal NLuc fusion was constructed with the c-Jun protein. We then evaluated binding to an Alexa-conjugated AP1 double stranded DNA consensus binding site in DIG permeabilized cells. Binding to the dsAP1 site was highly specific, with KD’s between 56.6 and 62.9 nM (Fig 7B, red lines). Binding in the BRET assay to AP1 was highly specific, with the affinity for AP1 mutated at the consensus binding site (green lines) or single stranded DNA (blue, black) approximately 2 orders of magnitude less than the ds consensus site. 10.1371/journal.pone.0161930.g007Fig 7 BRET with RNA and DNA binding proteins. A) Staufen 1 specifically binds RNA duplexes. An NLuc/STAU1 fusion was constructed and expressed as detailed in Materials and Methods. The protein was immunoprecipitated from the cell lysate using an STAU1 antibody (Millipore AB5781), then binding affinities were determined by incubating 106 RLU of immunopurified fusion protein with a 3’ ALEXA 594 conjugated PO RNA 20-mer hybridized with a 40 nt complementary RNA strand (RNA/RNA; red) or a 3’ ALEXA 594 conjugated PO DNA 20-mer hybridized with the 40 nt complementary RNA strand (RNA/DNA; green) at concentrations ranging from 10 pM to 3 μM. Binding of the ssRNA (blue) and ssDNA (black) was also evaluated. Data is plotted as BRET ratio vs concentration RNA and represent mean ± SEM from 3 independent experiments. B) C-Jun binding to AP1 dsDNA by BRET. An amino-terminal NLuc fusion was constructed with the C-Jun protein (NLuc/cJun). The fusion protein was expressed by transient transfection in Hela cells which were subsequently seeded in 96 well plates at 5000 cells/well. 24 hours after the initiation of transfection, cells were permeabilzed with 25 ng/mL digitonin in OPtiMEM media plus PO DNA comprising the ds AP1 consensus sequence or a mutant AP1 site (AP1-M) at concentrations between 1 nM to 10 μM. Single stranded DNAs without the complementary DNA strand were also included in the BRET assay. After a 15 minute incubation NLuc substrate was added and BRET ratios determined as detailed above. Concentration response curves and KD’s (nM) for two independent experiments are shown. Discussion In this manuscript we present a novel rapid throughput BRET assay designed to measure nucleic acid/protein interactions. The assay takes advantage of the high intensity luminescence generated with NLuc fusion proteins, which is transferred to an acceptor fluorophore conjugated to an oligonucleotide binding partner. This assay can generate KD’s and relative KD’s for interactions between ASOs and proteins (Fig 1). In addition, information can be generated which provides an estimate of the relative distance between the binding site and the end of the protein. Furthermore, in our experience, the assay is highly reproducible. For example, KD curves for the binding of a 3’ ALEXA 594 conjugated 5-10-5 ASO cEt gap-mer ASO generated using 3 independent preparations of immunopurified p54nrb-Nluc fusion protein, yielded highly similar results with little standard error (S5 Fig). The standard error was typically low for independent replicates in other experiments as well (Figs 1, 3, 4, 6 and 7). A FRET-based assay for the analysis of protein nucleic acids was recently described which relies on genetic incorporation of a fluorescent amino acid into the protein of interest [27]. There are several advantages of the ASO/BRET assay over this and other current methods [28] used to detect and quantify nucleic acid/protein interactions. First, an extensive purification of the protein of interest is not required. This eliminates the need for protein denaturation/renaturation and supports the evaluation of native ASO/protein interactions using immunoprecipitates. Further, for many proteins, crude cell lysates from cells transiently transfected with an NLuc fusion protein can even be used, eliminating the need for the development of specific antibodies to the protein of interest. Since extensive protein purification is not required, the assay also supports evaluation of ASO interactions with protein complexes rather than a single purified protein. In addition, mutant proteins can be rapidly generated by SDM, then quickly expressed and evaluated (Fig 4). It should be noted that as the size of the fused protein increased, the amplitude of the BRET signal was reduced, although this is not always the case (S2 Fig, Table 1). This is likely due to the increased distance between the bound ASO fluorophore and the amino- or carboxy-terminal fused NLuc. The decreased BRET amplitude results in less accurate BRET ratios, and suggests that there is a maximum protein size of which is practical for the assay, but which may also depend on the relative location of the ASO binding site on the protein. The ASO/BRET assay is highly specific and quantitative. We detected no ASO binding to Nluc alone or to NLuc fused to Barnase, a protein not known to interact with nucleic acids (Fig 3A). In addition, the large difference in ASO affinity for known nucleic acid binding proteins (Fig 2, S2 Fig, Table 1), indicates that the KD’s generated in the assay are accurate and specific, and that the assay has a linear dynamic range extending over several orders of magnitude. It is interesting to note that there was little correlation between the number of nucleic acid binding domains present in a given protein and overall affinity for ASOs (Table 1). Finally, PO ASO interactions with NLuc/RNAseH1 (Fig 3B) are consistent with previously published data showing a preference for binding of an RNA/DNA heteroduplex at the RNAse HI HBD [20,21]. We observed large differences in ASO affinity and in the site of interaction of the ASO on the protein based upon the chemistry of the 2’ modification. In general, ASOs with the more hydrophobic 2’ modified bases (2’F>cEt>MOE) bound with higher affinity (Table 1, Fig 2, S2 Fig), however this was not always the case as with RNAse H1 and Nucleolin. The interaction of La protein with 3' poly (U)-rich elements is known to occur primarily through the LaM domain and RRM1 [18], however, the PS ASOs we evaluated also appear to interact with the RRM2 domain as well as LaM and RRM1. (Fig 4A). Similarly, PS ASOs interact with P54nrb at both RRM1 and RRM2 (Fig 4B). Clearly, further studies need to be undertaken to understand the nature and specificity of these interactions; however these results emphasize the utility of this assay. It was also determined that the sequence of an ASO can have profound effects on affinity. For example, affinity of 12 cEt ASOs to P54nrb varied by over 2 orders of magnitude (Fig 5A). Similar sequence dependent differences in affinity were observed for other proteins (S3 Fig), however no clear sequence motifs emerged in the high affinity ASOs and there was little correlation between antisense activity and ASO affinity for any protein (Table 2). Clearly many more sequences must be evaluated in order for any rules to be determined or correlations to be observed. For certain high affinity proteins such as LRPPRC and NCL (Fig 6, S4B Fig) ASO/BRET can be performed with crude lysates or even in intact cells. However, for lower affinity proteins such as RpL5 and P54nrb, a large difference was observed in KD’s using the immunopurified fusion protein relative to performing the assay in intact cells (S4A and S4C Fig). It is unlikely that this difference is due to inefficient cell permeabilization or quenching by the media, since affinities obtained from permeabilized, intact cells and cell lysates were similar, especially for RPL5. More likely, reduced affinities may be the result of interaction of the ASO in the cell with more abundant or higher affinity proteins present in the intact cell or lysate which compete for ASO binding with lower affinity NLuc fusion proteins. It should also be noted that in the absence of efficient permeabilizing agents, ASO binding was not detectable (S4C Fig). Delivery of ASO to the cells by cationic lipid also did not result in significant BRET signal (data not shown). This may be the result of differences in cellular compartmentalization between the transfected ASO and fusion protein, or more likely, transfection does not deliver ASO in high enough concentrations to produce a detectable BRET signal. To expand the utility of the assay beyond ASO/protein interactions, we evaluated interactions of known DNA and RNA binding proteins with oligonucleotides by BRET. Stau1 has been demonstrated to bind double-stranded RNAs [23]. In the BRET assay, STAU1 was shown to bind preferentially to dsRNA over ssRNA or an RNA/DNA heteroduplex (Fig 7A). The affinities obtained by BRET were in close accord with those previously obtained using purified protein and a filter binding assay [24]. BRET was also demonstrated to be a high throughput alternative for transcription factor binding by EMSA. Binding of the NLuc/c-Jun fusion protein was shown to be highly specific to the dsAP1 site, with a KD of approximately 80 nM with virtually no binding to a mutant dsAP1 site or the single stranded DNA (Fig 7B). While this number is somewhat higher than that previously published for binding of AP1 by Fos/Jun heterodimer by EMSA (4-12nM) [29,30], it may be attributable to the much less complex binding environment in the EMSA as compared to the cellular BRET assay. Finally, the high intensity of ASO/BRET raises the potential for quantifying energy transfer between ASOs and specific proteins in individual cells by microscopic imaging. Such experiments would provide insight into ASO uptake, trafficking, and distribution, leading to improvements in specificity and potency of antisense therapeutics. Supporting Information S1 Fig Western blot of representative NLuc fusion proteins. 6 x 105 HeLa cells were transfected with NLuc fusion constructs using Effectene transfection reagent. After 24 hours cells were collected and lysates prepared for Western blot analysis as detailed in Materials and Methods. A) Anti NLuc pAb (Promega). B) Blot in panel A was stripped then re-probed with gene specific antibodies for P54nrb, Millipore 05–950; SFPQ, Abcam Ab38148); NPM1, Abcam Ab10530; C-Jun, Abcam Ab31419; Actin-B, Abcam Ab20272 (PDF) Click here for additional data file. S2 Fig ASO/BRET affinity for La protein. A) La fusion proteins were immunopurified as detailed in Materials and Methods and subsequently incubated with Alexa 594 conjugated 5-10-5 cEt gap-mer ASO at concentrations ranging from 10 pM to 10 μM. BRET ratios were determined for La-NLuc (green, black) or NLuc-La (red, blue), with either a 3’ conjugated (766636) or 5’ conjugated (766635) ASO. Concentration response curves and KD’s (nM) for two independent experiments are shown. B) ASO/BRET binding affinity varies with chemistry of the 2’ modification. ASO/BRET assay was performed with NLuc-La fusion and 5’ conjugated 5-10-5 ASOs at concentrations ranging from 10 pM to 10 μM. 2’F, red; cEt, blue; MOE, green. C) Relative affinities for 2’F (red), cET (blue), and MOE (green) gap-mer ASO as determined by competitive ASO binding to the NLuc-La fusion protein in the BRET assay. 10 nM 3’ Alexa conjugated cEt ASO (766636) was competed with unconjugated 5–10–5 2′-F (red), MOE (green), or cEt (blue) gap-mer ASO at concentrations from 0.1 to 1000 nM. Relative KD’s are shown. Data in panels B and C are mean ± SEM from 3–4 independent experiments. (PDF) Click here for additional data file. S3 Fig Sequence dependent binding of 3-10-3 cEt gap-mer ASOs. NLuc protein fusions were constructed for RNAse H1 (A), Nucleolin (B), and LRPPRC (C), then expressed, and immunopurified as detailed in Materials and Methods. Competitive binding affinities were determined by incubating 106 RLU of immunopurified NLuc fusion protein with 10 nM 3’ ALEXA 594 conjugated 5-10-5 cEt gap-mer ASO along with unconjugated TAAR5 3-10-3 cEt ASOs at concentrations ranging from 3 pM to 3 μM. Concentration curves are plotted for BRET ratios in the presence of each unconjugated ASO. TAAR5 ASO sequences and KD’s can be found in S4 Table. (PDF) Click here for additional data file. S4 Fig ASO NanoBRET in permeabilized cells. P54nrb-NLuc (A), NLuc-NCL (B), and NLuc/RPL5 (C) fusion proteins were expressed by transient transfection in Hela cells. Cell lysates were prepared and the fusion protein immunopreciptated with an antibody to NLuc. For ASO NanoBRET using the IP’ed protein (red) or crude lysates (green), binding affinities were determined by incubating 106 RLU of fusion protein with a 3’ ALEXA 594 conjugated 5-10-5 cEt gap-mer ASO at the indicated concentrations. Concentration curves were plotted for BRET ratios using GraphPad PRISM software. For Nano BRET in intact cells, the HeLa cells expressing the fusion protein were seeded in 96-well plates at 5000 cells/well. 24 hours after the initiation of transfection, cells were permeabilzed with 25–50 ng/mL digitonin in OptiMEM (blue) or PBS (blue dashed) plus the cEt gap-mer ASO at concentrations between 1 nM and 1 uM. After a 30 minute incubation NLuc substrate was added and BRET ratios determined as above. For thr RPL5 fusion, cells were also treated in OptiMEM in the presence of 0.2% saponin (black) or in OptiMEM with no permeabilizing agent (brown). (PDF) Click here for additional data file. S5 Fig ASO NanoBRET reproducibility. A) P54nrb/Nluc fusion protein was expressed by transient transfection of 3 different plates of Hela cells. Cell lysates were prepared and the fusion proteins immunoprecipitated with an antibody to P54nrb. Binding affinities were determined in separate experiments by incubating 106 RLU of fusion protein with a 3’ ALEXA 594 conjugated cEt gap-mer ASO at concentrations ranging from 10 pM to 3 uM. Binding curves are shown for each of the biological replicates and as well as the mean ± SEM for the 3 samples (black). (PDF) Click here for additional data file. S1 Table ASOs used in study. For gap-mer ASOs, 2’ modified bases (2’-alpha-flouro, (S)-cEt, and MOE) are indicated by bold type. The Ionis numbers of the gap-mer ASO are in parentheses. For AP1 oligonucleotides, the AP1 consensus site is underlined and consensus binding site mutations are indicated in red. (PDF) Click here for additional data file. S2 Table Sequences of PCR primers used to generate cDNAs for directional in-frame cloning with NLuc. (PDF) Click here for additional data file. S3 Table Sequences of SDM primers used to generate deletion mutants for p54nrb, La, and RNAse H1. (PDF) Click here for additional data file. S4 Table TAAR5 ASOs activity and protein affinity. 20,000 MEF cells were treated with 3.5 uM ASO by electroporation in 96 well pltes in a total volume of 100 ul. 24 hours later total RNA was purified and levels of TAAR5 mRNA accessed by qRT/PCR as described previously [11]. qRT/PCR FP: TGCTACCAGGTGAATGGGTCTT, RP: TGCGCAGGCCAGATAGATG, probe: AGGACAGTCCACCCGCTGGCC. For each ASO data is presented as the percent of mock treated control for 3 replicates. KD’s (nM) for NanoBRET ASO binding were determind us GraphPad PRISM software. (PDF) Click here for additional data file. The authors wish to thank Matt Robers and Richard Somberg for guidance with initial ASO BRET parameters and Tracy Reigle for help with preparation of the figures. ==== Refs References 1 Luscombe NM , Laskowski RA , Thornton JM (2001 ) Amino acid–base interactions: a three-dimensional analysis of protein–DNA interactions at an atomic level . Nucleic Acids Research 29 : 2860 –2874 . 11433033 2 Jayaram B , Jain T (2004 ) The Role of Water in Protein-DNA Recognition . Annual Review of Biophysics and Biomolecular Structure 33 : 343 –361 . 15139817 3 Tolstorukov MY , Jernigan RL , Zhurkin VB (2004 ) Protein–DNA Hydrophobic Recognition in the Minor Groove is Facilitated by Sugar Switching . Journal of Molecular Biology 337 : 65 –76 . 15001352 4 Koh YY , Wang Y , Qiu C , Opperman L , Gross L , Tanaka Hall T , et al (2011 ) Stacking interactions in PUF–RNA complexes . RNA 17 : 718 –727 . 10.1261/rna.2540311 21372189 5 Rhodes D , Burley SK (2000 ) Protein–nucleic acid interactions: Editorial overview . Current Opinion in Structural Biology 10 : 75 –77 . 10766517 6 Liang X-h , Sun H , Shen W , Crooke ST (2015 ) Identification and characterization of intracellular proteins that bind oligonucleotides with phosphorothioate linkages . Nucleic Acids Research 43 : 2927 –2945 . 10.1093/nar/gkv143 25712094 7 Hall MP , Unch J , Binkowski BF , Valley MP , Butler BL , Wood M , et al (2012 ) Engineered Luciferase Reporter from a Deep Sea Shrimp Utilizing a Novel Imidazopyrazinone Substrate . ACS Chemical Biology 7 : 1848 –1857 . 10.1021/cb3002478 22894855 8 Machleidt T , Woodroofe CC , Schwinn MK , Méndez J , Robers MB , Zimmerman K , et al (2015 ) NanoBRET—A Novel BRET Platform for the Analysis of Protein–Protein Interactions . ACS Chemical Biology 10 : 1797 –1804 . 10.1021/acschembio.5b00143 26006698 9 Masuho I , Ostrovskaya O , Kramer GM , Jones CD , Xie K , Martemyanov KA (2015 ) Distinct profiles of functional discrimination among G proteins determine the actions of G protein–coupled receptors . Science Signaling 8 : ra123 –ra123 . 10.1126/scisignal.aab4068 26628681 10 Wang J , Ren J , Wu B , Feng S , Cai G , Tuluc F , et al (2015 ) Activation of Rab8 guanine nucleotide exchange factor Rabin8 by ERK1/2 in response to EGF signaling . Proceedings of the National Academy of Sciences 112 : 148 –153 . 11 Vickers TA , Crooke ST (2015 ) The rates of the major steps in the molecular mechanism of RNase H1-dependent antisense oligonucleotide induced degradation of RNA . Nucleic Acids Research 43 : 8955 –8963 . 10.1093/nar/gkv920 26384424 12 Vickers TA , Crooke ST (2014 ) Antisense Oligonucleotides Capable of Promoting Specific Target mRNA Reduction via Competing RNase H1-Dependent and Independent Mechanisms . PLoS ONE 9 (10 ): e108625 10.1371/journal.pone.0108625 25299183 13 Vickers TA , Lima WF , Nichols JG , Crooke ST (2007 ) Reduced levels of Ago2 expression result in increased siRNA competition in mammalian cells . Nucleic Acids Research 35 : 6598 –6610 . 17905815 14 Shen W , Liang X-h , Crooke ST (2014 ) Phosphorothioate oligonucleotides can displace NEAT1 RNA and form nuclear paraspeckle-like structures . Nucleic Acids Research 42 : 8648 –8662 . 10.1093/nar/gku579 25013176 15 Shen W , Liang X-h , Sun H , Crooke ST (2015 ) 2′-Fluoro-modified phosphorothioate oligonucleotide can cause rapid degradation of P54nrb and PSF . Nucleic Acids Research 43 : 4569 –4578 . 10.1093/nar/gkv298 25855809 16 Chen H , Puhl HL , Koushik SV , Vogel SS , Ikeda SR (2006 ) Measurement of FRET Efficiency and Ratio of Donor to Acceptor Concentration in Living Cells . Biophysical Journal 91 : L39 –L41 . 16815904 17 Shav-Tal Y , Zipori D (2002 ) PSF and p54nrb/Nono–multi-functional nuclear proteins . FEBS Letters 531 : 109 –114 . 12417296 18 Alfano C , Sanfelice D , Babon J , Kelly G , Jacks A , et al (2004 ) Structural analysis of cooperative RNA binding by the La motif and central RRM domain of human La protein . Nat Struct Mol Biol 11 : 323 –329 . 15004549 19 Bousquet-Antonelli C , Deragon J-M (2009 ) A comprehensive analysis of the La-motif protein superfamily . RNA 15 : 750 –764 . 10.1261/rna.1478709 19299548 20 Wu H , Lima WF , Crooke ST (1999 ) Properties of Cloned and Expressed Human RNase H1 . J Biol Chem 274 : 28270 –28278 . 10497183 21 Wu H , Lima WF , Crooke ST (2001 ) Investigating the Structure of Human RNase H1 by Site-directed Mutagenesis . Journal of Biological Chemistry 276 : 23547 –23553 . 11319219 22 Park E , Maquat LE (2013 ) Staufen-mediated mRNA decay . Wiley interdisciplinary reviews RNA 4 : 423 –435 . 10.1002/wrna.1168 23681777 23 Wickham L , Duchaîne T , Luo M , Nabi IR , DesGroseillers L (1999 ) Mammalian Staufen Is a Double-Stranded-RNA- and Tubulin-Binding Protein Which Localizes to the Rough Endoplasmic Reticulum . Molecular and Cellular Biology 19 : 2220 –2230 . 10022909 24 Marión RM , Fortes P , Beloso A , Dotti C , Ortín J (1999 ) A Human Sequence Homologue of Staufen Is an RNA-Binding Protein That Is Associated with Polysomes and Localizes to the Rough Endoplasmic Reticulum . Molecular and Cellular Biology 19 : 2212 –2219 . 10022908 25 Hellman LM , Fried MG (2007 ) Electrophoretic Mobility Shift Assay (EMSA) for Detecting Protein-Nucleic Acid Interactions . Nature protocols 2 : 1849 –1861 . 17703195 26 Mechta-Grigoriou F , Gerald D , Yaniv M (2001 ) The mammalian Jun proteins: redundancy and specificity . Oncogene 20 : 2378 –2389 . 11402334 27 Park H , Kang H , Ko W , Lee W , Jo K , et al (2015 ) FRET-based analysis of protein-nucleic acid interactions by genetically incorporating a fluorescent amino acid . Amino Acids 47 : 729 –734 . 10.1007/s00726-014-1900-2 25540052 28 Dey B , Thukral S , Krishnan S , Chakrobarty M , Gupta S , Manghani C , et al (2012 ) DNA–protein interactions: methods for detection and analysis . Mol Cell Biochem 365 : 279 –299 . 10.1007/s11010-012-1269-z 22399265 29 Deng T , Karin M (1993 ) JunB differs from c-Jun in its DNA-binding and dimerization domains, and represses c-Jun by formation of inactive heterodimers . Genes & Development 7 : 479 –490 .8383624 30 Rauscher FJ , Voulalas PJ , Franza BR , Curran T (1988 ) Fos and Jun bind cooperatively to the AP-1 site: reconstitution in vitro . Genes & Development 2 : 1687 –1699 .2467839

PMC005xxxxxx/PMC5003357.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757097210.1371/journal.pone.0161008PONE-D-16-14790Research ArticleBiology and Life SciencesGeneticsFungal GeneticsFungal GenomicsBiology and Life SciencesMycologyFungal GeneticsFungal GenomicsBiology and Life SciencesGeneticsGenomicsFungal GenomicsBiology and Life SciencesOrganismsFungiResearch and Analysis MethodsDatabase and Informatics MethodsBiological DatabasesGenomic DatabasesBiology and Life SciencesComputational BiologyGenome AnalysisGenomic DatabasesBiology and Life SciencesGeneticsGenomicsGenome AnalysisGenomic DatabasesBiology and Life SciencesMicrobiologyMedical MicrobiologyMicrobial PathogensFungal PathogensMedicine and Health SciencesPathology and Laboratory MedicinePathogensMicrobial PathogensFungal PathogensBiology and Life SciencesMycologyFungal PathogensBiology and Life SciencesGeneticsFungal GeneticsBiology and Life SciencesMycologyFungal GeneticsResearch and Analysis MethodsDatabase and Informatics MethodsBiological DatabasesSequence DatabasesBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesSequencing TechniquesSequence AnalysisSequence DatabasesResearch and Analysis MethodsMolecular Biology TechniquesSequencing TechniquesSequence AnalysisSequence DatabasesBiology and Life SciencesComputational BiologyGenome AnalysisGenomic LibrariesBiology and Life SciencesGeneticsGenomicsGenome AnalysisGenomic LibrariesBiology and life sciencesBiochemistryProteinsDNA-binding proteinsTranscription FactorsBiology and Life SciencesGeneticsGene ExpressionGene RegulationTranscription FactorsBiology and Life SciencesBiochemistryProteinsRegulatory ProteinsTranscription FactorsGenomic Analyses of Cladophialophora bantiana, a Major Cause of Cerebral Phaeohyphomycosis Provides Insight into Its Lifestyle, Virulence and Adaption in Host Cladophialophora bantiana Isolated from Brain Abscesshttp://orcid.org/0000-0001-5828-7065Kuan Chee Sian 1Cham Chun Yoong 23http://orcid.org/0000-0002-1210-1138Singh Gurmit 2Yew Su Mei 1http://orcid.org/0000-0001-6221-6630Tan Yung-Chie 4Chong Pei-Sin 4Toh Yue Fen 1http://orcid.org/0000-0002-2547-4996Atiya Nadia 1Na Shiang Ling 1Lee Kok Wei 4Hoh Chee-Choong 4http://orcid.org/0000-0001-9196-0362Yee Wai-Yan 4Ng Kee Peng 1*1 Department of Medical Microbiology, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia2 Department of Neurosurgery, Hospital Pulau Pinang, Jalan Residensi, Georgetown, Pulau Pinang, Malaysia3 Department of Surgery, Neurosurgical Division, University of Malaya, Kuala Lumpur, Malaysia4 Codon Genomics SB, Selangor Darul Ehsan, MalaysiaStajich Jason E EditorUniversity of California Riverside, UNITED STATESCompeting Interests: Authors YCT, PSC, CCH, and WYY are employed by Codon Genomics SB. The other authors named on the manuscript are from University of Malaya. All other authors have declared that no competing interests exist. These affiliations do not alter the authors' adherence to all the PLOS ONE policies on sharing data and materials. Conceived and designed the experiments: KPN CSK. Performed the experiments: CSK YFT SMY SLN. Analyzed the data: CSK YCT PSC WYY CCH. Contributed reagents/materials/analysis tools: YCT PSC WYY CCH. Wrote the paper: CSK NA CYC GS KPN. Provide clinical specimen and clinical history: CYC GS. * E-mail: kpng@ummc.edu.my29 8 2016 2016 11 8 e016100813 4 2016 28 7 2016 © 2016 Kuan et al2016Kuan et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Cladophialophora bantiana is a dematiaceous fungus with a predilection for causing central nervous system (CNS) infection manifesting as brain abscess in both immunocompetent and immunocompromised patients. In this paper, we report comprehensive genomic analyses of C. bantiana isolated from the brain abscess of an immunocompetent man, the first reported case in Malaysia and Southeast Asia. The identity of the fungus was determined using combined morphological analysis and multilocus phylogeny. The draft genome sequence of a neurotrophic fungus, C. bantiana UM 956 was generated using Illumina sequencing technology to dissect its genetic fundamental and basic biology. The assembled 37.1 Mb genome encodes 12,155 putative coding genes, of which, 1.01% are predicted transposable elements. Its genomic features support its saprophytic lifestyle, renowned for its versatility in decomposing hemicellulose and pectin components. The C. bantiana UM 956 was also found to carry some important putative genes that engaged in pathogenicity, iron uptake and homeostasis as well as adaptation to various stresses to enable the organism to survive in hostile microenvironment. This wealth of resource will further catalyse more downstream functional studies to provide better understanding on how this fungus can be a successful and persistent pathogen in human. High Impact Research MoE GrantH-20001-00-E000070Ng Kee Peng This study was supported by High Impact Research MoE Grant UM.C/625/1/HIR/MOHE/MED/31 (Account no. H-20001-00-E000070) from the Ministry of Education Malaysia. Codon Genomics SB provided support in the form of salaries for authors KWL, CCH, and WYY, but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of these authors are articulated in the ‘author contributions’ section. Data AvailabilityThe ITS, SSU, and LSU nucleotide sequences of C. bantiana UM 956 were deposited in GenBank with accession numbers KU928131, KU928132, and KU928133, respectively. The assembly of the C. bantiana UM 956 was deposited in European Nucleotide Archive with BioProject number PRJEB13102 (FJVA01000000).Data Availability The ITS, SSU, and LSU nucleotide sequences of C. bantiana UM 956 were deposited in GenBank with accession numbers KU928131, KU928132, and KU928133, respectively. The assembly of the C. bantiana UM 956 was deposited in European Nucleotide Archive with BioProject number PRJEB13102 (FJVA01000000). ==== Body Introduction Phaeohyphomycosis (photo–Greek for dark) refers to mycotic infections caused by dematiaceous fungi. These fungi share a common feature of being darkly pigmented due to the presence of dihydroxynaphthalene melanin in their cell walls [1]. Till now, over 150 species and 70 genera of dematiaceous fungi are reported to cause human diseases, ranging from superficial infections to life-threatening infections, such as brain abscess and pulmonary infections [2]. Invasive and systemic phaeohyphomycosis is rare but its causative agents are increasingly recognized as a cause of serious disease such as CNS infections. CNS fungal infections are uncommon as the occurrence is very less and usually associated with devastating consequences in immunocompetent individuals [3]. CNS infections which are caused by fungi can cause one or more the symptoms such as acute or chronic meningitis, encephalitis, abscesses, or myelopathy [4, 5]. Cryptococcus neoformans is the predominant causes of cryptococcosis in immunocompetent individuals, while Zygomycetes, Aspergillus, and Candida species are the most common causes of such infection in immunocompromised patients [6, 7]. Additionally, a few dematiaceous fungi such as C. bantiana, Exophiala dermatitidis, Ochroconis gallopava and Rhinocladiella mackenziei are recognized as causative agents of primary CNS phaeohyphomycosis [2, 8]. C. bantiana, E. dermatitidis, O. gallopava and R. mackenziei are known as true neurotropic fungi [9]. Life-threatening CNS fungal infections are commonly associated with an immunocompromised state. Immunocompromised individuals with organ transplantations and acquired immune deficiency syndrome are susceptible to acquire the fungal infection, particularly in brain or meninges. However, primary cerebral phaeohyphomycosis caused by C. bantiana appears to be an exception to this rule, occurring more commonly in immunocompetent than in immunocompromised patients. Dixon et al. [10] used animal experiments to prove that the fungus is carried through bloodstream to the CNS, suggesting that it is disseminated via the haematogenous route to CNS [8, 11, 12]. Amongst the previous cases of cerebral phaeohyphomycosis, C. bantiana is responsible for causing 48% of the cases and associated with a high mortality rate of up to 70% and there are currently no standardized recommendations for treatment [8]. CNS infections caused by C. bantiana usually present with brain abscess, either single or multiple lesions [8, 13, 14]. Nonetheless, C. bantiana rarely causes cutaneous or subcutaneous infections [15]. The advancement and development of sequencing technology and bioinformatics have led to the generation of several neurotropic fungal genomes such as Cryptococcus gattii [16] and C. neoformans [17]. However, at present, the publicly available genome sequence of C. bantiana remains deficient. Additionally, relatively little is known about molecular mechanisms of pathogenicity and adaptibility of this neurotropic fungus in blood and cerebrospinal fluid (CSF) of the human body. At present, CNS infection caused by C. bantiana have not been reported in Southeast Asia, although the infections have increasingly been reported abroad [8, 11, 12, 14, 18–21], especially in Europe [12]. In this study, we describe a case of brain abscess caused by C. bantiana in an immunocompetent man which was successfully treated with surgical excision combined with voriconazole. Additionally, the draft genome of the C. bantiana was generated by a combined assembly of two different insert-size Illumina sequencing libraries (5-kb insert-size library and 500-bp insert-size library). To our knowledge, this report is the first comprehensive description of the C. bantiana genome. The thorough analysis of the neurotropic fungal genome will serve as a platform to further understand its basic biology, pathogenicity as well as adaptability in human host. Results and Discussion Clinical history A 49-year-old Malay male came with complaints of complex partial seizure, right sided weakness, fever, and headache. A computerized tomography (CT) of the brain showed a rim enhancing lesion over the left parietal region close to the motor strip measuring 4.4 cm (AP) × 2.8 cm (W) × 4.0 cm (H) with mass effect and surrounding edema (S1 Fig). Shortly, he developed slurring of speech, confusion resulting in deterioration of GCS to E1V1M5 (7/15). No history of recent transplantation, trauma and any history suggestive of immunocompromised state was elicited. Inflammatory markers (C-reactive protein and total white cells) were raised. A provisional diagnosis of pyogenic cerebral abscess was made and he underwent an image-guided aspiration of the abscess, in view of its close proximity to the motor strip under cover of intravenous cetriaxone. A total of 25 mL of viscid, yellow pus was aspirated. Gram and Ziehl-Neelsen smears of the pus did not show any bacteria or acid fast structures. Blood and sputum cultures were sterile. Fungal elements (septate and darkly pigmented fungal hyphae) were observed by a direct microscopic wet mount examination (40% potassium hydroxide). The initial intravenous cetriaxone was changed to empirical intravenous amphotericin B 35 mg OD and oral itraconazole 200 mg BD. The fungal isolate was identified as C. bantiana based on the typical morphological features and multilocus phylogenetic analysis. The patient completed a combination of intravenous amphotericin B and oral itraconazole for 45 days after underwent craniotomy and an uneventful excision. He was then discharged home with oral voriconazole for six months. Repeat contrasted CT brain after completion of antifungal treatment showed resolution of the abscess and recovery of motor power of the affected limbs allowing him to perform activities of daily living (power of 4/5). Possible transmission route of C. bantiana The environmental niche of C. bantiana is unknown and has seldom been isolated from nature other than nonhuman source. The occupational association of the infection by this neurotropic fungus with farming is suggestive of its origin from this environment as soil fungus [2, 8]. Relatively few reports of its isolation from the environment are on record [22]. A definite relation between an environmental and clinical strain still has to be proven. The patient is a lorry driver, whose payload involves soil, suggesting that soil is the most probable source of his infection, either through a trivial skin wound or inhalation of spores, followed by haematogenous dissemination to the brain. The findings of this case agree with previously reported predisposition of C. bantiana to people working with soil [8, 23]. Revankar et al. (2004) also revealed many cases of immunocompetent patients occur in rural settings, possibly due to more prevalent soil exposure [8]. Morphological characterization of C. bantiana UM 956 The growth rate of C. bantiana UM 956 on SDA was moderate and it takes around 2–3 days of incubation at 30°C to observe visible growth. The colonies mature and reached diameters of 30 mm after seven days of incubation at 30°C. The C. bantiana UM 956 colonies were compact, velvety textured and olivaceous to brown discolouration with a black undersurface (Fig 1A and 1B). The fungus did not produce any diffusible pigment. Microscopic examination with lactophenol cotton blue staining showed dark septate hyphae with sparsely branched conidiophores producing long, strongly coherent and wavy chains of conidia (Fig 1C). The conidia (1–5 × 1–2 μm) were pale olivaceous, ellipsoidal to spindle-shaped (Fig 1C). No chlamydospores were observed. On the basis of these characteristics, the isolate was found to be morphologically related to C. bantiana [24]. 10.1371/journal.pone.0161008.g001Fig 1 Colonial characteristic and microscopic morphology of C. bantiana. The (A) surface and (B) close-up view of the colonial morphology of C. bantiana after being cultured for seven days. Light micrograph showing (C) smooth walled, pale olivaceous, ellipsoidal to spindle-shaped conidia arranged in long, strongly coherent chains (400× magnification, bars 20 μm). Sequence-based identification and multilocus phylogenetic analysis The preliminary morphological identification of the UM 956 isolate was confirmed by PCR amplification of the ITS, SSU and LSU gene regions, followed by BLASTn search against those (ITS, SSU, and LSU nucleotide sequences) deposited in the NCBI-nucleotide database. The ITS sequence of UM 956 showed 100% (558/558) identical to the C. bantiana strain PWQ2235 isolate ISHAM-ITS_ID MITS1144. However, the SSU and LSU were 92% (925/1005) and 93% (1367/1470 bp) identical to the Chaetothyriales sp. TRN486 and the Cladophialophora carrionii isolate CBS 160.54, respectively. Multilocus phylogenetic analysis was subsequently used to identify UM 956 to the species-level. The sequenced ITS and LSU gene regions were used to construct a phylogram using combined gene analysis with an additional 12 ex-type strains of the Cladophialophora species (Table 1). The SSU gene region was excluded in the multilocus phylogenetic analysis due to the poor alignment between UM 956 and other Cladophialophora species (including other strains of C. bantiana). The multilocus phylogenetic tree consisted of members from the genus Cladophialophora and the Cladophialophora species are well separated (Fig 2). The analysis revealed that UM 956 was tightly clustered together with C. bantiana type strain CBS 173.52 (Fig 2). The high statistical support (1.0 Bayesian posterior probability) for the placement of UM 956 with C. bantiana type strain CBS 173.52 confirms its identity as C. bantiana. 10.1371/journal.pone.0161008.g002Fig 2 Bayesian phylogram generated using the combined gene sequences of ITS and LSU. The tree was rooted with Plococarpus schaereri AFTOL-ID 2289 as outgroup. The numbers on the nodes indicate Bayesian posterior probability based on 100 sampling frequencies for a total of 100,000 generations. 10.1371/journal.pone.0161008.t001Table 1 Details of Isolates Subjected to Multilocus Phylogenetic Analysis. ITS LSU Organism name Strain Origin Country KU928131 KU928133 Cladophialophora bantiana UM 956 Brain abscess Malaysia EU103989 KF155189 Cladophialophora bantiana CBS 173.52 Brain abscess USA EU103995 KC809995 Cladophialophora emmonsii CBS 640.96 Sub-cutaneous lesion, cat Netherlands EU103985 KC809989 Cladophialophora devriesii CBS 147.84 Disseminated infection USA EU137293 KC809991 Cladophialophora mycetomatis CBS 454.82 Culture contaminant Netherlands EU035406 KF928513 Cladophialophora chaetospira CBS 514.6 Wheat field soil Germany EU035403 KF928514 Cladophialophora chaetospira CBS 114747 Phyllostachys bambusoides China EU137322 KC809994 Cladophialophora yegresii CBS 114405 Stenocereus griseus Venezuela GU225939 KF928485 Cladophialophora modesta CBS 985.96 Brain USA FJ372388 FJ372405 Cladophialophora proteae CPC 14902 Encephalartos altensteinii South Africa EU137292 KF928518 Cladophialophora carrioni CBS 260.83 Skin lesion USA EU137266 KF928517 Cladophialophora carrioni CBS 160.54 Chromoblastomycosis Australia EU137268 KF928516 Cladophialophora carrioni CBS 114393 Chromoblastomycosis hand lesion Venezuela - EF643766 Placocarpus schaereri AFTOL-ID 2289 - USA Acquisition of the C. bantiana UM 956 genome sequence The genome of C. bantiana UM 956, was sequenced using Illumina HiSeq 2000 system. A total of 28,594,774 paired reads (2.57 Gb) of a 500-bp insert-size library and 12,972,712 paired reads (1.17 Gb) of a 5-kb insert-size library were generated by Illumina HiSeq 2000 Sequencing system. The sequencing coverage for the combined sequenced reads is 90-fold. The total assembly size of the genome is 37.1 Mb. The draft genome sequence consists of 439 contigs (≥ 200 bp) and orientated within 84 scaffolds (≥1,000 bp) (Table 2). The genome had an average GC content of 50.65%. A summary of the principal genome sequence data is provided in Table 2. 10.1371/journal.pone.0161008.t002Table 2 C. bantiana UM 956 genomic and assembly features. Details Paired-end and mate pair combined (500-bp and 5-kb) Sequencing depth ~90× Total length of sequences (bp) 37,087,895 Total number of contigs (≥200 bp) 439 Contigs N50 (bp) 172,990 Contigs GC content (%) 51.27 Total number of scaffolds (≥1,000 bp) 84 Scaffolds N50 (bp) 3,618,913 Scaffolds GC content (%) 51.19 tRNAs 58 8s rRNA 17 16s rRNA 1 28s rRNA 1 Number of predicted genes (≥ 99bp) 12,155 Annotated protein coding regions (nr) 10,543 Annotated protein coding regions (SwissProt) 1,464 Annotated protein coding regions (Interpro) 9,324 Hypothetical proteins 8,444 Gene content A total of 58 tRNAs and 19 rRNAs were identified in the UM 956 genome. Within the genome, a total of 12,155 putative genes (≥ 33 amino acids) were predicted. These putative genes consist of 48.5% of the assembly (1 gene per 3.1 kb). On average, there are 2.25 exons per gene and average size of protein coding genes is 1,481 bp in the genome. The basic genomic statistics, assembly size and the total number of predicted genes in the C. bantiana UM 956 genome were compared to other pathogenic dematiaceous fungi (Table 3, S1 Table). The genome size of C. bantiana UM 956, number of putative genes and gene density are comparable to the recently sequenced dematiaceous fungi (Table 3). 10.1371/journal.pone.0161008.t003Table 3 Genome content of C. bantiana UM 956 and other previously sequenced fungal genomes. Species Assembly size (Mb) Predicted genes Gene density (gene/10 kb) GC content (%) Isolation site Reference C. bantiana UM 956 37.1 12,155 3.3 51.27 Brain abscess In this study Bipolaris papendorfii UM 226 33.4 11,015 3.29 50.65 Skin scraping [26] Daldinia eschscholtzii UM 1020 35.5 11,120 3.1 46.81 Blood [27] Daldinia eschscholtzii UM 1400 35.8 10,822 3.0 46.8 Skin scraping [27] Ochroconis mirabilis UM 578 34.6 13,435 3.88 52.1 Skin scraping [28] Pyrenochaeta sp. UM 256 35.5 12,545 3.53 50.4 Skin scraping [29] Sporothrix schenckii strain 1099–18 32.4 10,293 3.17 54.96 Subcutaneous tissue [30] Exophiala dermatitidis NIH/UT8656 26.4 9269 3.51 51.51 Unknown [31] A total of 10,543, 1,464, and 9,324 gene-coding sequences are homologous to known proteins in the NCBI nr, SwissProt, and InterPro databases, respectively (S2 Table). A total of 8,444 hypothetical proteins (50% identity and 70% coverage cut-off) were identified based on the top hit of the BLAST result against NCBI nr database. We found that more than 95% (8085 sequences) of these hypothetical proteins hits to proteins sequences from C. psammophila CBS110553. InterPro protein sequence analysis revealed at least one domain was predicted in 6,527 hypothetical proteins in C. bantiana UM 956 (S3 Table). It is noted that C. bantiana UM 956 has 150 hypothetical proteins engaged in cytochrome P450 (CYP) superfamily. These enzymes perform various reactions in a wide variety of physiological processes, such as biosynthesis of secondary metabolites, detoxification, and degradation of xenobiotics [25]. The genome was further mapped to the Eukaryotic Clusters of Orthologs (KOG) database to further characterize the putative proteins. A total of 7,185 (59.1% of total predicted genes) of protein-coding genes were mapped in the KOG database and were classified to 26 different functional groups (Fig 3 and S4 Table). The UM 956 genome contains large amount of putative genes (1,728 genes) that were not categorized to a distinct group (categories “General functions prediction only” and “Function unknown”). Apart from the poorly characterized categories: categories “General functions prediction only” and “Function unknown”, the top five most abundant KOG groups were “Secondary metabolites biosynthesis, transport and catabolism” (578 genes), “Post-translational modification, protein turnover, chaperones” (499 genes), “Lipid transport and metabolism” (486 genes), “Energy production and conversion” (467 genes), and “Amino acid transport and metabolism” (368 genes). In contrast to other dematiaceous fungi, such as B. papendorfii UM 226 (334 genes), Pyrenochaeta sp. UM 256 (395 genes), Ochroconis mirabilis UM 578 (431 genes), D. eschscholtzii UM 1020 (358 genes), and D. eschscholtzii UM 1400 (356 genes), the predominant of the functionally annotated genes in UM 956 genome was engaged in biosynthesis and catabolism of secondary metabolites. The metabolism and biosynthesis of secondary metabolites in C. bantiana UM 956 are further discussed in the subsection of “Secondary metabolism”. 10.1371/journal.pone.0161008.g003Fig 3 KOG class annotation distribution of C. bantiana UM 956 genome. A total of 7,185 of protein-coding genes were annotated by KOG. The proteins were assigned into different KOG functionary categories as shown in the pie chart. Transposable elements Genomic plasticity allows organisms to survive and adapt to diverse environments, which is particularly related to pathogens to engage in novel niche. Transposable elements (TEs) have been proposed to actively promote genomic rearrangements [32], although their exact role in the evolution genomic DNA remain unknown. Dematiaceous fungal genomes contain greatly different amounts of TEs. B. papendorfii UM 226, Cochliobolus heterostrophus race O genome, Sporothrix schenckii, Sporothrix brasiliensis, O. mirabilis UM 578, D. eschscholtzii strains UM 1400 and UM 1020 contain 2.49%, 5.9%, 0.34%, 0.62%, 0.01%, 1.02%, and 1.42%, respectively [26, 27, 30, 33, 34]. In this study, the TEs occupy 1.01% (374,655 bp) of the C. bantiana UM 956 genome (Table 4). The class I, retrotransposons comprised 0.94% of the genome, whereas class II transposons, DNA transposons comprised only 0.07% of the genome. As previously noted in dematiaceous fungal pathogens [26, 30, 34], LTR (Long Terminal Repeat) is the major type of retrotransposons, in which the Gypsy-like element is more than Copia-like element. Our bioinformatics prediction revealed that 0.86% and 0.0058% of the sequenced UM 956 genome consists of gypsy and TY1_Copia, respectively. Furthermore, one copy of piggyBac transposable element was identified in UM 956. This transposable element was isolated from the cabbage looper moth Trichoplusia ni. PiggyBac transposon has been used as transposon-based mutagenesis tool for fission yeast [35] as well as for the genetic characterization of mammalian genomes [36–38]. 10.1371/journal.pone.0161008.t004Table 4 Putative transposable elements in the genome sequence of C. bantiana UM 956. Class Family name Total number Total bases Percentage of assembled genome I DDE_1 24 19,761 0.0533% gypsy 173 319,870 0.8625% LINE 8 5596 0.0151% TY1_Copia 5 2164 0.0058% II helitronORF 2 1363 0.0037% hAT 8 10657 0.0287% mariner 12 10680 0.0288% mariner_ant1 3 1587 0.0043% MuDR_A_B 3 1599 0.0043% cacta 1 89 0.0002% piggybac 1 1289 0.0035% Total 240 374,655 1.01% KEGG pathway analysis of C. bantiana UM 956 KEGG pathway analysis was carried out to further gain insight into the gene functions in C. bantiana UM 956. A total of 2,506 predicted proteins was assigned to their orthologous genes in metabolic pathways in the KEGG database. The full list of KEGG pathway annotation was listed in S5 Table. The top five categories in KEGG metabolic pathway are carbohydrate metabolism, amino acid metabolism, lipid metabolism, xenobiotics biodegradation and metabolism, and nucleotide metabolism (Fig 4). In particular, there are 590 unique reactions corresponding to carbohydrate metabolism. C. bantiana UM 956 contains many enzymes involved in glycolysis, gluconeogenesis, citrate cycle (TCA cycle), and other essential carbohydrate metabolism pathways, which suggest this fungus is capable in metabolize glucose, sucrose, galactose, fructose, mannose pyruvate, and starch (S5 Table). The potential of using sugars other than glucose as well as sulfur, nitrogen, and peptide nitrogen sources may partly explain the ubiquitous nature of C. bantiana in the environment. 10.1371/journal.pone.0161008.g004Fig 4 KEGG classifications of proteins in C. bantiana UM 956 genome. The proteins were assigned into different KEGG metabolic pathway categories as shown in the pie chart. A total of 2,506 protein-coding genes were involved in metabolic pathway based on the KEGG database. Carbohydrate-active enzymes The plant cell wall is composed of polysaccharides cellulose, hemicellulose, and pectin, which is important as a nutrient source for plant pathogens and saprophytes and act as physical barrier to plant pathogens. Fungi can produce diverse carbohydrate-active enzymes (CAZymes) to degrade enormous functional and structural diversity of complex plant polysaccharide materials for carbon sources [39]. Thus, CAZymes can be powerful reporters of the fungal lifestyle. C. bantiana has been reported widely distributed in soil and woody plant materials and readily as saprophyte, although the precise niche of the fungus remains unidentified [8, 22, 40]. As a matter of fact, saprophytes degrade plant cell wall materials to obtain nutrients for growth. In this work, we have probed the CAZyme repertoires of the C. bantiana UM 956 and compared with other 12 fungi of different lifestyles. Differences in the number and distribution of CAZymes among C. bantiana UM 956, non-plant pathogens (biotrophic, saprophytic, symbiotic fungi), and plant pathogens (necrotrophic, hemi-biotrophic, and facultative parasitic fungi) were analyzed. A total of 484 genes encoding putative CAZymes, comprising 14 carbohydrate binding module (CBM), 130 carbohydrate esterases (CE), 172 glycoside hydrolases (GH), and 97 glycosyl transferases (GT) (S6 Table). In general, C. bantiana UM 956 has fewer CAZymes than plant pathogenic fungi but comparable to saprotroph and symbiotic fungi (Fig 5A). The CAZyme content in UM 956 is larger than in biotrophic fungi, except for Cladosporium fulvum, renowned for its capability in degrading plant cell wall materials [39]. No polysaccharide lyase (PL) was identified in the C. bantiana genome. Zhao et al. [39] revealed that most of the fungi that lack of PL are saprophytic fungi. Moreover, saprophytic fungi tends to loss CE7, CE8, CE11, GH6, GH11, GH73, GH80 and GH82 families [39], which is also observed in C. bantiana genome. Although experimental supports were deficient, the absence of these CAZyme families may be correlated to the saprotrophic lifestyle of C. bantiana. 10.1371/journal.pone.0161008.g005Fig 5 CAZyme class annotation distribution of C. bantiana UM 956 genome. (A) Comparison of the distribution of CAZyme catalytic domains between C. bantiana and fungi from various lifestyles. (B) Comparison of the plant cell wall degrading potential from CAZyme analysis between C. bantiana and fungi from various lifestyles. AA: auxiliary activities; CBM: carbohydrate binding module; CE: carbohydrate esterase; GH: glycoside hydrolase; GT: glycosyltransferase and PL, polysaccharide lyase. The C. bantiana UM 956 genome contains 75 CAZymes unambiguously related to the breakdown of plant cell wall polysaccharides, such as cellulose, hemicellulose, pectin, and hemicellulose or pectin side chains (S7 Table). As shown in Fig 5B, the number of plant cell wall degrading enzymes in UM 956 was noticeably lower than necrotrophic and hemi-biotrophic fungi but equivalent in size to the saprotrophic fungi. Even if UM 956 displays the quantitative similarity to the saprotrophic fungi, but this neurotropic fungus shows qualitative differences to those of saprobes (Fig 5B and S7 Table). It is conspicuous that the reduction in genes encoding cellulose degrading CAZyme in C. bantiana is extreme (Fig 5B and S7 Table). The cellulose degrading capacity for C. bantiana is significantly smaller than necrotrophic, hemi-biotrophic, facultative parasitic, saprotrophic, and symbiotic fungi, but, not far-off to biotrophic fungi, Ustilago maydis and Blumeria graminis. In contrast, its potential hemicellulose and pectin degrading capacity are strikingly similar to that of Neurospora crassa and Trichoderma reesei. The CAZyme content in the C. bantiana genome suggests a preference of this fungus for hemicellulose and pectin rather than cellulose. Among the unique characteristics of the C. bantiana genome is the presence of an abundant CE1 proteins (31 CE1 proteins), whereas saprotrophic fungi have an average of four genes (minimum 2, maximum 6) (S7 Table). The CE1 family includes feruloyl esterase, acetyl xylan esterase, cinnamoyl esterase, carboxylesterase, S-formylglutathione hydrolase, diacylglycerol O-acyltransferase, and trehalose 6-O-mycolyltransferase. The members of GH61 family are copper-dependent lytic polysaccharide monooxygenases that degrade lignocellulose by catalysing the oxidative cleavage of cellulose [41]. Nonetheless, the GH61 was not determined in C. bantiana, whereas N. crassa and T. reesei have respective three and two genes. In particular, C. bantiana lacks most of the CAZyme families correlated with the plant biomass degradation, including CE5, GH6, GH7, GH36, GH53, GH54, GH62, PL1, and PL3 (S7 Table). Our data showed that UM 956 decompose plant cell wall polysaccharides in a manner strikingly different from the saprotrophic fungi. The number of cellulose degrading enzymes encoded in C. bantiana is smaller than N. crassa and T. reesei, indicative of a lower preference to cope with plant cellulose tissue. All these characteristics relate well to a saprotrophic lifestyle that utilizes plant biomass that has been pre-digested by previous colonizers or employs different approaches to degrade plant polysaccharides. Secreted peptidases One of the features of saprotrophic lifestyle is the lack of predominance of a specific protease family. In fact, they use a wide variety of extracellular proteases to degrade different types of substrate complexes in their environmental niches, indicative their less specialized nutritional status [42–44]. This observation is in agreement with our data that C. bantiana produces broad spectrum of protein degrading enzymes. Peptidase-encoding genes were categorized in the predicted proteomes of the C. bantiana UM 956 genome using the MEROPS database, with emphasis on the extracellular peptidases. We identified 136 peptidases in C. bantiana UM 956, of which, 19 were extracellular peptidases (nine serine peptidases, six aspartic peptidases, two metallopeptidases, one cysteine peptidase, and one threonine peptidase) (S8 Table). Cellular invasive processes involve the degradation of extracellular matrix and basement membranes by different extracellular proteases to enable microbial cells to spread through anatomic barriers to migrate into circulation [45]. Several studies indicated that metalloproteases belong to the M35 and M36 families are expanded in fungal pathogens as adaptation to human or animal hosts [46, 47]. Vu et al. [48] reveled that a secreted fungalysins (M36), known as Mpr1, play a vital role for C. neoformans to breach the endothelium of human blood-brain barrier (BBB) to establish fungal CNS disease [48]. However, we have not observed any deuterolysins (M35) and fungalysins (M36) in the neurotropic C. bantiana UM 956, suggesting it employs other approaches to cause fungal CNS infection. C. bantiana UM 956 was predicted to secrete a repertoire of different endo- and exopeptidases, including the peptidases of MEROPS subfamilies A01, S08, S09, S10, and M28. These extracellular peptidases were secreted by pathogenic fungi to digest proteins in hostile environments of the extracellular matrix with optimal activity at low pH [49]. Pathogenic Candida species comprise a gene family coding for candidapepsin (SAPs) which are virulence factors for localized and disseminated candida infections [50, 51]. The expression of SAPs are associated with hyphal formation, overcome the host immune system and adherence to numerous host tissues and cell types by digestion of host surface proteins [52]. In this study, we identified a putative secreted candidapepsin, SAP3 (UM956_10742) in C. bantiana UM 956 genome. The expression of SAP3 is associated with oral disease and vaginal infection [53]. Exhaustive experiments have been performed to reveal that the SAP3 is one of the principal C. albicans aspartyl proteinases that is involved in mucosal adherence for establishing mucocutaneous infections [54]. Sap3 is expressed during early stages of epithelial colonization and subsequently during mucosal tissue damage [52, 54–56], suggesting its adherence and invasive properties in fungal pathogenicity. The molecular weight of the C. bantiana SAP3 (CbSAP3) is 40.71733 kDa and the isoelectric point is 3.87, computationally predicted using the ExPasy software, which are comparable to experimentally derived results [57]. The protein was mainly composed of random coils and beta sheets based on the predicted secondary structure of the CbSAP3 (S2 Fig). An approximately of 61% of the CbSAP3 structure is random coils, with 234 of its residues making up 23 coils; 29% of its structure is consisted of β-strands, with 110 of its residues making up 30 beta strands; 10% of the CbSAP3 structure is helical, with 38 of its residues making up five helices (S2 Fig). Three-dimensional (3D) homology model of CbSAP3 was generated based on the experimentally solved structural homolog using a fully automated SWISS-MODEL server. The C. albicans SAP3 (PDB: 2h6t.1.A) was used as the template for homology modeling of CbSAP3. The 40% sequence identity between CbSAP3 and C. albicans SAP3 (PDB: 2h6t.1.A) was above the 30% limit that is considered to be the threshold limit for an accurate homology modeling [58]. CbSAP3 appear as a kidney-shaped bilobed protein primarily composed of β-strands that separated into an N-terminal and a C-terminal domain (Fig 6A) which is similar to C. albicans SAP3 protein. The conserved catalytic aspartic acid active sites (Asp82 and Asp272) were located at the groove formed between the domains that is partly covered by an antiparallel β-hairpin made from β-strands 6 and 7 (121–142 residues) (Fig 6A). This β-hairpin structure is known as the active site flap in aspartyl proteinases [59, 60]. The catalytic Asp residues (Asp82 and Asp270) are located in an Asp-Thr-Gly-Ser/Tyr motif in both the N-terminal and C-terminal domains. As previously described [60], water molecule directly binds to both Asp82 and Asp272 catalytic sites surrounded by other water molecules, which interact to Gly84, Gly274, and Thr276 active site residues (Fig 6B and S3 Fig). The structures of catalytic active sites of both CbSAP3 and C. albicans SAP3 were compared (Fig 6B). In consistent with the C. albicans SAP3 structure, the conserved Asp82 residue (OD1 atom) of the CbSAP3 forms a hydrogen bond with N atom of Gly84 residue. The OD1 atom of Asp272 is hydrogen bonded to the Gly274 N atom (Fig 6B). CbSAP3 has a disulfide loop that builds up the N-terminal loop (Cys96 and Cys99) (Fig 6A). In contrast to C. albicans SAP3, no C-terminal loop was observed although a disulfide bond was predicted between Cys306 and Cys343. Apparently the most striking difference in CbSAP3 is the absence of zinc ion and pepstatin A binding sites. Accuracy of the CbSAP3 model was assessed by PROCHECK program, a protein structure validation program. The stereochemistry of main chains and side chains of the CbSAP3 model are showed in S9 Table. All the parameters of main chains and side chains were indicated to be within the acceptable limit and it can serve as a good hypothetical domain. The model could serve as a hypothetical CbSAP3 structure for further investigation of the invasive property of this pathogenic fungus. 10.1371/journal.pone.0161008.g006Fig 6 Homology model of the CbSAP3. (A) Three-dimensional ribbon structures of CbSAP3 and C. albicans SAP3. α-helices are shown in red; β-sheets are shown in yellow and random coils are shown in black. The N- and C- terminal end are labeled. (B) Active sites comparison of the CbSAP3 and C. albicans SAP3. Carbon atoms are shown in green; nitrogen atoms are shown in blue; oxygen atoms are shown in red. Hydrogen bonds are shown in pink dotted lines. Secondary metabolism: melanin Secondary metabolite production is a hallmark of filamentous fungi. These compounds are important chemicals, ranging from fatal mycotoxins to medicinal biologically active compounds that are beneficial to humankind. Secondary metabolite biosynthetic genes are frequently appeared in clusters with a main backbone gene [61]. SMURF tool was used to annotate secondary metabolite biosynthetic gene clusters and backbone genes. A total of 14 gene clusters encoding backbone enzymes were predicted in C. bantiana UM 956 genome, including five genes encoding polyketide synthase (PKS) or PKS-like, eight genes encoding non-ribosomal peptide synthetase (NRPS) or NRPS-like, and a gene encoding dimethylallyl tryptophan synthase (DMATS). Dematiaceous fungi are remarkable microorganisms that readily produce melanins during formation of fungal spore for deposition in the cell wall. The melanins are virulence factors for many pathogenic fungi. Most fungi biosynthesized melanins through 1,8-dihydroxynaphthalene (DHN) pathway. These melanins are known as DHN-melanins. Recognized human pathogens that form DHN-melanins include Aspergillus nidulans, A. niger, Alternaria alternata, C. bantiana, Cladosporium carrionii, E. dermatitidis, E. jeanselmei, Fonsecaea compacta, Phialophora richardsiae, and Neoscytalidium dimidiatum [62]. Polyketide synthases (PKs) are important components to make DHN-melanin precursors. Here, we identified two PKS genes (UM956_2591 and UM956_7690), which is best matched to a characterized conidial yellow pigment biosynthesis PKS (WA) from A. nidulans FGSC A4 [63]. Both the PKSs contain features of a non-reducing fungal type I PKS with a starter unit of ACP transacylase (SAT), β-ketoacyl synthase (KS), acyltransferase (AT), product template (PT), acyl carrier protein (ACP), and thioesterase (TE) [64]. The PKSs have domain order of KS-AT-ACP-ACP-TE similar to the PKS that is involved in melanin biosynthesis [65]. Additionally, a set of DHN melanin-synthesis related proteins, including a scytalone dehydratase, SCD1 (UM956_4290), a tetrahydroxynaphthalene reductase, BRN1 (UM956_1427), and a transcription factor CMR1 (UM956_9060) were identified in C. bantiana UM 956. We also identified two mitogen-activated protein kinases (MAPKs), ChK1 (UM956_1033), and Mps1 (UM956_3770) that have been reported to regulate the expression of CMR1 for pigmentation [66]. Taken together, these data revealed that C. bantiana UM 956 may produce melanin via the DHN-melanin pathway, which regulated by ChK1 and Mps1, although a gene cluster comprising PKS, BRN1, and CMR1 was not determined in the genome. Iron Uptake and Homeostasis Iron is an important biological paradox of survival factor which is indispensable for virtually all organisms including fungi. Importantly, iron acquisition represents a key step in the infection process for pathogens [67]. However, inappropriate storage of iron or iron overload can generate oxidative stress through Haber-Weiss/Fenton chemistry [68]. Thus, fungi usually employ two high-affinity iron uptake mechanisms, nonribosomally synthesized secreted iron chelators (siderophores) and non-siderophore reductive iron assimilation (RIA) [69–71]. Here, we identified two genes encoding L-ornithine-N5-monooxygeneses (UM956_953 and UM956_3728) that catalyze hydroxylation of L-ornithine, the first committed step for the biosynthesis of extracellular fusarinine C (FSC) and triacetylfusarinine C (TAFC), and intracellular ferricrocin (FC). Apart from SidC (UM956_3727), the NRPS crucial for FC synthesis, the FSC biosynthetic enzymes NRPS SidD (UM956_958) and SidF (UM956_956) were also identified in C. bantiana UM956. Nevertheless, the orthologue of TAFC synthetase SidG gene was not identified, indicating that TAFC was not synthesized by C. bantiana UM 956. SidC and SidD lack of typical domain arrangement that observed in A. fumigatus [72]. SidC has an adenylation (A), six thiolation (T), and six condensation (C) domains with the arrangement of T-C-T-C-T-C-A-T-C-T-C-T-C. The gene encoding SidC is located in close proximity to the L-ornithine-N5-monooxygenese (Fig 7A). SidD has a domain arrangement of A-T-C-T-C. The NRPS is located in the FSC biosynthetic gene cluster: siderophore transporter, L-ornithine-N5-monooxygeneses, acetyltransferase SidF, ABC multidrug transporter, and SidD (Fig 7B). Moreover, we identified a GATA-like transcription factor (UM956_2655), a homolog of URBS1 of Ustilago maydis [73] and of SREA of A. nidulans [74], that modulate siderophore biosynthesis and iron uptake. As reported in Yuan et al. [73], multiple GATA binding sites are present in the intergenic region between the genes encoding L-ornithine-N5-monooxygenese and SidC (S10 Table). 10.1371/journal.pone.0161008.g007Fig 7 Siderophore genes of C. bantiana UM 956 genome. (A) FC synthetase SidC gene. (Top panel) Schematic map of SidC (black arrow) and adjacent genes. Gray arrows are genes which are expected to function during siderophore biosynthesis. Numbers are in kilobases. (Bottom panel) Domain setup of SidC. (B) FSC synthetase SidD gene. (Top panel) Schematic map of SidD (black arrow) and adjacent genes. Numbers are in kilobases. (Bottom panel) Domain setup of SidD. Alternatively, RIA is an important backup mechanism for fungi to support extracellular siderophore-driven iron homeostasis [71]. RIA is a three-step process which is regulated by metalloreductase Fre1, ferroxidase Fet3, and iron permease Ftr1. In this study, a metalloreductase Fre8 (UM956_11306), three ferroxidases Fet3 (UM956_46, UM956_979, UM956_3628), and two iron permease Ftr1 (UM956_47 and UM956_980) were identified in the genome. It should be noted that two ferroxidase and iron permease encoding gene clusters were identified. As previously reported [75, 76], the ferroxidases Fet3, UM956_46 and UM956_979 were located adjacent to iron permeases Ftr1, UM956_47 and UM956_980, respectively. Stress adaptation in C. bantiana Pathogenic fungi may contain a battery of stress-responsive proteins to promote the pathogenicity and tune the physiological fitness to their diverse host niches. It is clear that C. bantiana can adapt well in animal as well as human host microenvironment. Thus, the C. bantiana UM 956 genome was mapped to the fungal stress response database (FSRD) to obtain information of stress biology of C. bantiana. A total of 311 putative proteins was assigned to their orthologous genes in the FSRD. The full list of FSRD annotation was listed in S11 Table. C. bantiana can grow at human body temperature and up to 43°C [77]. Therefore, it is not surprising that C. bantiana contains an armory of thermal stress-responsive proteins to enable the fungus to survive in human host. C. bantiana UM 956 genome was predicted to contain HSP60 (UM956_2121), HSP70 (UM956_8246), HSP78 (UM956_6295), HSP83 (UM956_4074), HSP98 (UM956_1152), and heat shock protein ssb1 (UM956_3475) to promote degradation of damaged proteins, folding of client proteins, and to stabilize proteins and membranes [78]. Indeed, heat shock transcription factor Hsf1 is activated in response to the thermal stress and bind to the canonical heat shock elements (HSEs) in the promoters of heat shock proteins [79]. Here, an Hsf1 (UM956_7666) was identified in C. bantiana UM 956. Thermotolerance is an important virulence factor that is responsible for its pathogenicity [40]. Nicholls et al. [79] indicated that the activation of Hsf1 is required despite during slow thermal changes such as those suffered by febrile patients. The Hsf1-HSE regulons in C. bantiana may function to promote cellular adaptation in response to the thermal stress. In human, phagocytic cells tend to produce an array of reactive oxygen species (ROS) such as superoxide (O2−) to kill fungal pathogens by damaging their deoxyribonucleic acid, proteins and membranes [80]. Thus, fungal pathogens harbor some detoxification mechanisms to respond to these chemicals. These ROS stress responses have been reported to promote the fungal pathogens’ virulence [81, 82]. In this work, we found that C. bantiana UM 956 may detoxify ROS using superoxide dismutase (SOD), catalase, and components of the thioredoxin and glutaredoxin systems to thrive against the oxidative environments created by human host. The fungus may produce two SODs: SOD2 (Mn/Fe) (UM956_1425) and SODC (Cu/Zn) (UM956_6712) that catalyze dismutation of the O2− species into molecular oxygen (O2) or hydrogen peroxide (H2O2). C. bantiana UM 956 also contains other important genes encoding antioxidant enzymes, including seven catalases (UM956_6359, UM956_6429, UM956_7993, UM956_8082, UM956_10667, UM956_11123, and UM956_4627), thioredoxin reductase (UM956_11048), monothiol glutaredoxin, grx5 (UM956_6203), grx3 (UM956_9386), glutathione peroxidase, gpx1 (UM956_328), and methionine sulfoxide reductase (UM956_12128) were identified in the genome. Furthermore, we identified a tpsA gene encoding trehalose-6-phosphate synthase (UM956_10499) that involved in the first step biosynthesis of trehalose (antioxidant). Based on NCBI nr and SwissProt annotations, a series of signaling proteins and transcription factors that required for oxidative stress resistance in the fungus were determined (S12 Table). There is undoubtedly the neurotropic C. bantiana UM 956 comprises of a number of effective defense mechanisms to respond combinations of different host-generated stresses that relevant to the host-pathogen interaction. Antifungal susceptibility profile and drug resistance genes According to the Epsilometer Test, the isolate is susceptible to caspofungin (MIC = 0.032 μg/mL), anidulafungin (MIC = 0.003 μg/mL), posaconazole (MIC = 0.004 μg/mL), voriconazole (MIC = 0.19 μg/mL) and itraconazole (MIC = 0.125 μg/mL). However, the isolate showed low resistance to amphotericin B (MIC = 2 μg/mL) and high resistance to fluconazole (MIC >256 μg/mL). Compelling evidence has been gathered for the genetic basis of azole resistance development in yeast and molds. Azole resistance can be caused by mutations in the drug target enzyme, lanosterol 14α demethylase, which is encoded by the ERG11/CYP51 gene or by overexpression of genes coding for membrane transport proteins [83–85]. In this study, two lanosterol 14α demethylases (UM956_1899, 97.4% and UM956_10619, 96.7%) were identified in the C. bantiana UM 956 genome. These two protein sequences were compared with the previously published ERG11 protein sequences from azole-susceptible C. albicans isolates (GenBank accession no. XM_711668 and AIX03623) to investigate the Erg11 mutations of fluconazole resistant isolate. Common ERG11 mutations (A114S, Y132F, K143R, F145L, S405F, D446E, G448E, G450E, and G464S) that related to azole resistance [86, 87] were not observed. However, nonsynonymous Erg11 mutations (D153E, E266D, and D116E) that reported in fluconazole resistant C. albicans isolates [88] were identified in UM956_1899 and UM956_10619. Single mutation (D116E) was detected in UM956_10619, while more than one nucleotide changes (D153E, E266D, and D116E) were identified in UM956_1899 (S4 Fig). Moreover, we determined two putative NDT80 transcription factors (UM956_5350 and UM956_11916) in the genome, which are important in regulating sterol metabolism and fluconazole tolerance in C. albicans by binding to the promoters of Erg11 gene [89]. The 500 bp upstream regions of the UM956_1899 and UM956_10619 genes were analysed in silico for NDT80 transcription factor binding site. Among them, only UM956_1899 gene contains a putative Ndt80 binding consensus located between −182 and −202 (S5 Fig). Azoles resistance can be also mediated by overexpression of major facilitator superfamily (MFS) transporter encoded by MDR1 (multidrug resistance) and ATP-binding cassette (ABC) transporters encoded by CDR1 and CDR2 (Candida drug resistance). MFS transporters contain two type of proton antiporters, including drug: H+ antiport 1 (DHA1), consisting of 12 transmembrane domains and drug: H+ antiport 2 (DHA2), consisting of 14 transmembrane domains [90, 91]. These efflux pumps play important role to efflux drugs in exchange with one or more H+ with a substrate molecule to confer azoles resistance in prokaryotes as well as eukaryotes [92]. In this study, a DHA1 (UM956_ 913) was identified in the C. bantiana UM 956 genome, although its expression level remain unknown. TMpred analysis predicted 12 transmembrane helices in UM956_913 (S6 Fig), suggesting it harbours transmembrane transporter activity to efflux drugs. In addition, neither CDR1 nor CDR2 was found in the C. bantiana UM 956 genome. Taken together, the ERG11, NDT80 transcription factor, and the DHA1 may work in concert to regulate ergosterol metabolism to confer fluconazole resistance in C. bantiana UM 956. Comparative genomics of C. bantiana and other dematiaceous fungal pathogens Phylogenomic tree was constructed using nine publicly available dematiaceous fungal and a yeast pathogens from different classes, including two Eurotiomycetes, three Dothideomycetes, three Sodariomycetes and a Tremellomycetes as outgroup. A total of 98,789 proteins were clustered into 15,953 orthologous clusters with 2,199 single-copy orthologous genes determined in all the analyzed genomes. Concatenated alignments of 2,199 (13.78%) single-copy orthologous genes were used to construct Bayesian trees. From the phylogenomic trees, Sodariomycetes fungi (D. eschscholzii UM 1020, D. eschscholzii UM 1400, and S. schenckii strain 1099–18) were clustered into one clade; Dothideomycetes fungi (O. mirabilis UM 578, Pyrenochaeta sp. UM 256, and B. papendorfii UM 226) were grouped into a different clade, while Eurotiomycetes fungi (Exophiala dermatitidis NIH/UT8656 and C. bantiana UM 956) form a separate branch (Fig 8). 10.1371/journal.pone.0161008.g008Fig 8 Comparative phylogenomic analysis of C. bantiana UM 956 along with eight previously published dematiaceous fungal genomes using Bayesian. Number at the node referring to Bayesian posterior probability. The tree is rooted with C. neoformans var. grubii H99 as outgroup. The core gene families between C. bantiana UM 956 and other nine genomes were identified using the OrthoMCL analysis (S13 Table). The other nine genomes used for comparison were divided into dematiaceous fungal pathogens (B. papendorfii, D. eschscholtzii, O. mirabilis, E. dermatitidis, Pyrenochaeta sp., and S. schenckii) and a neurotropic yeast pathogen (C. neoformans var. grubii) groups. Based on the orthology analysis, the gene families in C. bantiana UM 956 can be classified into three categories: i.) found in C. bantiana UM 956 and other dematiaceous fungi, 1477 gene families; ii.) found in C. bantiana UM 956 and E. dermatitidis NIH/UT8656, 921 gene families; iii.) only found in C. bantiana UM 956, 148 gene families (S14 Table). The C. bantiana UM 956 and E. dermatitidis NIH/UT8656 shared 21 family clusters with known functions (Table 5). Of these, only gene encoding calmodulin (UM956_2895) was related to fungal pathogenicity based on the pathogen-host interaction (PHI) gene analysis. Calmodulin plays a vital role in the Ca2+ signaling pathways. Disruption of Ca2+/calmodulin-dependent kinase in Magnaporthe oryzae not only hindered the formation of appressoria and conidial germination but also reduced its ability to infect rice plants [93]. Additionally, C. bantiana UM 956 and E. dermatitidis NIH/UT8656 shared gene families associated with stress responses, including, genes encoding myosin-cross-reactive antigen (mcra) (family Clado12877) and glucose-repressible protein, Grg-1 (family Clado13011). Myosin-cross-reactive antigen provides additional function in stress protection in bacteria [94]. Bischoff et al. [95] postulated that high temperature at the site of infection triggered stress in Staphylococcus aureus to up-regulate the expression of mcra gene. Fungal pathogens and saprophytes contain various glucose-repressive activities to obtain nutrients from their environments [96]. Previous studies revealed that the grg-1 gene is regulated by glucose and its mRNA level was markedly increased when glucose was deprived [96]. 10.1371/journal.pone.0161008.t005Table 5 Shared gene families in both C. bantiana UM 956 and E. dermatitidis NIH/UT8656. Families Gene ID Annotation Clado12617 UM956_2895 Calmodulin Clado12419 UM956_11691 Enolase Clado13343 UM956_10094 3-oxoacyl-[acyl-carrier protein] reductase Clado12502 UM956_1607 Hydroxymethylglutaryl-CoA lyase Clado8358 UM956_11042 Gluconate 5-dehydrogenase Clado12892 UM956_5601 Alpha-ketoglutarate-dependent 2,4-dichlorophenoxyacetate dioxygenase Clado12552 UM956_2102 Glucarate dehydratase Clado12877 UM956_5482 Myosin-cross-reactive antigen Clado13319 UM956_1006 Serine/threonine protein kinase Clado12819 UM956_4926 Transcription initiation factor TFIID subunit 12 Clado13293 UM956_9670 Alpha-methylacyl-CoA racemase Clado12881 UM956_5516 myb-like DNA-binding protein Flb Clado12590 UM956_2498 Serine/threonine protein kinase Clado12337 UM956_10723 Regulatory protein SWI5 Clado13216 UM956_102 Alcohol dehydrogenase Clado12541 UM956_210 Lysyl oxidase-like protein 2/3/4 Clado8464 UM956_9770 Biphenyl-2,3-diol 1,2-dioxygenase Clado13011 UM956_6681 Glucose-repressible protein, Grg-1 Clado12706 UM956_3739 Carboxypeptidase Clado13191 UM956_876 Glucan 1,3-beta-glucosidase Clado13300 UM956_9747 Acetolactate synthase, catabolic Whole proteome alignment of the C. bantiana UM 956 and the other nine fungal gene contents showed that C. bantiana UM 956 harbored 679 unique genes. Among the 148 unique gene families, only four clusters with known functions (Table 6). The predominant of the gene families were hypothetical proteins or unannotated proteins. Thus, we performed domain analysis on these unknown proteins to obtain clue of their role or function in C. bantiana UM 956. The analysis revealed that most of the hypothetical proteins related to MFS (eight clusters) and cytochrome P450 (seven clusters). Major facilitator superfamily is one of the largest families of transporters that ubiquitously found in bacteria, archaea, and eukaryotes [97]. Apart from export host-derived antimicrobial compounds as described in the previous section, MFS also involved in transport secondary metabolites, a wide array of organic and inorganic anions and cations, and essential nutrients [97]. The cytochrome P450 proteins play many roles in physiological processes, such as secondary metabolites biosynthesis, detoxification of host defense compounds, and xenobiotics degradation. In this work, we found that 14 of the unique hypothetical proteins contain cytochrome P450 domain or conserved sites. Nonetheless, the exact roles of these proteins remain to be determined. In particular, we identified two hypothetical proteins (UM956_2358 and UM956_9078) with Clr5 domain (Clado10284 family). Clr5 play a role in gene silencing in fission yeast heterochromatin by modulation of chromatin structure in the mating-type region and regulate sexual differentiation [98]. Overall, the putative unique Clr5-containing proteins might involve in the epigenetic regulation to affect gene silencing at heterochromatic loci. 10.1371/journal.pone.0161008.t006Table 6 Specific functional family clusters in C. bantiana UM 956. Families Gene ID Annotation Clado8347 UM956_2269 Acetoacetate-CoA ligase UM956_7538 Acetoacetate-CoA ligase UM956_10163 Acetoacetate-CoA ligase Clado12590 UM956_2498 Serine/threonine protein kinase Clado13242 UM956_9168 Serine/threonine protein kinase Clado10301 UM956_4621 Succinate-semialdehyde dehydrogenase (NADP+) UM956_9931 Succinate-semialdehyde dehydrogenase (NADP+) Conclusions The report represents the first case of brain abscess caused by C. bantiana in Malaysia and Southeast Asia. C. bantiana UM 956 was identified using combined morphological examination and multigene phylogeny. The infected patient was successfully treated with a combination of surgical excision and antifungal therapy. Excision rather than aspiration combined with systemic long term antifungal therapy is essential for resolution of fungal cerebral phaeohyphomycosis infection. In this study, we successfully produced a high-quality draft genome sequence of C. bantiana UM 956. The C. bantiana UM 956 constitutes a repertoire of genes for the saprophytic lifestyle as well as invasion action of the pathogen. Moreover, our genome analysis revealed that the neurotrophic fungus can respond to multiple environmental stresses to enable it to survive in human host. Materials and Methods Ethic statement Approval for this study was obtained from the Medical Research and Ethic Committee (MREC), Ministry of Health (MOH) (reference number: (7) KKM/NIHSEC/P151131), and written consent was obtained. Fungal isolate C. bantiana UM 956 was recovered from the brain tissue of a patient with brain abscess in the Mycology Unit, Department of Medical Microbiology, UMMC, Kuala Lumpur. The isolate was processed according to the laboratory’s standard operating procedures (SOP) with direct wet mount microscopy followed by inoculation on SDA for incubation at 30°C up to seven days, with alternate day examination for fungal growth. Macroscopic examination was performed to observe its colonial morphology, such as color, texture, and topography. Tease mount and slide culture were performed to identify the arrangement of conidia and conidiophores. In vitro antifungal susceptibility The in vitro antifungal susceptibility of the fungal isolate was evaluated by the Epsilometer Test (Etest, Biomerieux, France) by determining the minimum inhibitory concentrations (MICs) according to the previous study [99]. The isolate was tested against nine antifungal drugs, including fluconazole, itraconazole, posaconazole, voriconazole, anidulafungin, caspofungin, micafungin, 5-flucytosine, and amphotericin B. DNA sequencing and phylogenetic analysis The internal transcribed spacer region (ITS), the small subunit of the ribosomal RNA gene (SSU), and the large subunit of the ribosomal RNA gene (LSU) were used as targets for molecular identification of the clinical isolate. Total DNA extraction, PCR amplification, sequencing, and BLASTn search were performed as described previously [26, 99]. Unique ITS and LSU sequences from the isolate, together with an additional 12 species of Cladophialophora and an outgroup strain of P. schaereri (Table 1) were subjected to phylogenetic analysis. Multiple sequence alignments of collected ITS and D1/D2 nucleotide sequences were generated using M-Coffee [100]. Individual alignments were concatenated for Bayesian Markov Chain Monte Carlo (MCMC) analysis partitioned by the gene. Bayesian tree analyses were performed using MrBayes v3.2.2 with reversible jump MCMC averaging over the entire general time reversible (GTR) rates and gamma-distributed rate heterogeneity for all subsets of the partitioned scheme. A total of 500,000 generations were run with a sampling frequency of 100, and diagnostics were calculated for every 1,000 generations. The first 1,250 trees were discarded with a burn-in setting of 25%. Convergence was assessed with a standard deviation of split frequencies below 0.01, no noticeable trend for the plot of the generation versus the log probability of the data, and a potential scale reduction factor (PSRF) close to 1.0 for all parameters. Genomic DNA extraction, genome sequencing and assembly Genomic DNA of C. bantiana UM 956 was extracted as described previously [26, 34]. The genome was sequenced using Illumina HiSeq 2000 Sequencer (Illumina Inc., San Diego, CA USA) in a 2×90 bp paired-end mode on 500-bp and 5-kb library sizes. Illumina library was prepared using TruSeq v3 Reagent Kits (Illumina Inc., San Diego, CA USA). The selected library fragments were purified through gel electrophoresis, which then selectively enriched and amplified by PCR. The 500-bp Illumina sequenced read was then combined with the 5-kb Illumina sequenced read for further processing. Both sets of sequenced reads were first pre-processed using FASTX-Toolkit (http://hannonlab.cshl.edu/fastx_toolkit/) trimming bases with a Phred quality below Qv20 from the 3’-end of the reads. The trimmed reads shorter than 30 bp and reads with 40% bases having Qv ≤ 20 were filtered out, retaining small-insert reads ≥ 80 bp and large-insert reads ≥ 30 bp. Two bases were trimmed from the 5’-terminal of all reads. Pre-processed reads from both libraries were assembled with Velvet version 1.2.07 [101] with k-mer setting = 73, insert length = 568, -ins_length_sd = 100, -min_pair_count = 15, insert_length2 = 5000, ins_length2_sd = 500, and min_contig_lgth = 200. Additional parameter of -shortMatePaired = yes was set for large insert library. The generated contig assembled from the Velvet were further scaffolded using SSPACE Basic v2.0 [102] with more stringent parameters than software default to achieve higher accuracy assembly (parameters: -z 200, -k 15, -a 0.3, -n 30 and -T 10). GapFiller v1.10 (-m = 60, -o = 15, -r = 0.8,–n = 30, -t = 30 and -T = 10) was used to perform gap filling by utilizing paired-end sequencing data from both libraries [103, 104]. Genome annotation of C. bantiana UM 956 Interspersed repetitive elements and low complexity DNA sequences were masked using RepeatMasker version open-3.3.0 with the Repbase fungal library version rm-20120418, followed by masking off the RNA sequences. GeneMark-ES version 2.3e [105] was used for gene prediction in C. bantiana UM 956 complete genome. Annotation of coding sequences for UM 956 was completed using BLAST (Basic Local Alignment Search Tool) searches against the NCBI nr protein and SwissProt databases. Protein domain families were matched to Pfam database using InterProScan 5 [106]. Individual rRNA and tRNA were predicted using RNAmmer v1.2 [107] and tRNAscan-SE v1.3.1 [108], respectively. Putative transposable elements were identified using Transposon-PSI (http://transposonpsi.sourceforge.net) by PSI-TBLASTN searches with a collection of (retro-) transposon open reading frame (ORF) homology profiles. Functional annotation of predicted genes Kyoto Encyclopedia of Genes and Genomes (KEGG) metabolic pathways matches was carried out using local BLAST2GO tools [109]. Protein classification was performed using EuKaryotic Orthologous Group (KOG) [110]. The total number of predicted proteins involved in the KOG category “Secondary metabolites biosynthesis, transport and catabolism” for other pathogenic dematiaceous fungi was obtained from the DemaDB database (fungaldb.um.edu.my) [111]. The best protein models of the C. bantiana UM 956 genome was subjected to database of automated Carbohydrate-active enzyme ANnotation (dbCAN) annotation pipeline [112]. Comparative analysis was performed against fungi with different lifestyles, including saprophytic fungi (Neurospora crassa OR74A and Trichoderma reesei QM6a), facultative parasitic fungi (A. nidulans FGSC A4), biotrophic fungi (C. flavum, Ustilago maydis, and Blumeria graminis), necrotrophic fungi (Cochliobolus heterostrophus C4 and C. heterostrophus C5), hemi-biotrophic fungi (Cochliobolus sativus ND90Pr, Magnaporthe oryzae 70–15 version 8, and Fusarium graminearum PH-1), and symbiotic fungus (Laccaria bicolor) [39, 113]. Peptidases was identified by a batch blast of C. bantiana UM 956 protein models against MEROPS database [114]. Secreted peptidases were determined by the cleavage sites prediction and signal peptide/non-signal peptide using SignalP version 4.12 [115] after discarding proteins with transmembrane domains that were determined by TMHMM version 2.0 [116]. Secreted peptidases were selected based on the proteins without transmembrane domain or the presence of a transmembrane domain located at in the N-terminal 40 amino acids as it is responsible to the secretion signal. Genome mapping of secondary metabolite backbone genes and associated genes for secondary metabolite biosynthesis cluster were carried out using web-based SMURF (Secondary Metabolite Unknown Regions Finder) (www.jcvi.org/smurf/) [117]. The organisation SidC and SidD gene clusters were retrieved from the sequenced genome using Artemis v12.0 sequence viewer [118]. Stress responsive genes and pathogenicity-associated genes were predicted by BLASTP search against fungal stress response database (FSRD) and pathogen-host interaction database (PHI-base). Amino acid sequences with e-value threshold ≤1e-5, alignment length over 70% of its own length and over 50% match identity were assigned as the annotation of putative genes. Orthologous genes and comparative genomic analysis The protein sequences of all current publicly available dematiaceous fungal genomes (B. papendorfii UM 226 [26], D. eschscholtzii UM 1020 and UM 1400 [27], O. mirabilis UM 578 [28], Pyrenochaeta sp. UM 256 [29], S. schenckii strain 1099–18 [30], and E. dermatitidis NIH/UT8656 [31]) and a neurotrophic yeast genome (C. neoformans var. grubii H99 [17]) were used to determine the orthologous genes in C. bantiana UM 956. The genome sequences of B. papendorfii UM 226, D. eschscholtzii UM 1020 and UM 1400, O. mirabilis UM 578, and Pyrenochaeta sp. UM 256 were downloaded from DemaDb database (fungaldb.um.edu.my) [111]; the S. schenckii strain 1099–18 genome sequence was acquired from Laboratorio Nacional de Computacao Cientifica; the C. neoformans var. grubii H99 genome sequence was obtained from Broad Institute. The OrthoMCL version 2.02 [119] was used in the analysis of protein sequences clustering (≥33 amino acids) for C. bantiana UM 956 and the nine genome references by all-against-all BLASTp searches of all proteins. The reciprocal best hits from distinct genomes were identified as orthologous genes. Phylogenomic analysis A phylogenomic tree was made using all proteome clusters of the seven dematiaceous fungal species produced from comparative analysis. The tree is rooted with C. neoformans var. grubii H99 as outgroup. A total of 2,272 single-copy orthologous genes containing one member in each species was subjected to individual sequence alignments using ClustalW version 2.0 [120]. TrimAL (with–gt 0.5) was used to discard all spurious sequences or poorly aligned regions. The filtered multiple alignments were then concatenated into a superalignment with 1,239,944 characters. Bayesian phylogenetic analysis was run using MrBayes v3.2.2 [121] with mixed amino acid model, gamma-distributed rate variation across sites, and a proportion of invariable sites. The MCMC was run using a sampling frequency of 100 for 100,000 generations with a burn-in setting of 25% Prediction of secondary structure and homology modelling Secondary structure of the CbSAP3 was predicted using PSIPRED v3.3 [122]. Amino acid sequence alignment between CbSAP3 and C. albicans SAP3 was performed using Clustal Omega. Three-dimensional structural models of the CbSAP3 was produced using SWISS-MODEL server [123] using the template (C. albicans SAP3, Protein Data Bank entry 2h6t.1.A) [60] selected from template identification function in the SWISS-MODEL program. The generated model was stored as a PDB output file and the predicted structure and C. albicans SAP3 protein were visualized and compared using Swiss-PdbViewer 4.0.1. The quality of the protein model generated was validated using the PROCHECK program as described in Kuan et al. [124]. Identification of GATA-like and NDT80 transcription factors binding elements The intergenic region between the genes encoding L-ornithine-N5-monooxygenese and SidC and the 500 bp of upstream regions of Erg11 genes were retrieved from the C. bantiana UM 956 genome using Artemis v12.0 sequence viewer. Consensus GATA-like and NDT80 transcription factor binding elements were predicted using the JASPAR database (http://jaspar.genereg.net/cgi-bin/jaspar_db.pl). Nucleotide sequence accession numbers The ITS, SSU, and LSU nucleotide sequences of C. bantiana UM 956 were deposited in GenBank with accession numbers KU928131, KU928132, and KU928133, respectively. The assembly of the C. bantiana UM 956 was deposited in European Nucleotide Archive with BioProject number PRJEB13102 (sequence accession: FLVJ01000001-FLVJ01000426). Supporting Information S1 Fig Cranial CT scan showing a large, irregular rim-enhancing necrotic lesion in the left parietal region close to the motor strip causing significant mass effect with surrounding edema. (PDF) Click here for additional data file. S2 Fig Secondary structure prediction of the CbSAP3. (PDF) Click here for additional data file. S3 Fig Amino acid sequence alignment between CbSAP3 and C. albicans SAP3. Sequence alignment was performed using Clustal Omega. Asterisk (*) indicates positions of conserved catalytic active sites. (PDF) Click here for additional data file. S4 Fig Amino acid sequence alignment of C. bantiana Erg11 (UM956_1899 and UM956_10619) and C. albicans Erg11 (XM_711668 and AIX03623). Sequence alignment was performed using Clustal Omega. Asterisk (*) indicates positions of point mutations that related to azole resistance. (PDF) Click here for additional data file. S5 Fig Sequence analysis of Erg11’-flanking region. Underlined sequences are the Ndt80 transcription factor binding sites predicted using the JASPAR database. Translation start codon ATG is indicated with boldface and boxed. (PDF) Click here for additional data file. S6 Fig Transmembrane domains of DHA1 (UM956_ 913) as predicted by TMpred. The horizontal line represents the level of hydrophobicity (score ≥ 500) that predicts membrane-spanning domains with high probability. Predicted transmembrane domains are indicated with Roman numerals. The X axis is the amino acid sequences of the enzyme and the Y axis is the hydrophobicity of the residues. (PDF) Click here for additional data file. S1 Table Basic genome statistics of C. bantiana UM 956 and other previously sequenced fungal genomes. (XLSX) Click here for additional data file. S2 Table List of protein-coding genes identified in C. bantiana genome based on the NCBI nr, SwissProt, and InterPro databases. (XLSX) Click here for additional data file. S3 Table List of hypothetical proteins predicted in C. bantiana UM 956 genome. (XLSX) Click here for additional data file. S4 Table KOG classification of predicted proteins in C. bantiana UM 956 genome. (XLT) Click here for additional data file. S5 Table Distribution of predicted proteins from C. bantiana UM 956 genome that involved in in KEGG metabolic pathway. (XLSX) Click here for additional data file. S6 Table CAZyme class annotation distribution in C. bantiana UM 956 genome. (XLS) Click here for additional data file. S7 Table Plant cell wall degrading and modifying CAZyme families predicted in C. bantiana UM 956 genome. (XLS) Click here for additional data file. S8 Table List of genes encoding peptidases predicted using MEROPS analysis. (XLS) Click here for additional data file. S9 Table Stereochemistry of the CbSAP3 protein model validated using PROCHECK. (XLS) Click here for additional data file. S10 Table Predicted GATA binding sites in the intergenic region between the genes encoding L-ornithine-N5-monooxygenese and SidC. (XLSX) Click here for additional data file. S11 Table List of predicted genes encoding stress response proteins identified based on FSRD. (XLS) Click here for additional data file. S12 Table List of signaling proteins and transcription factors that involved in adaptive responses to osmotic stress. (XLS) Click here for additional data file. S13 Table List of gene families and the distribution of the genes in each families between C. bantiana UM 956 and other tested fungi/yeast. BP226: B. papendorfii UM 226; Cl956: C. bantiana UM 956; CN99: C. neoformans var. grubii H99; DE1020: D. eschscholtzii UM 1020; DE1400: D. eschscholtzii UM 1400; OM578: O. mirabilis UM 578; Py256: Pyrenochaeta sp. UM 256 and SS1099: S. schenckii strain 1099–18; ED: E. dermatitidis NIH/UT8656. (XLSM) Click here for additional data file. S14 Table Gene families shared between C. bantiana UM 956 and other tested fungi/yeast. BP226: B. papendorfii UM 226; Cl956: C. bantiana UM 956; CN99: C. neoformans var. grubii H99; DE1020: D. eschscholtzii UM 1020; DE1400: D. eschscholtzii UM 1400; OM578: O. mirabilis UM 578; Py256: Pyrenochaeta sp. UM 256 and SS1099: S. schenckii strain 1099–18; ED: E. dermatitidis NIH/UT8656. (XLSX) Click here for additional data file. ==== Refs References 1 Geis PA , Wheeler MH , Szaniszlo PJ . Pentaketide metabolites of melanin synthesis in the dematiaceous fungus Wangiella dermatitidis . Arch Microbiol . 1984 ;137 (4 ):324 –8 . 6539583 2 Revankar SG , Sutton DA . Melanized fungi in human disease . Clin Microbiol Rev . 2010 ;23 (4 ):884 –928 . 10.1128/CMR.00019-10 20930077 3 Rajshekhar V . Surgical management of intracranial fungal masses . Neurol India . 2007 ;55 (3 ):267 –73 . 17921656 4 Ramesha KN , Kate MP , Kesavadas C , Radhakrishnan VV , Nair S , Thomas SV . Fungal infections of the central nervous system in HIV-negative patients: experience from a tertiary referral center of South India . Ann Indian Acad Neurol . 2010 ;13 (2 ):112 –6 . 10.4103/0972-2327.64635 20814494 5 Baddley JW , Salzman D , Pappas PG . Fungal brain abscess in transplant recipients: epidemiologic, microbiologic, and clinical features . Clin Transplant . 2002 ;16 (6 ):419 –24 . 12437621 6 Revankar SG , Sutton DA , Rinaldi MG . Primary central nervous system phaeohyphomycosis: a review of 101 cases . Clin Infect Dis . 2004 ;38 (2 ):206 –16 . 14699452 7 Davis LE . Fungal infections of the central nervous system . Neurol Clin . 1999 ;17 (4 ):761 –81 . 10517927 8 Lui G , Lee N , Ip M , Choi KW , Tso YK , Lam E , et al Cryptococcosis in apparently immunocompetent patients . QJM . 2006 ;99 (3 ):143 –51 . 16504989 9 Chakrabarti A . Epidemiology of central nervous system mycoses . Neurol India . 2007 ;55 (3 ):191 –7 . 17921647 10 Dixon DM , Shadomy HJ , Shadomy S . Dematiaceous fungal pathogens which were isolated from nature . Mycopathologia . 1980 ;70 (3 ):153 –61 . 7189579 11 Revankar SG . Cladophialophora bantiana brain abscess in an immunocompetent patient . Can J Infect Dis Med Microbiol . 2011 ;22 (4 ):149 –50 . 23205028 12 Kantarcioglu AS , de Hoog GS . Infections of the central nervous system by melanized fungi: a review of cases presented between 1999 and 2004 . Mycoses . 2004 ;47 (1–2 ):4 –13 . 14998393 13 Garzoni C , Markham L , Bijlenga P , Garbino J . Cladophialophora bantiana: a rare cause of fungal brain abscess. Clinical aspects and new therapeutic options . Med Mycol . 2008 ;46 (5 ):481 –6 . 10.1080/13693780801914906 18608882 14 Levin TP , Baty DE , Fekete T , Truant AL , Suh B . Cladophialophora bantiana brain abscess in a solid-organ transplant recipient: case report and review of the literature . J Clin Microbiol . 2004 ;42 (9 ):4374 –8 . 15365048 15 Werlinger KD , Yen Moore A . Eumycotic mycetoma caused by Cladophialophora bantiana in a patient with systemic lupus erythematosus . J Am Acad Dermatol . 2005 ;52 (5 Suppl 1):S114 –7 . 15858506 16 D'Souza CA , Kronstad JW , Taylor G , Warren R , Yuen M , Hu G , et al Genome variation in Cryptococcus gattii, an emerging pathogen of immunocompetent hosts . MBio . 2011 ;2 (1 ):e00342 –10 . 10.1128/mBio.00342-10 21304167 17 Janbon G , Ormerod KL , Paulet D , Byrnes EJ 3rd, Yadav V , Chatterjee G , et al Analysis of the genome and transcriptome of Cryptococcus neoformans var. grubii reveals complex RNA expression and microevolution leading to virulence attenuation . PLoS Genet . 2014 ;10 (4 ):e1004261 10.1371/journal.pgen.1004261 24743168 18 Harrison DK , Moser S , Palmer CA . Central nervous system infections in transplant recipients by Cladophialophora bantiana . South Med J . 2008 ;101 (3 ):292 –6 . 10.1097/SMJ.0b013e318164e289 18364660 19 Fica A , Diaz MC , Luppi M , Olivares R , Saez L , Baboor M , et al Unsuccessful treatment with voriconazole of a brain abscess due to Cladophialophora bantiana . Scand J Infect Dis . 2003 ;35 (11–12 ):892 –3 . 14723373 20 Silveira ER , Resende MA , Mariano VS , Coura WA , Alkmim LD , Vianna LB , et al Brain abscess caused by Cladophialophora (Xylohypha) bantiana in a renal transplant patient . Transpl Infect Dis . 2003 ;5 (2 ):104 –7 . 12974792 21 Huang WM , Fan YM , Li W , Yang WW . Brain abscess caused by Cladophialophora bantiana in China . J Med Microbiol . 2011 ;60 (Pt 12 ):1872 –4 . 10.1099/jmm.0.032532-0 21852529 22 Dixon DM , Shadomy HJ , Shadomy S . Dematiaceous fungal pathogens isolated from nature . Mycopathologia . 1980 ;70 (3 ):153 –61 . 7189579 23 Suri P , Chhina DK , Kaushal V , Kaushal RK , Singh J . Cerebral Phaeohyphomycosis due to Cladophialophora bantiana—A Case Report and Review of Literature from India . J Clin Diagn Res . 2014 ;8 (4 ):DD01 –5 . 10.7860/JCDR/2014/7444.4216 24959445 24 de Hoog GS , Guarro J , Gené J , Figueras MJ , editors. Atlas of Clinical Fungi . 2nd ed. Centraalbureau voor Schimmelcultures, Baarn , the Netherlands 2005 . 25 Cresnar B , Petric S . Cytochrome P450 enzymes in the fungal kingdom . Biochim Biophys Acta . 2011 ;1814 (1 ):29 –35 . 10.1016/j.bbapap.2010.06.020 20619366 26 Kuan CS , Yew SM , Toh YF , Chan CL , Ngeow YF , Lee KW , et al Dissecting the fungal biology of Bipolaris papendorfii: from phylogenetic to comparative genomic analysis . DNA Res . 2015 ;22 (3 ):219 –32 . 10.1093/dnares/dsv007 25922537 27 Chan CL , Yew SM , Ngeow YF , Na SL , Lee KW , Hoh CC , et al Genome analysis of Daldinia eschscholtzii strains UM 1400 and UM 1020, wood-decaying fungi isolated from human hosts . BMC Genomics . 2015 ;16 (1 ):966 .26581579 28 Chan CL , Yew SM , Na SL , Tan YC , Lee KW , Yee WY , et al Draft Genome Sequence of Ochroconis constricta UM 578, isolated from human skin scraping . Genome Announc . 2014 ;2 (2 ). 29 Yew SM , Chan CL , Soo-Hoo TS , Na SL , Ong SS , Hassan H , et al Draft genome sequence of dematiaceous coelomycete Pyrenochaeta sp. strain UM 256, isolated from skin scraping . Genome Announc . 2013 ;1 (3 ). 30 Teixeira MM , de Almeida LG , Kubitschek-Barreira P , Alves FL , Kioshima ES , Abadio AK , et al Comparative genomics of the major fungal agents of human and animal Sporotrichosis: Sporothrix schenckii and Sporothrix brasiliensis . BMC Genomics . 2014 ;15 :943 10.1186/1471-2164-15-943 25351875 31 Chen Z , Martinez DA , Gujja S , Sykes SM , Zeng Q , Szaniszlo PJ , et al Comparative genomic and transcriptomic analysis of wangiella dermatitidis, a major cause of phaeohyphomycosis and a model black yeast human pathogen . G3 (Bethesda) . 2014 ;4 (4 ):561 –78 .24496724 32 Seidl MF , Thomma BP . Sex or no sex: evolutionary adaptation occurs regardless . Bioessays . 2014 ;36 (4 ):335 –45 .24531982 33 Santana MF , Silva JC , Mizubuti ES , Araujo EF , Condon BJ , Turgeon BG , et al Characterization and potential evolutionary impact of transposable elements in the genome of Cochliobolus heterostrophus . BMC Genomics . 2014 ;15 :536 10.1186/1471-2164-15-536 24973942 34 Yew SM , Chan CL , Kuan CS , Toh YF , Ngeow YF , Na SL , et al The genome of newly classified Ochroconis mirabilis: Insights into fungal adaptation to different living conditions . BMC Genomics . 2016 ;17 (1 ):91 .26842951 35 Li J , Zhang JM , Li X , Suo F , Zhang MJ , Hou W , et al A piggyBac transposon-based mutagenesis system for the fission yeast Schizosaccharomyces pombe . Nucleic Acids Res . 2011 ;39 (6 ):e40 10.1093/nar/gkq1358 21247877 36 Ding S , Wu X , Li G , Han M , Zhuang Y , Xu T . Efficient transposition of the piggyBac (PB) transposon in mammalian cells and mice . Cell . 2005 ;122 (3 ):473 –83 . 16096065 37 Wu SC , Meir YJ , Coates CJ , Handler AM , Pelczar P , Moisyadi S , et al piggyBac is a flexible and highly active transposon as compared to sleeping beauty, Tol2, and Mos1 in mammalian cells . Proc Natl Acad Sci U S A . 2006 ;103 (41 ):15008 –13 . 17005721 38 Wilson MH , Coates CJ , George AL , Jr. PiggyBac transposon-mediated gene transfer in human cells . Mol Ther . 2007 ;15 (1 ):139 –45 . 17164785 39 Zhao Z , Liu H , Wang C , Xu JR . Correction: Comparative analysis of fungal genomes reveals different plant cell wall degrading capacity in fungi . BMC Genomics . 2014 ;15 :6 10.1186/1471-2164-15-6 24422981 40 Badali H , Gueidan C , Najafzadeh MJ , Bonifaz A , van den Ende AH , de Hoog GS . Biodiversity of the genus Cladophialophora . Stud Mycol . 2008 ;61 :175 –91 . 10.3114/sim.2008.61.18 19287540 41 Langston JA , Shaghasi T , Abbate E , Xu F , Vlasenko E , Sweeney MD . Oxidoreductive cellulose depolymerization by the enzymes cellobiose dehydrogenase and glycoside hydrolase 61 . Appl Environ Microbiol . 2011 ;77 (19 ):7007 –15 . 10.1128/AEM.05815-11 21821740 42 Raymond JSL , Lokesh J , Donald WR . Adaptation of proteases and carbohydrases of saprophytic, phytopathogenic and entomopathogenic fungi to the requirements of their ecological niches . Microbiology . 1997 ;143 :1983 –92 . 9202474 43 Sriranganadane D , Waridel P , Salamin K , Reichard U , Grouzmann E , Neuhaus JM , et al Aspergillus protein degradation pathways with different secreted protease sets at neutral and acidic pH . J Proteome Res . 2010 ;9 (7 ):3511 –9 . 10.1021/pr901202z 20486678 44 da Silva BA , dos Santos AL , Barreto-Bergter E , Pinto MR . Extracellular peptidase in the fungal pathogen Pseudallescheria boydii . Curr Microbiol . 2006 ;53 (1 ):18 –22 . 16775782 45 Singh B , Fleury C , Jalalvand F , Riesbeck K . Human pathogens utilize host extracellular matrix proteins laminin and collagen for adhesion and invasion of the host . FEMS Microbiol Rev . 2012 ;36 (6 ):1122 –80 . 10.1111/j.1574-6976.2012.00340.x 22537156 46 Sharpton TJ , Stajich JE , Rounsley SD , Gardner MJ , Wortman JR , Jordar VS , et al Comparative genomic analyses of the human fungal pathogens Coccidioides and their relatives . Genome Res . 2009 ;19 (10 ):1722 –31 . 10.1101/gr.087551.108 19717792 47 Desjardins CA , Champion MD , Holder JW , Muszewska A , Goldberg J , Bailao AM , et al Comparative genomic analysis of human fungal pathogens causing paracoccidioidomycosis . PLoS Genet . 2011 ;7 (10 ):e1002345 10.1371/journal.pgen.1002345 22046142 48 Vu K , Tham R , Uhrig JP , Thompson GR 3rd, Na Pombejra S , Jamklang M , et al Invasion of the central nervous system by Cryptococcus neoformans requires a secreted fungal metalloprotease . MBio . 2014 ;5 (3 ):e01101 –14 . 10.1128/mBio.01101-14 24895304 49 Monod M , Capoccia S , Lechenne B , Zaugg C , Holdom M , Jousson O . Secreted proteases from pathogenic fungi . Int J Med Microbiol . 2002 ;292 (5–6 ):405 –19 . 12452286 50 Schaller M , Korting HC , Borelli C , Hamm G , Hube B . Candida albicans-secreted aspartic proteinases modify the epithelial cytokine response in an in vitro model of vaginal candidiasis . Infect Immun . 2005 ;73 (5 ):2758 –65 . 15845479 51 Schaller M , Schackert C , Korting HC , Januschke E , Hube B . Invasion of Candida albicans correlates with expression of secreted aspartic proteinases during experimental infection of human epidermis . J Invest Dermatol . 2000 ;114 (4 ):712 –7 . 10733678 52 Naglik JR , Challacombe SJ , Hube B . Candida albicans secreted aspartyl proteinases in virulence and pathogenesis . Microbiol Mol Biol Rev . 2003 ;67 (3 ):400 –28 . 12966142 53 Naglik JR , Rodgers CA , Shirlaw PJ , Dobbie JL , Fernandes-Naglik LL , Greenspan D , et al Differential expression of Candida albicans secreted aspartyl proteinase and phospholipase B genes in humans correlates with active oral and vaginal infections . J Infect Dis . 2003 ;188 (3 ):469 –79 . 12870130 54 Naglik J , Albrecht A , Bader O , Hube B . Candida albicans proteinases and host/pathogen interactions . Cell Microbiol . 2004 ;6 (10 ):915 –26 . Epub 2004/09/02. 15339267 55 Schaller M , Schafer W , Korting HC , Hube B . Differential expression of secreted aspartyl proteinases in a model of human oral candidosis and in patient samples from the oral cavity . Mol Microbiol . 1998 ;29 (2 ):605 –15 . 9720876 56 Schaller M , Korting HC , Schafer W , Bastert J , Chen W , Hube B . Secreted aspartic proteinase (Sap) activity contributes to tissue damage in a model of human oral candidosis . Mol Microbiol . 1999 ;34 (1 ):169 –80 . 10540295 57 Smolenski G , Sullivan PA , Cutfield SM , Cutfield JF . Analysis of secreted aspartic proteinases from Candida albicans: purification and characterization of individual Sap1, Sap2 and Sap3 isoenzymes . Microbiology . 1997 ;143 (Pt 2 ):349 –56 . 9043112 58 Marti-Renom MA , Stuart AC , Fiser A , Sanchez R , Melo F , Sali A . Comparative protein structure modeling of genes and genomes . Annu Rev Biophys Biomol Struct . 2000 ;29 :291 –325 . 10940251 59 Davies DR . The structure and function of the aspartic proteinases . Annu Rev Biophys Biophys Chem . 1990 ;19 :189 –215 . 2194475 60 Borelli C , Ruge E , Schaller M , Monod M , Korting HC , Huber R , et al The crystal structure of the secreted aspartic proteinase 3 from Candida albicans and its complex with pepstatin A . Proteins . 2007 ;68 (3 ):738 –48 . 17510964 61 Keller NP , Turner G , Bennett JW . Fungal secondary metabolism—from biochemistry to genomics . Nat Rev Microbiol . 2005 ;3 (12 ):937 –47 . 16322742 62 Wheeler MH , Bell AA . Melanins and their importance in pathogenic fungi . Curr Top Med Mycol . 1988 ;2 :338 –87 . 3288360 63 Mayorga ME , Timberlake WE . The developmentally regulated Aspergillus nidulans wA gene encodes a polypeptide homologous to polyketide and fatty acid synthases . Mol Gen Genet . 1992 ;235 (2–3 ):205 –12 . 1465094 64 Wang Y , Wang J , Cheong YH , Hur JS . Three new non-reducing polyketide synthase genes from the lichen-forming fungus Usnea longissima . Mycobiology . 2014 ;42 (1 ):34 –40 . 10.5941/MYCO.2014.42.1.34 24808732 65 Kroken S , Glass NL , Taylor JW , Yoder OC , Turgeon BG . Phylogenomic analysis of type I polyketide synthase genes in pathogenic and saprobic ascomycetes . Proc Natl Acad Sci U S A . 2003 ;100 (26 ):15670 –5 . 14676319 66 Eliahu N , Igbaria A , Rose MS , Horwitz BA , Lev S . Melanin biosynthesis in the maize pathogen Cochliobolus heterostrophus depends on two mitogen-activated protein kinases, Chk1 and Mps1, and the transcription factor Cmr1 . Eukaryot Cell . 2007 ;6 (3 ):421 –9 . 17237364 67 Weinberg ED . The role of iron in protozoan and fungal infectious diseases . J Eukaryot Microbiol . 1999 ;46 (3 ):231 –8 . 10377984 68 Halliwell B , Gutteridge JM . Oxygen toxicity, oxygen radicals, transition metals and disease . Biochem J . 1984 ;219 (1 ):1 –14 . 6326753 69 Schrettl M , Bignell E , Kragl C , Joechl C , Rogers T , Arst HN Jr., et al Siderophore biosynthesis but not reductive iron assimilation is essential for Aspergillus fumigatus virulence . J Exp Med . 2004 ;200 (9 ):1213 –9 . 15504822 70 Oide S , Moeder W , Krasnoff S , Gibson D , Haas H , Yoshioka K , et al NPS6, encoding a nonribosomal peptide synthetase involved in siderophore-mediated iron metabolism, is a conserved virulence determinant of plant pathogenic ascomycetes . Plant Cell . 2006 ;18 (10 ):2836 –53 . 17056706 71 Condon BJ , Oide S , Gibson DM , Krasnoff SB , Turgeon BG . Reductive iron assimilation and intracellular siderophores assist extracellular siderophore-driven iron homeostasis and virulence . Mol Plant Microbe Interact . 2014 ;27 (8 ):793 –808 . 10.1094/MPMI-11-13-0328-R 24762221 72 Stack D , Neville C , Doyle S . Nonribosomal peptide synthesis in Aspergillus fumigatus and other fungi . Microbiology . 2007 ;153 (Pt 5 ):1297 –306 . 17464044 73 Yuan WM , Gentil GD , Budde AD , Leong SA . Characterization of the Ustilago maydis sid2 gene, encoding a multidomain peptide synthetase in the ferrichrome biosynthetic gene cluster . J Bacteriol . 2001 ;183 (13 ):4040 –51 . 11395469 74 Oberegger H , Schoeser M , Zadra I , Abt B , Haas H . SREA is involved in regulation of siderophore biosynthesis, utilization and uptake in Aspergillus nidulans . Mol Microbiol . 2001 ;41 (5 ):1077 –89 . 11555288 75 Lian T , Simmer MI , D'Souza CA , Steen BR , Zuyderduyn SD , Jones SJ , et al Iron-regulated transcription and capsule formation in the fungal pathogen Cryptococcus neoformans . Mol Microbiol . 2005 ;55 (5 ):1452 –72 . 15720553 76 Eichhorn H , Lessing F , Winterberg B , Schirawski J , Kamper J , Muller P , et al A ferroxidation/permeation iron uptake system is required for virulence in Ustilago maydis . Plant Cell . 2006 ;18 (11 ):3332 –45 . 17138696 77 Ganavalli SA , Raghavendra DK . Cladophialophora bantiana, the neurotropic fungus–A mini review . Journal of Clinical and Diagnostic Research . 2011 ;5 (6 ):1301 –6 . 78 Brown AJ , Budge S , Kaloriti D , Tillmann A , Jacobsen MD , Yin Z , et al Stress adaptation in a pathogenic fungus . J Exp Biol . 2014 ;217 (Pt 1 ):144 –55 . 10.1242/jeb.088930 24353214 79 Nicholls S , MacCallum DM , Kaffarnik FA , Selway L , Peck SC , Brown AJ . Activation of the heat shock transcription factor Hsf1 is essential for the full virulence of the fungal pathogen Candida albicans . Fungal Genet Biol . 2011 ;48 (3 ):297 –305 . 10.1016/j.fgb.2010.08.010 20817114 80 Brown AJ , Haynes K , Quinn J . Nitrosative and oxidative stress responses in fungal pathogenicity . Curr Opin Microbiol . 2009 ;12 (4 ):384 –91 . 10.1016/j.mib.2009.06.007 19616469 81 Missall TA , Lodge JK . Function of the thioredoxin proteins in Cryptococcus neoformans during stress or virulence and regulation by putative transcriptional modulators . Mol Microbiol . 2005 ;57 (3 ):847 –58 . 16045626 82 Hwang CS , Rhie GE , Oh JH , Huh WK , Yim HS , Kang SO . Copper- and zinc-containing superoxide dismutase (Cu/ZnSOD) is required for the protection of Candida albicans against oxidative stresses and the expression of its full virulence . Microbiology . 2002 ;148 (Pt 11 ):3705 –13 . 12427960 83 White TC . Increased mRNA levels of ERG16, CDR, and MDR1 correlate with increases in azole resistance in Candida albicans isolates from a patient infected with human immunodeficiency virus . Antimicrob Agents Chemother . 1997 ;41 (7 ):1482 –7 . 9210670 84 Perea S . [Azole resistance in Candida albicans ]. Rev Esp Quimioter . 2000 ;13 (3 ):314 –7 . 11086283 85 Howard SJ , Cerar D , Anderson MJ , Albarrag A , Fisher MC , Pasqualotto AC , et al Frequency and evolution of Azole resistance in Aspergillus fumigatus associated with treatment failure . Emerg Infect Dis . 2009 ;15 (7 ):1068 –76 . 10.3201/eid1507.090043 19624922 86 Flowers SA , Colon B , Whaley SG , Schuler MA , Rogers PD . Contribution of clinically derived mutations in ERG11 to azole resistance in Candida albicans . Antimicrob Agents Chemother . 2015 ;59 (1 ):450 –60 . 10.1128/AAC.03470-14 25385095 87 Xiang MJ , Liu JY , Ni PH , Wang S , Shi C , Wei B , et al Erg11 mutations associated with azole resistance in clinical isolates of Candida albicans . FEMS Yeast Res . 2013 ;13 (4 ):386 –93 . 10.1111/1567-1364.12042 23480635 88 Manastir L , Ergon MC , Yucesoy M . Investigation of mutations in Erg11 gene of fluconazole resistant Candida albicans isolates from Turkish hospitals . Mycoses . 2011 ;54 (2 ):99 –104 . 10.1111/j.1439-0507.2009.01766.x 19732347 89 Sellam A , Tebbji F , Nantel A . Role of Ndt80p in sterol metabolism regulation and azole resistance in Candida albicans . Eukaryot Cell . 2009 ;8 (8 ):1174 –83 . 10.1128/EC.00074-09 19542309 90 Lamping E , Monk BC , Niimi K , Holmes AR , Tsao S , Tanabe K , et al Characterization of three classes of membrane proteins involved in fungal azole resistance by functional hyperexpression in Saccharomyces cerevisiae . Eukaryot Cell . 2007 ;6 (7 ):1150 –65 . 17513564 91 Costa C , Dias PJ , Sa-Correia I , Teixeira MC . MFS multidrug transporters in pathogenic fungi: do they have real clinical impact? Front Physiol . 2014 ;5 :197 10.3389/fphys.2014.00197 24904431 92 Paulsen IT , Skurray RA . Topology, structure and evolution of two families of proteins involved in antibiotic and antiseptic resistance in eukaryotes and prokaryotes—an analysis . Gene . 1993 ;124 (1 ):1 –11 . 8440470 93 Liu XH , Lu JP , Dong B , Gu Y , Lin FC . Disruption of MoCMK1, encoding a putative calcium/calmodulin-dependent kinase, in Magnaporthe oryzae . Microbiol Res . 2010 ;165 (5 ):402 –10 . 10.1016/j.micres.2009.08.007 19837571 94 Rosberg-Cody E , Liavonchanka A , Gobel C , Ross RP , O'Sullivan O , Fitzgerald GF , et al Myosin-cross-reactive antigen (MCRA) protein from Bifidobacterium breve is a FAD-dependent fatty acid hydratase which has a function in stress protection . BMC Biochem . 2011 ;12 :9 10.1186/1471-2091-12-9 21329502 95 Bischoff M , Dunman P , Kormanec J , Macapagal D , Murphy E , Mounts W , et al Microarray-based analysis of the Staphylococcus aureus sigmaB regulon . J Bacteriol . 2004 ;186 (13 ):4085 –99 . 15205410 96 McNally MT , Free SJ . Isolation and characterization of a Neurospora glucose-repressible gene . Curr Genet . 1988 ;14 (6 ):545 –51 . 2977301 97 Pao SS , Paulsen IT , Saier MH Jr. Major facilitator superfamily . Microbiol Mol Biol Rev . 1998 ;62 (1 ):1 –34 . 9529885 98 Hansen KR , Hazan I , Shanker S , Watt S , Verhein-Hansen J , Bahler J , et al H3K9me-independent gene silencing in fission yeast heterochromatin by Clr5 and histone deacetylases . PLoS Genet . 2011 ;7 (1 ):e1001268 10.1371/journal.pgen.1001268 21253571 99 Yew SM , Chan CL , Lee KW , Na SL , Tan R , Hoh CC , et al A five-year survey of dematiaceous fungi in a tropical hospital reveals potential opportunistic species . PLoS One . 2014 ;9 (8 ):e104352 10.1371/journal.pone.0104352 25098697 100 Moretti S , Armougom F , Wallace IM , Higgins DG , Jongeneel CV , Notredame C . The M-Coffee web server: a meta-method for computing multiple sequence alignments by combining alternative alignment methods . Nucleic Acids Res . 2007 ;35 (Web Server issue):W645 –W8 . 17526519 101 Zerbino DR , Birney E . Velvet: algorithms for de novo short read assembly using de Bruijn graphs . Genome Res . 2008 ;18 (5 ):821 –9 . 10.1101/gr.074492.107 18349386 102 Boetzer M , Henkel CV , Jansen HJ , Butler D , Pirovano W . Scaffolding pre-assembled contigs using SSPACE . Bioinformatics . 2011 ;27 (4 ):578 –9 . 10.1093/bioinformatics/btq683 21149342 103 Boetzer M , Pirovano W . Toward almost closed genomes with GapFiller . Genome Biol . 2012 ;13 (6 ):R56 10.1186/gb-2012-13-6-r56 22731987 104 Nadalin F , Vezzi F , Policriti A . GapFiller: a de novo assembly approach to fill the gap within paired reads . BMC Bioinformatics . 2012 ;13 Suppl 14 :S8 10.1186/1471-2105-13-S14-S8 23095524 105 Lomsadze A , Ter-Hovhannisyan V , Chernoff YO , Borodovsky M . Gene identification in novel eukaryotic genomes by self-training algorithm . Nucleic Acids Res . 2005 ;33 (20 ):6494 –506 . 16314312 106 Quevillon E , Silventoinen V , Pillai S , Harte N , Mulder N , Apweiler R , et al InterProScan: protein domains identifier . Nucleic Acids Res . 2005 ;33 (Web Server issue):W116 –20 . 15980438 107 Lagesen K , Hallin P , Rodland EA , Staerfeldt HH , Rognes T , Ussery DW . RNAmmer: consistent and rapid annotation of ribosomal RNA genes . Nucleic Acids Res . 2007 ;35 (9 ):3100 –8 . 17452365 108 Lowe TM , Eddy SR . tRNAscan-SE: a program for improved detection of transfer RNA genes in genomic sequence . Nucleic Acids Res . 1997 ;25 (5 ):955 –64 . 9023104 109 Conesa A , Gotz S , Garcia-Gomez JM , Terol J , Talon M , Robles M . Blast2GO: a universal tool for annotation, visualization and analysis in functional genomics research . Bioinformatics . 2005 ;21 (18 ):3674 –6 . 16081474 110 Tatusov RL , Fedorova ND , Jackson JD , Jacobs AR , Kiryutin B , Koonin EV , et al The COG database: an updated version includes eukaryotes . BMC Bioinformatics . 2003 ;4 :41 12969510 111 Kuan CS , Yew SM , Chan CL , Toh YF , Lee KW , Cheong WH , et al DemaDb: an integrated dematiaceous fungal genomes database . Database (Oxford) . 2016 ;2016 . 112 Yin Y , Mao X , Yang J , Chen X , Mao F , Xu Y . dbCAN: a web resource for automated carbohydrate-active enzyme annotation . Nucleic Acids Res . 2012 ;40 (Web Server issue):W445 –51 . 10.1093/nar/gks479 22645317 113 Ohm RA , Feau N , Henrissat B , Schoch CL , Horwitz BA , Barry KW , et al Diverse lifestyles and strategies of plant pathogenesis encoded in the genomes of eighteen Dothideomycetes fungi . PLoS Pathog . 2012 ;8 (12 ):e1003037 10.1371/journal.ppat.1003037 23236275 114 Rawlings ND , Barrett AJ , Bateman A . MEROPS: the database of proteolytic enzymes, their substrates and inhibitors . Nucleic Acids Res . 2012 ;40 (Database issue):D343 –50 . 10.1093/nar/gkr987 22086950 115 Petersen TN , Brunak S , von Heijne G , Nielsen H . SignalP 4.0: discriminating signal peptides from transmembrane regions . Nat Methods . 2011 ;8 (10 ):785 –6 . 10.1038/nmeth.1701 21959131 116 Krogh A , Larsson B , von Heijne G , Sonnhammer EL . Predicting transmembrane protein topology with a hidden Markov model: application to complete genomes . J Mol Biol . 2001 ;305 (3 ):567 –80 . 11152613 117 Khaldi N , Seifuddin FT , Turner G , Haft D , Nierman WC , Wolfe KH , et al SMURF: Genomic mapping of fungal secondary metabolite clusters . Fungal Genet Biol . 2010 ;47 (9 ):736 –41 . 10.1016/j.fgb.2010.06.003 20554054 118 Rutherford K , Parkhill J , Crook J , Horsnell T , Rice P , Rajandream MA , et al Artemis: sequence visualization and annotation . Bioinformatics . 2000 ;16 (10 ):944 –5 . 11120685 119 Li L , Stoeckert CJ Jr., Roos DS . OrthoMCL: identification of ortholog groups for eukaryotic genomes . Genome Res . 2003 ;13 (9 ):2178 –89 . 12952885 120 Larkin MA , Blackshields G , Brown NP , Chenna R , McGettigan PA , McWilliam H , et al Clustal W and Clustal X version 2.0 . Bioinformatics . 2007 ;23 (21 ):2947 –8 . 17846036 121 Ronquist F , Huelsenbeck JP . MrBayes 3: Bayesian phylogenetic inference under mixed models . Bioinformatics . 2003 ;19 (12 ):1572 –4 . 12912839 122 Buchan DW , Ward SM , Lobley AE , Nugent TC , Bryson K , Jones DT . Protein annotation and modelling servers at University College London . Nucleic Acids Res . 2010 ;38 (Web Server issue):W563 –8 . 10.1093/nar/gkq427 20507913 123 Biasini M , Bienert S , Waterhouse A , Arnold K , Studer G , Schmidt T , et al SWISS-MODEL: modelling protein tertiary and quaternary structure using evolutionary information . Nucleic Acids Res . 2014 ;42 (Web Server issue):W252 –8 . 10.1093/nar/gku340 24782522 124 Kuan CS , Tien CF , See Too WC , Few LL . Sequence analysis and homology modeling of glyceraldehyde-3-phosphate dehydrogenase of an antarctic psychrophilic Pseudomonas sp . African Journal of Microbiology Research . 2012 ;6 (23 ):881 –91 .

PMC005xxxxxx/PMC5003358.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 10.1371/journal.pone.0162119PONE-D-16-33158CorrectionCorrection: Arabidopsis thaliana Contains Both Ni2+ and Zn2+ Dependent Glyoxalase I Enzymes and Ectopic Expression of the Latter Contributes More towards Abiotic Stress Tolerance in E. coli The PLOS ONE Staff 29 8 2016 2016 11 8 e0162119© 2016 The PLOS ONE Staff2016The PLOS ONE StaffThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Arabidopsis thaliana Contains Both Ni2+ and Zn2+ Dependent Glyoxalase I Enzymes and Ectopic Expression of the Latter Contributes More towards Abiotic Stress Tolerance in E. coli ==== Body The affiliation for the second author is incorrect. Rituraj Batth is not affiliated with #2 but with #1 Faculty of Life Sciences and Biotechnology, Plant Molecular Biology Laboratory, South Asian University, Akbar Bhawan, Chanakyapuri, New Delhi 110021, India. The publisher apologizes for the error. ==== Refs Reference 1 Jain M , Batth R , Kumari S , Mustafiz A (2016 ) Arabidopsis thaliana Contains Both Ni2+ and Zn2+ Dependent Glyoxalase I Enzymes and Ectopic Expression of the Latter Contributes More towards Abiotic Stress Tolerance in E. coli . PLoS ONE 11 (7 ): e0159348 doi:10.1371/journal.pone.0159348 27415831

PMC005xxxxxx/PMC5003359.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757125510.1371/journal.pone.0161691PONE-D-16-07183Research ArticleMedicine and Health SciencesDiagnostic MedicineSigns and SymptomsLesionsMedicine and Health SciencesPathology and Laboratory MedicineSigns and SymptomsLesionsMedicine and Health SciencesDiagnostic MedicineDiagnostic RadiologyMagnetic Resonance ImagingResearch and Analysis MethodsImaging TechniquesDiagnostic RadiologyMagnetic Resonance ImagingMedicine and Health SciencesRadiology and ImagingDiagnostic RadiologyMagnetic Resonance ImagingPeople and PlacesPopulation GroupingsProfessionsRadiologistsBiology and Life SciencesAnatomyCardiovascular AnatomyBlood VesselsArteriesMedicine and Health SciencesAnatomyCardiovascular AnatomyBlood VesselsArteriesMedicine and Health SciencesOncologyCancers and NeoplasmsBreast TumorsBreast CancerMedicine and Health SciencesOncologyCancers and NeoplasmsMalignant TumorsMedicine and Health SciencesDiagnostic MedicineCancer Detection and DiagnosisMedicine and Health SciencesOncologyCancer Detection and DiagnosisMedicine and Health SciencesRadiology and ImagingBreast Contrast Enhanced MR Imaging: Semi-Automatic Detection of Vascular Map and Predominant Feeding Vessel Vascular Map Detection on Breast Contrast Enhanced MR ImagingPetrillo Antonella 1Fusco Roberta 1*Filice Salvatore 1Granata Vincenza 1Catalano Orlando 1Vallone Paolo 1Di Bonito Maurizio 2D’Aiuto Massimiliano 3Rinaldo Massimo 3Capasso Immacolata 3Sansone Mario 41 Division of Radiology, Department of Diagnostic Imaging, radiant and metabolic Therapy, “Istituto Nazionale Tumori Fondazione Giovanni Pascale–IRCCS”, Via Mariano Semmola, Naples, Italy2 Division of Diagnostic Pathology, Department of Diagnostic and Laboratory Pathology “Istituto Nazionale Tumori Fondazione Giovanni Pascale–IRCCS”, Via Mariano Semmola, Naples, Italy3 Divison of Senology Surgery, Department of Senology “Istituto Nazionale Tumori Fondazione Giovanni Pascale–IRCCS”, Via Mariano Semmola, Naples, Italy4 Department of Electrical Engineering and Information Technologies, University “Federico II” of Naples, Via Claudio, Naples, ItalyAlberich-Bayarri Angel EditorGeneralitat Valenciana, SPAINCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: AP RF MS. Data curation: AP RF VG SF OC PV MDB MDA MR IC MS. Formal analysis: RF MS. Investigation: AP RF VG SF OC PV MDB MDA MR IC MS. Methodology: AP RF MS. Software: RF MS. Supervision: AP RF MS. Validation: AP RF MS. Visualization: AP RF MS. Writing – original draft: AP RF MS. Writing – review & editing: AP RF MS. * E-mail: r.fusco@istitutotumori.na.it29 8 2016 2016 11 8 e01616914 3 2016 10 8 2016 © 2016 Petrillo et al2016Petrillo et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Purpose To obtain breast vascular map and to assess correlation between predominant feeding vessel and tumor location with a semi-automatic method compared to conventional radiologic reading. Methods 148 malignant and 75 benign breast lesions were included. All patients underwent bilateral MR imaging. Written informed consent was obtained from the patients before MRI. The local ethics committee granted approval for this study. Semi-automatic breast vascular map and predominant vessel detection was performed on MRI, for each patient. Semi-automatic detection (depending on grey levels threshold manually chosen by radiologist) was compared with results of two expert radiologists; inter-observer variability and reliability of semi-automatic approach were assessed. Results Anatomic analysis of breast lesions revealed that 20% of patients had masses in internal half, 50% in external half and the 30% in subareolar/central area. As regards the 44 tumors in internal half, based on radiologic consensus, 40 demonstrated a predominant feeding vessel (61% were supplied by internal thoracic vessels, 14% by lateral thoracic vessels, 16% by both thoracic vessels and 9% had no predominant feeding vessel—p<0.01), based on semi-automatic detection, 38 tumors demonstrated a predominant feeding vessel (66% were supplied by internal thoracic vessels, 11% by lateral thoracic vessels, 9% by both thoracic vessels and 14% had no predominant feeding vessel—p<0.01). As regards the 111 tumors in external half, based on radiologic consensus, 91 demonstrated a predominant feeding vessel (25% were supplied by internal thoracic vessels, 39% by lateral thoracic vessels, 18% by both thoracic vessels and 18% had no predominant feeding vessel—p<0.01), based on semi-automatic detection, 94 demonstrated a predominant feeding vessel (27% were supplied by internal thoracic vessels, 45% by lateral thoracic vessels, 4% by both thoracic vessels and 24% had no predominant feeding vessel—p<0.01). An excellent agreement between two radiologic assessments (k = 0.81) and between radiologic consensus and semi-automatic assessment (k = 0.80) was found to identify origin of predominant feeding vessel. An excellent reliability for semi-automatic assessment (Cronbach's alpha = 0.96) was reported. Conclusions Predominant feeding vessel location was correlated with breast lesion location: internal thoracic artery supplied the highest proportion of breasts with tumor in internal half and lateral thoracic artery supplied the highest proportion of breasts with lateral tumor. The authors received no specific funding for this work. Data AvailabilityAll relevant data are within the paper.Data Availability All relevant data are within the paper. ==== Body Introduction Magnetic resonance imaging (MRI) has become an important imaging modality in the breast cancer diagnostic work-up. MR is the most accurate imaging method for non-invasive breast cancer detection, having nearly 100% sensitivity [1]. The specificity, however, is only moderate reaching an overall specificity of 72% in 44 studies as reported by Peters et al. [1] and varies widely across studies in relation with cancer prevalence and the criteria used to differentiate malignant and benign lesions [1]. Breast Contrast Enhanced Magnetic Resonance Imaging (CE-MRI) provides the study of breast vascular map by means of the enhancement of an injected paramagnetic contrast agent. Maximum intensity projections (MIPs) obtained from post-processing of subtraction images can reveal not only the presence of enhancing lesions but also the angiographic vessel vascular map of the whole bilateral breast. The diagnostic value of breast vascular maps has recently been explored [2–3] and an increased vascularity adjacent to lesion and in the ipsilateral breast as a whole has been found [4–9]. The presence of vessels adjacent to one or more lesions has been investigated in several papers [7, 10]. In 2002, a study by Carriero et al. [5] reported 89% of sensitivity and 83% of specificity for perilesional or intralesional vessels. Kul et al. [7] showed that both ipsilateral increased vascularity (sensitivity and specificity of 62% and 79%, respectively) and the adjacent vessel sign (sensitivity and specificity of 74% and 89%, respectively) are associated with breast cancer in a significant percentage of patients. In 2005, Malich et al. [10] defined the adjacent vessel as “a prominent vessel leading to the lesion, seen in subtraction images” and reported it as “a minor sign of malignancy” present in 63% of 268 malignant lesions and in 28% of 60 benign lesions. Fischer et al. [11] obtained a sensitivity of 83% and a specificity of 87% on 122 patients (71 malignant lesion and 51 benign lesion). Dietzel et al. [12] examined the diagnostic value of the adjacent vessel on a large population (1.084 cases) reporting a sensitivity of the adjacent vessel sign of 47%, a specificity of 88%, a positive likelihood ratio of 3.8, a negative likelihood ratio of 0.6 and diagnostic odds ratio of 6.3. Finally, Grubstein et al. [13] characterized the alterations in blood supply by location of the tumor (medial versus lateral) within the breast using MRI. The study involved 105 patients with a cancer prevalence of 49%. A dominant vessel was noted in 47 of 51 cases (92%). A predominant medial vascular supply, connected to the internal mammary vessels, was seen in 87% of medial tumors and 48% of lateral tumors. In most studies present in literature, the critical issues were the subjectivity of visual methods for vessel analysis, the quantification suffered from intra- and inter-observer variability and the approach could be time consuming considering the need to measure the size of each vessel and then to identify the predominant feeding vessel. In this scenario, it is desirable the development of a software dedicated to the semi-automatic detection of vessels within the breast and of the predominant feeding vessel. Moreover, it is desirable to assess the correlation between breast cancer site and predominant feeding vessel location. In this study, we propose a semi-automatic method to obtain the breast vascular map and the predominant feeding vessel and to correlate the latter with breast lesion location. We have compared the findings of the semi-automatic approach with the results of two readers expert radiologists and we have further evaluated the inter-observer variability and a measure of semi-automatic approach reliability. Material and Methods Study population Retrospectively, a single-institution database was reviewed for patients diagnosed with breast cancer and who were surgically treated for breast cancer between 2011 and 2014. The records of 3499 consecutive patients were analyzed. Of these, 298 patients underwent breast contrast-enhanced MRI, between 7th and 14th day of menstrual cycle for women in premenopausal phase, with a dedicated equipment because of abnormal mammographic or ultrasound or clinical findings. Among these, patients with unilateral and unifocal histo-pathologically-confirmed lesions were included in this study. Patients with bilateral or multifocal breast cancer or with a history of radiation therapy or breast biopsy within 6 months or with no histologic confirmation of the lesion, who had undergone unilateral mastectomy for a previous diagnosis of breast cancer, or who had undergone unilateral imaging were excluded by consensus of two expert radiologists. An overall of 223 patients (age range, 20–73 years; mean, 45 years), with a defined breast lesion, visible vessels on contrast enhanced MR acquisitions and no evident motion artifacts on subtraction images were included. 148 had a malignant tissue diagnosis and 75 had a benign tissue diagnosis. All patients underwent bilateral imaging. Written informed consent was obtained from each patient before MRI. No personal medical information about an identifiable living individual was presented in this study. Approval for this study was granted by the ethics committee of National Cancer Institute of Naples “Pascale Foundation”. Magnetic Resonance Imaging Protocol Dynamic Contrast Enhanced Magnetic Resonance Imaging was performed with a 1.5T breast dedicated system (Aurora; Aurora Imaging Technology, North Andover, MA), enclosing an integrated in-table coil [14]. A pre-contrast three-dimensional (3D) non-spoiled SPIRAL-RODEO fat-sat sequence (TR 29 ms, TE 4.8 ms, flip angle 45°, matrix 512x512, thickness 1.13 mm, gap 0 mm) and four dynamic post-contrast 3D spoiled SPIRAL-RODEO fat-sat acquisitions (TR 29 ms, TE 4.8 ms, flip angle 45°, matrix 512x512, thickness 1.13 mm, gap 0 mm) were obtained with temporal resolution of 90 seconds (160 slice to encompass the entire volume of interest). Gadobenate dimeglumine (Multihance, Gd-BOPTA Bracco; AtlantaPharma, Costance, Germany) was administered intravenously, as a bolus injection, at a dose of 0.2 mL⁄kg body weight, followed by flushing with 20 mL of saline via an Optistar Elite (Covidien Imaging Solution, Hazelwood, MO) automated contrast delivery system. Histopathologic Analysis and Reference Standard Breast pathologists with experience of at least 10 years performed histopathological verification. Pathology of surgical specimens served as reference standard for imaging findings. Benign breast lesions were divided into the following subcategories: fibroadenoma, phyllodes tumor, papilloma and fibrocystic changes. Malignant lesions were histopathologically classified according to the World Health Organization (WHO) classification of breast carcinoma. The size of the tumor was measured across the largest cross sectional dimension. The intensity, extent, and subcellular distribution of estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2) were evaluated as previously described by Crispo et al [15]. Image Analysis Radiologic Evaluation Two dedicated radiologists with 20 years of experience (A.P.) and 15 years of experience (S.F.) interpreted breast MRI scans. The radiologists provided subjective evaluation of images both individually and in consensus; in particular, they proceeded by means of visual inspection of the vertical Maximum Intensity Projection (MIP) of the subtraction between the first post-contrast (1.5 min after IV injection) and the pre-contrast acquisition. The radiologists were uninformed about the pathological findings. The evaluation focused on two main aspects: Presence of a mass and location of lesion (internal, external half and subareolar/central area). Presence of a dominant feeding vessel and origin of the vessel (internal, lateral or internal/lateral thoracic artery). To individuate the dominant feeding vessel each radiologist proceeded as follows: each reader evaluated the MIPs of subtraction images from several different angles of projection in order to become acquainted with the vascular map of the patient; then he/she subjectively individuated the dominant vessels. The qualitative criteria used for establishing the presence of dominant vessels were both a “sufficient” (in the opinion of the radiologist) length and diameter: when a single vessel had both characteristics a single dominant vessel was identified (presence of a single dominant vessel); in some cases, two vessels had a comparable size and the radiologist decided for presence of two dominant vessels; in other cases, no vessel had “sufficient” size to be considered dominant (absence of dominant vessel). When a single dominant vessel was found the radiologist individuated its origin as “internal” or “lateral” thoracic; when two dominant vessels were found the origin was set to “internal and lateral” thoracic. Semi-automatic evaluation Image analysis included a blood vessel semi-automatic detection procedure. In order to obtain an optimal angiographic effect, the maximum intensity projection (MIP) was calculated by DCE-MRI data. A single vertical MIP projection was generated using the subtraction volume between the first post-contrast sequence and the pre-contrast sequence. Semi-automatic evaluation of breast vascular map was previously reported in detail by Fusco et al. [16]. Briefly, our method is composed of two steps: first, a multi-scale vessel enhancement filtering, as reported in [17], was applied to the subtracted MIP; second, morphologic operators were applied on the resulting images in order to extract vessel skeletons. In the first step, the local image structure of the subtracted MIP (Fig 1A) is analyzed at different scales using a “vesselness” measure involving local Hessian eigenvalues: per each pixel the “vesselness” can be considered the probability that it belongs to a vessel. In order to magnify the vascular map and to suppress background noise the “vesselness” image (Fig 1B) is processed by global histogram equalization (Fig 1C) and then converted into a binary (black-white) image using the Otsu’s global threshold method [18]. 10.1371/journal.pone.0161691.g001Fig 1 Semiautomatic vascular map extraction: (a) MIP obtained subtracting the first post-contrast sequence from pre-contrast sequence; (b) multiscale vessel enhancement filtering; (c) optimal contrast image after global histogram equalization; and (d) skeleton with extracted vessel in different colors (color figure online); (e) Semiautomatic vessel length and diameter measure: a vessel tracking; (f) normal segments to vessel direction. Morphologic operators were applied on the resulting image in order to detect connected regions (vessels) and to eliminate spurious pixels. Specifically, thinning operator was applied repeatedly until convergence: the obtained skeleton identified the longitudinal direction of vessels. Length of vessels was measured along the longitudinal direction; vessel diameters were measured orthogonally to longitudinal direction. In order to mimic radiologists criteria for dominant vessel identification, significant vessels were defined as having a length of at least 3 cm and a diameter of at least 2 mm [6–8, 19]. When no vessel had these requirements, no dominant vessel was identified. The dominant feeding vessel was identified as having the highest length among significant vessels. Dominant vessel was marked in red (Fig 1D). Looking at this last image the radiologist identified the origin of vascular supply (internal or lateral thoracic). However, putting vessels in descending ordered based on diameter size, it was found that in some cases the first two vessels have very similar diameter: in the case this difference was no more than 5%, they were considered both as dominant vessels and the origin was set to both internal and lateral thoracic. No training was performed by expert radiologist for both radiological and semi-automatic evaluation. Image analyses were carried out using the MATLAB R2007a software (The MathWorks Inc., Natick, MA). Statistical Analysis In order to evaluate variability between the two readers and between the radiologic consensus interpretation and the semi-automatic detection, inter-reading concordance assessment was assessed by means of Cohen's kappa coefficient (k) for categorical items (0–0.20, poor agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61–0.80, good agreement; and 0.81–1.00, excellent agreement). Cohen's kappa coefficient was used as metric to validate the predominant feeding vessel origin obtained by semi-automatic approach compared with qualitative perspective. Conbach's alpha [20] was used as a (lower bound) estimate of the reliability (internal consistence) of semi-automatic breast vascular map assessment (α<0.5 unacceptable, 0.5≤ α< 0.6 poor, 0.6≤ α< 0.7 acceptable, 0.7≤ α< 0.9 good, α≥ 0.9 excellent). Chi-square and Fisher exact test was used to assess statistically significant differences in the number of internal thoracic and lateral thoracic artery as predominant feeding vessel and correlation with lesion location. Median number of automatically detected vessels per each breast in malignant and benign lesions were compared using the Mann Whitney test for unpaired data. Chi-square test was used to assess a statistical significance between vascular map detection in the breast with lesion compared with the contralateral. Results Histological results 148 patients had malignant tumors (66.4%): 99 cases of invasive ductal carcinoma, 9 cases of invasive lobular carcinoma, 11 cases of mixed invasive ductal and invasive lobular carcinomas and 29 ductal carcinoma in situ. 75 patients had benign tumors (34.6%): 47 cases of fibroadenoma, 18 of atypical hyperplasia, six of fibrocystic change and four of intraductal papilloma (Table 1). In Table 2 were reported histopathological features of malignant lesions. 10.1371/journal.pone.0161691.t001Table 1 Pathologic diagnosis of all tumors. Malignant N. Pts Benign N. Pts IDC 99 Fibroadenoma 47 ILC 9 Fibrocystic changes 6 IDLC 11 Atypical Hyperplasia 18 DCIS 29 Intraductal Papilloma 4 Note—IDC = Invasive Ductal Carcinoma; ILC = Invasive Lobular Carcinoma; IDLC = Invasive Ductal-Lobular Carcinoma; DCIS = Ductal Carcinoma In Situ. 10.1371/journal.pone.0161691.t002Table 2 Histopatologic features of malignant lesions. Histopathologic Features N. Pts. ER/PR positive 52 HER2 positive 21 Triple positive 12 Triple negative 63 Note—ER = Estrogen Receptors; PR = Progesterone Receptors; HER2 = Human Epidermal Growth Factor Receptor 2. Lesion Location and presence of dominant feeding vessel Table 3 shows lesion location and presence of dominant feeding vessel for malignant and benign lesions. Anatomic analysis of breast lesions revealed that 20% (44/223) of patients had masses in internal half, 50% (111/223) of patients had masses in external half and the 30% (68/223) of patients had masses in subareolar/central area. 10.1371/journal.pone.0161691.t003Table 3 Tumor location and malignancy. Table entries: number of lesions. Malignancy Malignant Benign Tumor location Internal half 32 12 External half 64 47 Subareolar/Central Area 52 16 According to semi-automatic approach, the median number of vessels in the breasts containing malignant lesions was 3.9±1.6, compared with 2.8±1.4 in breasts containing benign lesions (p<0.05), and median length of predominant feeding vessel was 12.5±2.4 mm in patients with malignant tumors and 7.4±3.5 mm in patients with benign tumors. These differences were statistically significant (p<0.05 at Mann Whitney test). There were not differences statistically significant in the median number of vessels and in the median size (length and diameter) of predominant feeding vessel among different subgroups of malignant lesions (p>0.05 at Mann Whitney test). Instead there was a difference statistically significant (p<0.05 at Mann Whitney test) between fibroadenoma lesions and other benign lesions in the median number of breast vessel (2.9±1.5 versus 2.0±1.1) and in the median size of predominant feeding vessel (length 8.7±2.4 mm versus 6.3±1.8 mm; diameter 2.4±0.4 mm versus 2.1±0.3 mm) (Table 4). A dominant feeding vessel was present in 91% of malignant cases for both semi-automatic and radiologic evaluation, while predominant feeding vessel was present with lower percentage in benign lesion, respectively 67% and 65% (Table 4). 10.1371/journal.pone.0161691.t004Table 4 Relationship between lesion subtype and size with dominant feeding vessels. The number of feeding vessels, the dominant vessel length and diameter were evaluated exclusively with the semi-automatic approach. Lesion type Lesion Size [cm] (median±SD) Number of feeding vessels (median±SD) Dominant vessel length [mm] (median±SD) Dominant vessel diameter [mm] (median±SD) Presence of dominant vessel (semiautomatic) Presence of dominant vessel (consensus reading) Malignant IDC 3.5±1.4 3.8±1.3 12.3±4.3 3.4±0.6 128 127 ILC IDLC DCIS 1.3±3.8 3.1±1.1 11.5±4.9 3.2±0.5 7 7 Benign Fibroadenoma 3.2±1.8 2.9±1.4 8.7±2.4 2.4±0.4 40 38 Fibrocystic changes 2.4±1.1 2.0±1.1 6.3±1.8 2.1±0.3 10 11 Atypical hyperplasia Intraductal papilloma A good concordance between two radiologic assessments (k = 0.78 with 95% interval confidence 0.68–0.88) and an excellent agreement between radiologic consensus interpretation and semi-automatic assessment (k = 0.88 with 95% interval confidence 0.79–0.96) were reported to identify presence or absence of predominant feeding vessel. Dominant feeding vessel location and correlation with tumor location As regards the 44 tumors in internal half, based on radiologic consensus interpretation, 40 demonstrated a predominant feeding vessel (61% were supplied by internal thoracic vessels, 14% by lateral thoracic vessels, 16% by both thoracic vessels and 9% had no predominant feeding vessel—p<0.01), based on semi-automatic detection, 38 tumors demonstrated a predominant feeding vessel (66% were supplied by internal thoracic vessels, 11% by lateral thoracic vessels, 9% by both thoracic vessels and 14% had no predominant feeding vessel—p<0.01). As regards the 111 tumors in external half, based on radiologic consensus, 91 demonstrated a predominant feeding vessel (25% were supplied by internal thoracic vessels, 39% by lateral thoracic vessels, 18% by both thoracic vessels and 18% had no predominant feeding vessel—p<0.01), based on semi-automatic detection, 94 demonstrated a predominant feeding vessel (27% were supplied by internal thoracic vessels, 45% by lateral thoracic vessels, 4% by both thoracic vessels and 24% had no predominant feeding vessel—p<0.01). The difference in the proportion of internal and lateral thoracic dominant vessels was statistically significant (p<0.01), for both internal and external masses, as shown in Table 5. No significant statistical difference was found between the proportion of internal and lateral thoracic artery when the masses were localized in subareolar/central area (Table 5). 10.1371/journal.pone.0161691.t005Table 5 Tumor location and dominant feeding vessel origin findings for single reader, for consensus reading and for semi-automatic evaluation. Table entries: number of lesions. Dominant feeding vessel origin Internal Thoracic Lateral Thoracic Internal and Lateral Thoracic No predominant feeding vessel Reader 1 Tumor location Internal half 24 6 6 8 External half 27 42 20 22 Subareolar/Central Area 20 14 18 16 Reader 2 Tumor location Internal half 25 5 7 7 External half 27 42 19 23 Subareolar/Central Area 19 16 17 16 Consensus reading Tumor location Internal half 27 6 7 4 External half 28 43 20 20 Subareolar/Central Area 20 15 17 16 Semi-automatic evaluation Tumor location Internal half 29 5 4 6 External half 30 50 14 17 Subareolar/Central Area 21 15 17 15 An excellent concordance between two radiologic assessments (k = 0.82 with 95% interval confidence 0.73–0.89) and an excellent agreement between radiologic consensus interpretation and semi-automatic assessment (k = 0.80 with 95% interval confidence 0.72–0.87) were reported to identify origin of predominant feeding vessel. An excellent reliability for semi-automatic assessment (Cronbach's alpha equal to 0.96) was also assessed. Fig 2 shows examples of MIP reconstruction with tumor location and relative dominant vessel. 10.1371/journal.pone.0161691.g002Fig 2 MIP reconstructions and predominant vessel: a) A lateral tumor (external quadrants) and dominant feeding vessel with origin in internal thoracic artery (automatic and radiologic assessment); b) A subareolar/central and lateral mass and dominant vessel with origin in internal and lateral thoracic artery (automatic and radiologic assessment); c) A lateral tumor (external quadrants) without dominant feeding vessel (automatic and radiologic assessment); d) A lateral tumor (external quadrants) with dominant feeding vessel with origin in lateral thoracic artery (automatic and radiologic assessment); e) A lateral tumor (external quadrants) with dominant feeding vessel with origin in lateral thoracic artery (automatic and radiologic assessment); f) A subareolar mass and dominant vessel with origin in internal thoracic artery (recognized only by automatic evaluation). Note- the triangle indicates the lesion and the arrow indicates the predominant feeding vessel. Discussion Angiogenesis is an important characteristic of breast carcinoma growth and it allows to observe differences in contrast enhancement between breast tumors and fibro-glandular tissue on CE-MRI [1–13]. The branches of internal thoracic artery provide most of the blood supply to the normal breast, especially to the inner compartments. The branches of lateral thoracic artery supply mainly the outer quadrants of the breast, as reported in [13]. MR is the best imaging modality for breast cancer detection and diagnosis. The location of the primary tumor influences survival in breast cancer, with a low outcome for tumors in inner and periareolar quadrants [21]. Specifically, patients with inner quadrant tumors had higher hazards for systemic disease relapse when compared to those located in the outer quadrants [21, 22]. We assumed that a breast with a malignant mass might show an altered blood supply according to the tumor location; this could be used as indicator of prognosis and survival and could be detected on the MIP obtained by MRI processing. In literature [1–13], the critical issues were the subjectivity of visual methods for vessel analysis and intra- and inter-observer variability of the predominant feeding vessel quantification. Orgüç et al [23] compared retrospectively bilateral 3D vascular maps obtained by dynamic CE-MRI and determined the association of one-sided vascular increase with ipsilateral breast cancer. They reported that MRI vascular mapping is an accurate method (sensitivity of 69% and specificity of 92%) for characterizing breast tissue vascularization. Radiological interpretation of predominant feeding vessel could be time consuming considering the need to measure the size of each vessel and then to identify the dominant vessel. In this scenario, it is desirable to assess the reliability of a semi-automatic approach for breast vascular map detection and for identification of predominant feeding vessel origin in correlation with tumor site. Vignati et al [24–25] proposed an automatic algorithm for vessel detection from breast DCE-MRI and evaluated the clinical impact of this algorithm to assess neo-adjuvant therapy response. Their preliminary findings suggested the use of automatic vascular maps as biomarker to assess neo-adjuvant therapy pathologic response. Lin et al [26] developed a computer-based algorithm for detecting blood vessels appearing in breast dynamic CE-MRI analyzing 34 cases: the median correct-detection rate was 85.6% (mean 84.9% +/- 7.8%), the incorrect-detection rate was 13.1% (mean 15.1% +/- 7.8%), and the missed-detection rate was 19.2% (mean 21.3% +/- 12.8%). Both Vignati et al. [24–25] and Lin et al. [26] have not assessed the correlation between predominant feeding vessel and lesion location as a sign of malignity. In this study, we proposed a semi-automatic method to obtain breast vascular map and predominant feeding vessel and to correlate dominant vessel with lesion location. We have compared the findings of semi-automatic approach with the results of two readers expert radiologists evaluating the inter-observer variability. The results indicated an excellent agreement between two radiologic assessments, radiologic consensus interpretation and semi-automatic assessment (k = 0.82 and k = 0.80, respectively); moreover, the findings showed that semi-automatic breast vascular map and predominant feeding vessel detection identified correctly, with excellent reliability, the lesion dominant vessel and its anatomic location (Cronbach's alpha equal to 0.96). Our investigation, based on semi-automatic assessment, revealed the presence of a prominent feeding vessel in 91% of the patients with a malignant tumor compared to 67% of patients with benign lesions (p<0.05 at Chi square test). Therefore, this latter could be used as sign of malignity. Moreover, our findings of semi-automatic approach revealed that an internal thoracic artery supplied a high percentage of the breasts with a tumor in internal half and that the lateral thoracic artery supplied a high percentage of the breasts with a lateral tumor (external half). These results were in accordance with Grubstein et al [13], reporting a predominant medial vascular supply connected to the internal mammary vessels in 87% of medial tumors and 48% of lateral tumors. Some limitations of this study could be mentioned. Radiologists evaluated only subjectively the size of dominant vessels, therefore it was not possible to assess inter-observer variabilities regarding vessel size. The correlation of others factors on breast tissue vascularity such as age, menopausal situation, in which the measurements have been made, habit of smoking was not evaluated and could be investigated in next studies. Moreover, the tool was tested for specific situations including unilateral and unifocal histo-pathologically confirmed enhancing breast lesions. However, next endpoint is to assess the tool for bilateral and/or multifocal breast cancer. In these case using combination of different MIP projection such as anterior-posterior and right-left some limitations including overlapping lesions and predominant vessel could be resolved. Another limit is due to vessel size being measured on projection images (MIPs). This might potentially give distorted measures, in fact, for example, measuring the diameter of a vessel having elliptical section it can be projected on the minor axis giving a wrong estimation of the vessel diameter; similarly, when measuring vessel lengths we are neglecting curvature of the vessels which might be longer in reality. Our analysis on MIP images implicitly assumed that vessel had circular section and curvature could be neglected. Future algorithm developments should consider a full 3D processing. The study is a retrospective study and we are not able to increase sample size; the dataset could be considered “unbalanced” (high number of low-vascular tumours is less than the number of high-vascular tumours) in the sense of pattern recognition jargon. It is well known that the performance of an algorithm tested on an “unbalanced” dataset could be misleading and the algorithm might be poorly generalizable. However, we considered retrospectively a study population of more than 3000 subjects from which we excluded a large number of patients in order to extract a homogeneous sample. From a statistical point of view, the sample size of this study (223 patients) allowed to perform all the statistical tests reported in the manuscript at a level of significance p<0.05 or better. Clinical impact of this study could be assessed observing in a prospective study the added diagnostic value of local and global breast vascularity in comparison with standard MR functional descriptors (such as wash-in and wash-out rate) on large consecutive series of patients for prognosis, early and presurgical therapy response assessment in advanced breast cancer. Moreover, it might interesting to evaluate the vascularity of breast tissue using MRI angiography in correlation with Doppler Ultrasound. Finally, it must be highlighted that, in this preliminary study, we focussed on defined and clearly visible breast lesions in order to evaluate the feasibility of the semi-automatic approach; in future prospective studies hard to evaluate studies will be included in order to assess the clinical impact of the method. Conclusions This study demonstrated that semi-automatic breast vascular map and predominant feeding vessel detection can identify correctly with a good reliability the lesion dominant vessel and its anatomic origin. Results of semi-automatic vascular map assessment showed a differential blood supply to breasts with malignancies according to the tumor location and an increased vascularity compared to breasts without malignant tumors: internal thoracic artery supplied the highest proportion of breasts with a tumor in internal half and lateral thoracic artery supplied the highest proportion of breasts with a lateral tumor. ==== Refs References 1 Peters NH , Borel Rinkes IH , Zuithoff NP , Mali WP , Moons KG , Peeters PH . Meta-analysis of MR imaging in the diagnosis of breast lesions . Radiology 2008 , 246 :116 –124 . 18024435 2 Schmitz AC , Peters NH , Veldhuis WB , Gallardo AM , van Diest PJ , Stapper G , et al Contrast-enhanced 3.0-T breast MRI for characterization of breast lesions: increased specificity by using vascular maps . Eur Radiol 2008 , 18 :355 –364 . 17882425 3 Wright H , Listinsky J , Quinn C , Rim A , Crowe J , Kim J . Increased ipsilateral whole breast vascularity as measured by contrast-enhanced magnetic resonance imaging in patients with breast cancer . Am J Surg 2005 , 190 :576 –579 16164924 4 Mahfouz AE , Sherif H , Saad A , Taupitz M , Filimonow S , Kivelitz D , et al Gadolinium-enhanced MR angiography of the breast: is breast cancer associated with ipsilateral higher vascularity? Eur Radiol 2001 , 11 :965 –969 . 11419172 5 Carriero A , Di Credico A , Mansour M , Bonomo L . Maximum intensity projection analysis in magnetic resonance of the breast . J Exp Clin Cancer Res 2002 , 21 :77 –81 . 12585659 6 Sardanelli F , Iozzelli A , Fausto A , Carriero A , Kirchin MA . Gadobenate dimeglumine-enhanced MR imaging breast vascular maps: association between invasive cancer and ipsilateral increased vascularity . Radiology 2005 , 235 :791 –797 . 15845796 7 Kul S , Cansu A , Alhan E , Dinc H , Reis A , Çan G . Contrast-enhanced MR angiography of the breast: Evaluation of ipsilateral increased vascularity and adjacent vessel sign in the characterization of breast lesions . AJR Am J Roentgenol . 2010 11 ;195 (5 ):1250 –4 . 10.2214/AJR.10.4368 20966336 8 Sardanelli F , Fausto A , Menicagli L , Esseridou A . Breast vascular mapping obtained with contrast-enhanced MR imaging: implications for cancer diagnosis, treatment, and risk stratification . Eur Radiol 2007 , 17 :F48 –51 . 18376457 9 Verardi N , Di Leo G , Carbonaro LA , Fedeli MP , Sardanelli F . Contrast-enhanced MR imaging of the breast: association between asymmetric increased breast vascularity and ipsilateral cancer in a consecutive series of 197 patients . Radiol Med 2012 , 118 :239 –250 . 10.1007/s11547-012-0864-0 22872456 10 Malich A , Fischer DR , Wurdinger S , Boettcher J , Marx C , Facius M , et al Potential MRI interpretation model: differentiation of benign from malignant breast masses . AJR Am J Roentgenol 2005 , 185 :964 –970 . 16177416 11 Fischer DR , Wurdinger S , Boettcher J , Malich A , Kaiser WA . Further signs in the evaluation of magnetic resonance mammography: a retrospective study . Invest Radiol 2005 , 40 :430 –435 . 15973134 12 Dietzel M , Baltzer PA , Vag T , Herzog A , Gajda M , Camara O , et al The adjacent vessel sign on breast MRI: new data and a subgroup analysis for 1,084 histologically verified cases . Korean J Radiol 2010 , 11 :178 –186 . 10.3348/kjr.2010.11.2.178 20191065 13 Grubstein A , Yepes M , Kiszonas R . Magnetic resonance imaging of breast vascularity in medial versus lateral breast cancer . Eur J Radiol 2010 , 75 :9 –11 . 14 Petrillo A , Petrillo M , Fulciniti F , Raso MM , Thomas R , Botti G , Siani A . Primary phyllodes tumor of the axilla: DCE-MRI findings with 1.5T breast-dedicated system and pathological correlation . Breast J . 2011 ,17 :525 –527 . 10.1111/j.1524-4741.2011.01121.x 21726349 15 Crispo A , Barba M , D'Aiuto G , De Laurentiis M , Grimaldi M , Rinaldo M , et al Molecular profiles of screen detected vs. symptomatic breast cancer and their impact on survival: results from a clinical series . BMC Cancer 2013 , 13 :15 10.1186/1471-2407-13-15 23305429 16 Fusco R , Sansone M , Filice S , Petrillo A . Breast contrast-enhanced MR imaging: semiautomatic detection of vascular map . Breast Cancer 2016 , 23 :266 –72 . 10.1007/s12282-014-0565-8 25239166 17 Frangi AF, Niessen WJ, Vinc KL, Viergever MA. Multiscale Vessel Enhancement Filtering, Medical Image Computing and Computer-Assisted Interventation—MICCAI’98 Lecture Notes in Computer Science 1998, 1496/1998: 130 18 Otsu Nobuyuki . A threshold selection method from gray-level histograms . IEEE Trans Sys Man Cyber 1979 , 9 (1 ):62 –66 . 19 Sardanelli F . Vessel Analysis on Contrast-Enhanced MRI of the Breast: Global or Local Vascularity? AJR Am J Roentgenol . 2010 ;195 : 1246 –1249 . 10.2214/AJR.10.4984 20966335 20 Cronbach LJ . Coefficient alpha and the internal structure of tests . Psychometrika 1951 , 16 : 297 –334 . 21 Lohrisch C , Jackson J , Jones A , Mates D , Olivotto IA . Relationship between tumor location and relapse in 6,781 women with early invasive breast cancer . J Clin Oncol , 2000 , 18 :2828 –2835 . 10920130 22 Krag D , Weaver D , Ashikaga T , Moffat F , Klimberg VS , Shriver C , et al The sentinel node in breast cancer—a multicenter validation study . N Engl J Med 1998 , 339 :941 –946 . 9753708 23 Orgüç Ş , Başara I , Coşkun T , Pekindil G . Three-dimensional vascular mapping of the breast by using contrast-enhanced MRI: association of unilateral increased vascularity with ipsilateral breast cancer . Diagn Interv Radiol . 2012 , 18 :454 –459 . 10.4261/1305-3825.DIR.5280-11.2 22865198 24 Vignati A , Giannini V , Bert A , Borrelli P , De Luca M , Martincich L , et al A fully automatic multiscale 3-dimensional Hessian-based algorithm for vessel detection in breast DCE-MRI . Invest Radiol 2012 , 47 :705 –710 . 10.1097/RLI.0b013e31826dc3a4 23070098 25 Vignati A , Giannini V , Carbonaro LA , Bertotto I , Martincich L , Sardanelli F , et al A new algorithm for automatic vascular mapping of DCE-MRI of the breast: Clinical application of a potential new biomarker . Comput Methods Programs Biomed 2014 , 117 :482 –488 . 10.1016/j.cmpb.2014.09.003 25262335 26 Lin M , Chen JH , Nie K , Chang D , Nalcioglu O , Su MY . Algorithm-based method for detection of blood vessels in breast MRI for development of computer-aided diagnosis . J Magn Reson Imaging . 2009 , 30 :817 –824 . 10.1002/jmri.21915 19787727

PMC005xxxxxx/PMC5003360.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757151110.1371/journal.pone.0161760PONE-D-16-04101Research ArticleBiology and Life SciencesAnatomyRenal SystemKidneysMedicine and Health SciencesAnatomyRenal SystemKidneysBiology and Life SciencesGeneticsGene ExpressionMedicine and Health SciencesVascular MedicineBlood PressureMedicine and Health SciencesSurgical and Invasive Medical ProceduresSurgical ExcisionNephrectomyMedicine and Health SciencesSurgical and Invasive Medical ProceduresUrinary System ProceduresNephrectomyBiology and Life SciencesAnatomyBody FluidsBloodBlood PlasmaMedicine and Health SciencesAnatomyBody FluidsBloodBlood PlasmaBiology and Life SciencesPhysiologyBody FluidsBloodBlood PlasmaMedicine and Health SciencesPhysiologyBody FluidsBloodBlood PlasmaMedicine and Health SciencesHematologyBloodBlood PlasmaBiology and Life SciencesDevelopmental BiologyFibrosisBiology and Life SciencesBiochemistryProteinsCollagensMedicine and health sciencesPharmacologyDrugsACE inhibitorsBiology and life sciencesBiochemistryEnzymologyEnzyme inhibitorsACE inhibitorsDiminazene Aceturate Improves Cardiac Fibrosis and Diastolic Dysfunction in Rats with Kidney Disease DIZE Has Cardioprotective Effects in Experimental Kidney DiseaseVelkoska Elena Patel Sheila K. Griggs Karen http://orcid.org/0000-0003-1863-7539Burrell Louise M. *Department of Medicine, The University of Melbourne, Austin Health, Heidelberg, Victoria, AustraliaJoles Jaap A. EditorUniversity Medical Center Utrecht, NETHERLANDSCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: EV LMB. Data curation: EV. Formal analysis: EV SKP. Funding acquisition: EV. Investigation: EV KG. Methodology: EV KG LMB. Project administration: EV. Supervision: LMB. Writing – original draft: EV LMB. Writing – review & editing: EV SKP LMB. * E-mail: l.burrell@unimelb.edu.au29 8 2016 2016 11 8 e016176029 1 2016 11 8 2016 © 2016 Velkoska et al2016Velkoska et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Angiotensin converting enzyme (ACE) 2 is a negative regulator of the renin angiotensin system (RAS) through its role to degrade angiotensin II. In rats with subtotal nephrectomy (STNx), adverse cardiac remodelling occurs despite elevated cardiac ACE2 activity. We hypothesised that diminazene aceturate (DIZE), which has been described as having an off-target effect to activate ACE2, would have beneficial cardiac effects in STNx rats. STNx led to hypertension, diastolic dysfunction, left ventricular hypertrophy, cardiac fibrosis, and increased cardiac ACE, ACE2, Ang II and Ang 1–7 levels. Cardiac gene expression of ADAM17 was also increased. In STNx, two-weeks of subcutaneous DIZE (15mg/kg/d) had no effect on blood pressure but improved diastolic dysfunction and cardiac fibrosis, reduced ADAM17 mRNA and shifted the cardiac RAS balance to a cardioprotective profile with reduced ACE and Ang II. There was no change in cardiac ACE2 activity or in cardiac Ang 1–7 levels with DIZE. In conclusion, our results suggest that DIZE exerts a protective effect on the heart under the pathological condition of kidney injury. This effect was not due to improved kidney function, a fall in blood pressure or a reduction in LVH but was associated with a reduction in cardiac ACE and cardiac Ang II levels. As in vitro studies showed no direct effect of DIZE on ACE2 or ACE activity, the precise mechanism of action of DIZE remains to be determined. http://dx.doi.org/10.13039/501100000925National Health and Medical Research CouncilGNT1048285Velkoska Elena This work was supported by funding granted to EV from National Health and Medical Research Council of Australia (APP1048285). Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction The prevalence of kidney disease is increasing worldwide and cardiovascular disease is the major cause of morbidity and mortality [1, 2]. Activation of the renin angiotensin system (RAS) plays an important role in the development and progression of kidney and heart disease. In the classic RAS, angiotensin converting enzyme (ACE) converts angiotensin (Ang) I to Ang II, which mediates its adverse effects via the angiotensin type 1 receptor [3]. RAS blockade is recommended as first-line therapy in patients with kidney disease to improve outcomes [4], but as kidney and cardiac disease continue to progress, new therapeutic approaches are needed. In the alternate arm of the RAS [5, 6], ACE2 counterbalances an activated ACE/Ang II pathway through degradation of Ang II, and generation of the antifibrotic peptide, Ang 1–7 [7, 8]. Increasing evidence suggests that the relative balance between the deleterious ACE/Ang II pathway and the protective ACE2/Ang 1–7 pathway is an important determinant of tissue injury. In kidney disease secondary to subtotal nephrectomy (STNx), kidney ACE and Ang II are increased [9–13] and kidney ACE2 activity is decreased [12, 14]. These findings have led to strategies to replenish ACE2 using recombinant ACE2 or to activate ACE2 using compounds such as diminazene aceturate (DIZE) [15]. Both approaches have beneficial effects on the kidney in experimental kidney disease. Recombinant human ACE2 prevented Ang II induced kidney disease and tubulointerstitial fibrosis [16] and slowed the progression of diabetic nephropathy through a reduction in renal Ang II and increased Ang 1–7 levels [17]. DIZE had reno-protective effects after renal ischemia/reperfusion injury through enhancement of antioxidant activity [18], and we reported that short-term DIZE decreased kidney cortical ACE activity and ameliorated the reduction in kidney ACE2 expression and activity in STNx rats [13]. Kidney disease is also associated with activation of the cardiac RAS and cardiac damage. We previously reported that STNx rats have increased cardiac ACE activity and adverse cardiac remodelling despite elevations in cardiac ACE2 activity [19, 20]. We also showed that ramipril lowered blood pressure and inhibited cardiac ACE activity, and that these changes were associated with a reduction in LVH and cardiac ACE2 activity [19]. In the current paper, we hypothesised that DIZE which has been described as having an off-target effect to activate ACE2, would have beneficial cardiac effects in STNx rats. Full cardiac hemodynamic data was available in a subset of rats from our previously published study [13] along with cardiac tissue. Therefore, the focus of this paper was to investigate the effect of short-term DIZE on cardiac structure and function, cardiac ACE/ACE2 gene expression/activity and cardiac angiotensin peptide (Ang II/Ang 1–7) levels in STNx rats with an activated RAS and in Control rats with a balanced RAS. In addition, we assessed gene expression of cardiac ADAM17 [21], a proteinase that is responsible for the cleavage or “shedding” of the catalytically active ectodomain of ACE2 from the cell membrane. Materials and Methods Experimental Protocol Experimental procedures were performed in accordance with the National Health and Medical Research Council of Australia guidelines for animal experimentation and were approved by the Animal Ethics Committee, Austin Health. Female Sprague Dawley (SD) rats (body weight of 190-210g) were housed in a 12:12h light-dark cycle, with ad libitum food containing 0.4–0.6% NaCl (Norco) and water. STNx (n = 16) was performed as described previously [12, 13, 19, 20], with a right nephrectomy, and ligation of all but one of the extra-renal branches of the left renal artery. STNx rats were randomly allocated to the presumed ACE2 activator DIZE (2 weeks s.c. 15mg/kg/day, n = 8) via osmotic minipump (Model # 2002, Alzet, Cupertino, CA, USA) or Vehicle (n = 8), which was implanted at the time of STNx surgery. We used the same dose and mode of delivery of DIZE as in previously published studies [13, 22, 23]. Control rats received DIZE (2 weeks s.c. 15mg/kg/day, n = 8) or Vehicle (n = 8). On day 14, rats were anaesthetised with intraperitoneal (i.p.) sodium pentobarbitone (60 mg/kg/body weight), and cardiac haemodynamics were determined using a micro-tipped pressure transducer catheter (Millar, 1.5F) inserted into the left carotid artery and advanced into the left ventricle (LV) as described previously [14, 20]. Data were stored and analysed using Millar conductance data acquisition and analysis software, and heart rate, systolic blood pressure and maximal and minimal rate of ventricular contraction (±dP/dt) were determined [14, 20]. We also measured the time constant of isovolumic relaxation (Tau), which measures active relaxation, with higher values of Tau implying impaired relaxation [24]. Rats were killed by a lethal dose of sodium pentobarbitone, decapitated and trunk blood collected into either lithium heparin tubes or into EDTA tubes containing 20 μl/ml of blood of an enzyme inhibitor cocktail [50 mM EDTA, 0.2 M N-ethylmaleimide and 1–2 TIU (trypsin inhibitory units)/ml aprotinin made up in saline] [20]. After centrifugation, plasma was separated, snap-frozen and stored at −80°C. The heart was removed, weighed and the LV was transversely dissected into 3 pieces, with one piece fixed in 4% paraformaldehyde and embedded in paraffin for histopathology. The remainder of the LV was snap-frozen in isopentane and stored at -80°C for mRNA extraction, peptide analysis and activity assays. Drugs Sodium pentobarbitone was obtained from Boehringer Ingelheim, Artarmon, NSW, Australia), DIZE from Sigma-Aldrich Australia. Biochemical analysis Plasma creatinine (Cr) was measured using an autoanalyser (Beckman Instruments, Palo Alta, CA, USA). Determination of cardiac collagen Cardiac (LV) paraffin sections 4μm thick were deparaffinized, rehydrated and stained with 0.1% Sirius Red (Polysciences Inc) in saturated picric acid (picrosirius red) for 1 hour, differentiated in 0.01% HCl for 30 seconds, and rapidly dehydrated. Interstitial collagen volume fraction was determined by measuring the area of stained tissue within a given field, excluding vessels, artefacts, minor scars or incomplete tissue fields; 15–20 fields were analysed per animal in a blinded manner [14, 20]. To measure perivascular collagen, all arteries in the LV section were analysed, and the whole artery including the adventitia was selected for assessment. For both interstitial and perivascular collagen, the area stained was calculated as a percentage of the total area within a given field [25, 26]. Plasma and cardiac angiotensin peptides Blood for the measurement of angiotensin peptides were collected and stored as described above. Frozen LV sections were homogenised in 5mM EDTA solution containing a protease inhibitor (P8340, Sigma Aldrich; components: 104 mM AEBSF [4-(2-Aminoethyl) benzenesulfonyl fluoride hydrochloride], 80μM aprotinin, 4mM bestatin, 1.4mM E-64, 2mM leupeptin and 1.5mM pepstatin A), frozen at -80°C to aid in cell disruption, thawed and centrifuged prior to assaying (Prosearch, Australia). Angiotensin peptide content is expressed per mg of protein content of each individual sample. The radioimmunoassays for Ang II and Ang 1–7 have been described previously [12, 20]. The antibodies used for Ang II and Ang 1–7 were raised in rabbit and guinea pig respectively, and the specific radioisotopes, 125I-Ang II and 125I-Ang 1–7, were made by Prosearch (Melbourne, Australia). The intra- and inter-assay coefficients of variation were 7.6 and 8.3% for Ang II and 4.5 and 10% for Ang 1–7. In vivo cardiac ACE and ACE2 activity LV membrane preparations were prepared as previously described [19], and ACE activity measured using an enzymatic assay [20]. Briefly, 100μg of membrane protein was incubated at 37°C with the ACE substrate hippuryl-His-Leu (1mM) in a total volume of 50μl in the presence and absence of EDTA (10μM) for 60 min. The rate of substrate cleavage was determined by comparison to a standard curve of the product His-Leu and expressed as nmole of substrate cleaved/mg of protein/hr. LV ACE2 was measured using an enzymatic assay as previously described [19, 20]. Briefly, 100μg of membrane protein was incubated in duplicate with an ACE2 quenched fluorescent substrate (QFS), (7-methoxycoumarin-4-yl)-acetyl-Ala-Pro-Lys (2, 4-dintirophenyl); Auspep, Parkville, Victoria, Australia), as previously described [19, 20, 27]. The specific activity was determined using 100μM EDTA which has a similar effect to inhibit ACE2 as the specific ACE2 inhibitor, MLN4760 [28, 29]. In addition, as QFS can be cleaved by prolyl endopeptidase, an inhibitor of this enzyme, Z-Pro-prolinal (1μM) was included in all wells [30]. The rate of substrate cleavage was determined by comparison to a standard curve of the free fluorophore, 4-amino-methoxycoumarin (MCA; Sigma, MO, USA) and expressed as nmole of substrate cleaved/mg of protein/hr. For plasma ACE and ACE2 activity, blood collected into heparinised tubes was centrifuged at 4°C and assayed as above. Results are expressed as nmole of substrate/ml of plasma/hr. In vitro effect of DIZE on cardiac (LV) ACE2 and ACE activity and ACE binding Cardiac (LV) membranes from STNx (n = 4) and Control (n = 4) rats were incubated with increasing concentrations of DIZE (0.1mM, 0.1μM, 0.1nM) or vehicle. ACE and ACE2 activity was measured as described above and results expressed as nmole of substrate cleaved/mg of protein/hr. The effect of DIZE on cardiac ACE binding was also assessed using ex vivo autoradiography on LV sections (20μm) using the specific ACE inhibitor radioligand 125I-MK351A (Ki = 30 pmol/l) [14, 19, 20]. Sections were incubated with serial dilutions of DIZE or the ACE inhibitor ramiprilat (10−10 to 10−3 mol/l) (n = 4 per concentration). Quantification of ACE binding density was performed using a microcomputer-imaging device (MCID, Imaging Research, UK) which measures the relative absorbance of the radioactive labelling. Cardiac (LV) ACE2 immunohistochemistry Immunohistochemical staining for ACE2 (polyclonal antibody, T17, from Santa Cruz Biotechnology, Santa Cruz, CA, USA; diluted 1:100) was performed in rat LV sections as described previously [14, 19, 25]. Staining was quantified using computerized image analysis (Imaging Research, Linton, Cambridge, UK). All sections used for quantification were fixed, processed, sectioned and immunolabelled at the same time and under the same conditions to limit variability. Images were imported into the AIS Imaging program using a colour video camera and a standard light microscope (magnification ×20). The detection level threshold for positively stained areas (brown for 3,3-Diaminobenzidine staining) was set so that the processed image accurately reflected the positively stained areas as visualized by light microscopy on the unprocessed digital image. Myocyte ACE2 staining was determined by measuring the area of stained tissue within a given field, excluding vessels, artifacts, minor scars or incomplete tissue fields. Fifteen to 20 fields were analysed per animal. The percentage area of chromogen staining was determined by calculating the number of selected pixels (positively stained areas) as a percentage of the total area within a given field [26]. Cardiac (LV) gene expression Total RNA was isolated from the LV using the RNeasy fibrous kit method (Qiagen). cDNA was synthesized with a reverse transcriptase reaction using standard techniques (Superscript III kit; Life Technologies) as described previously [19, 20]. The primers and probes for ACE, ACE2 and brain natriuretic peptide (BNP) were designed using the software program Primer Express (PE Applied Biosystems). Collagen 1A (Col1A) and ADAM17 TaqMan gene expression assay was purchased from Applied Biosystems. Quantitative RT-PCR was performed using the TaqMan system based on real-time detection of accumulated fluorescence (7500 Real Time PCR System, Applied Biosystems, CA, USA). Gene expression was normalized to 18S VIC and reported as ratios compared with the level of expression in Control rats, which were given an arbitrary value of 1. Statistical Analysis Data are presented as mean ± standard error of mean (SEM). P values were calculated using a two-way analysis of variance (ANOVA), followed by post hoc Bonferroni tests (GraphPad Prism 6) comparing effect of STNx (Control vs. STNx) and effect of treatment (Control vs Control+DIZE; STNx vs. STNx+DIZE). For data with unequal variance, results were log-transformed and analysed using Kruskal-Wallis test with Dunn’s multiple comparison test. The Pearson correlation coefficient was determined for the associations between variables using the data from untreated Control and STNx rats. Two-tailed P-values <0.05 were considered significant. Results Physiological, biochemical and plasma RAS parameters after STNx Following STNx, rats had reduced body weight (P<0.01) and kidney impairment with elevated plasma Cr (P<0.001) compared to Controls (Table 1). Plasma ACE2 activity, Ang II and Ang 1–7 were increased in STNx rats (P<0.05), with no change in plasma ACE activity compared to Control rats. Treatment with DIZE did not alter circulating RAS components or improve kidney function in Control or STNx rats. 10.1371/journal.pone.0161760.t001Table 1 Renal function and plasma RAS components. Control Subtotal nephrectomy Vehicle DIZE Vehicle DIZE (15 mg/kg/day) (15 mg/kg/day) (n = 8) (n = 8) (n = 8) (n = 8) Body weight (g) 239 ± 4 232 ± 3 214 ± 8** 210 ± 10 Plasma creatinine (μmol/L) 17 ± 1 16 ± 1 46 ± 5*** 40 ± 2 Plasma RAS Components ACE activity (nmol/ml/hr) 1714 ± 67 1529 ± 116 1589 ± 98 1582 ± 143 ACE2 activity (nmol/ml/hr) 5.6 ± 0.6 5.3 ± 0.2 7.1 ± 0.4* 6.5 ± 0.4 Ang II (fmol/ml) 28.4 ± 3.4 29.6 ± 2.9 70.9 ± 18.4* 60.6 ± 11.2 Ang 1–7 (fmol/ml) 183 ± 16 196 ± 24 327 ± 50* 456 ± 72 Data expressed as mean ± SEM. *P<0.05 **P<0.01 ***P<0.001 disease effect (Control Vehicle vs. STNx Vehicle). Blood pressure and cardiac function and structure STNx rats had increased systolic blood pressure (Fig 1A, P<0.001) and cardiac hypercontractility (Fig 1B, P<0.05). Diastolic dysfunction was present with impaired active relaxation shown by a reduction in min dP/dt (Fig 1C, P<0.05) and an increased time constant for isovolumic relaxation (Tau) (Fig 1D, P<0.05). In STNx, DIZE had no effect on blood pressure or contractility but improved diastolic function compared to vehicle treated rats (Fig 1C and 1D, P<0.05). STNx was associated with LV hypertrophy (LVH, Fig 2A, P<0.001) and significant interstitial (Fig 2B, P<0.001) and perivascular fibrosis (Fig 2C, P<0.01). DIZE reduced both interstitial and perivascular fibrosis (P<0.05), with no effect on LVH. DIZE had no effect on blood pressure, cardiac structure or function in Control rats. 10.1371/journal.pone.0161760.g001Fig 1 DIZE improved diastolic dysfunction in STNx. Systolic blood pressure (A), ventricular contractility (B), ventricular relaxation (C) and time constant of isovolumic relaxation (Tau; D) in Control (Cont) and subtotal nephrectomy (STNx) rats (n = 8/group). Data expressed as mean±SEM. *P<0.05, ***P<0.001 disease effect (Control vs. STNx) and # P<0.05 treatment effect (Vehicle vs. DIZE) 10.1371/journal.pone.0161760.g002Fig 2 DIZE was associated with a reduction in cardiac fibrosis in STNx. Left ventricular hypertrophy (A), interstitial collagen (B) and perivascular collagen (C) in Control (Cont) and subtotal nephrectomy (STNx) rats (n = 8/group). Right hand panel consists of representative photomicrographs of left ventricular total collagen content (red staining) (magnification x200). Data expressed as mean±SEM. **P<0.01, ***P<0.001 disease effect (Control vs. STNx) and # P<0.05 treatment effect (Vehicle vs. DIZE) Cardiac (LV) gene expression Cardiac ACE and ACE2 gene expression were increased in STNx rats compared to Control (both P<0.05), and were unchanged with DIZE (Table 2). Cardiac ADAM17 gene expression was also elevated in STNx rats and reduced with DIZE (Table 2). BNP mRNA expression, a marker of cardiac damage was increased in the LV of STNx rats (P<0.001) and reduced by DIZE (P<0.05). Col1A (P<0.00) gene expression was increased in the LV of STNx rats, and reduced by DIZE (P<0.05) in keeping with the reduction in collagen protein. In Control rats, DIZE had no effect on cardiac ACE, ACE2, BNP or Col1A gene expression. 10.1371/journal.pone.0161760.t002Table 2 Cardiac (LV) gene expression. Control Subtotal nephrectomy Vehicle DIZE Vehicle DIZE (15 mg/kg/day) (15 mg/kg/day) (n = 8) (n = 8) (n = 8) (n = 8) ACE (arbitrary units) 1.00 ± 0.22 1.34 ± 0.24 1.99 ± 0.32* 1.69 ± 0.33 ACE2 (arbitrary units) 1.00 ± 0.13 1.16 ± 0.18 1.46 ± 0.17* 1.41 ± 0.12 BNP (arbitrary units) 1.00 ± 0.22 0.97 ± 0.19 4.13 ± 0.56*** 2.56 ± 0.41# Col1A (arbitrary units) 1.00 ± 0.23 1.48 ± 0.24 3.92 ± 0.54*** 2.52 ± 0.37# ADAM17 (arbitrary units) 1.00 ± 0.16 1.12 ± 0.14 1.80 ± 0.27* 1.13 ± 0.09# Data expressed as mean±SEM. *P<0.05 ***P<0.001 disease effect (Control Vehicle vs. STNx Vehicle) #P<0.05 treatment effect (Vehicle vs. DIZE) Cardiac (LV) RAS components Cardiac ACE activity (Fig 3A) and ACE2 activity (Fig 3B) were increased in STNx rats compared to Control rats (both, P<0.05). Cardiac ACE2 protein was also elevated in STNx rat and unchanged with DIZE treatment (S1 Fig). In STNx, DIZE treatment was associated with a reduction in cardiac ACE activity (P<0.05), but no change in cardiac ACE2 activity compared to the untreated STNx group. This resulted in a decrease in the ACE/ACE2 activity ratio towards a more favourable and cardioprotective profile (Fig 3C, P<0.05). In Control rats, DIZE had no effect on cardiac ACE or ACE2 protein or activity. 10.1371/journal.pone.0161760.g003Fig 3 DIZE shifts the cardiac RAS balance to a cardioprotective profile in STNx. Left ventricular (LV) ACE (A) and ACE2 activity (B), ACE/ACE2 activity ratio (C), LV Ang II (D) and Ang 1–7 (E) peptide content in Control (Cont) and subtotal nephrectomy (STNx) rats (n = 8/group). Data expressed as mean ± SEM. *P<0.05 disease effect (Control vs. STNx) and # P<0.05 treatment effect (Vehicle vs. DIZE) Cardiac Ang II (Fig 3D; P<0.01) and Ang 1–7 (Fig 3E, P<0.05) were increased in STNx rats compared to Control. In STNx, DIZE treatment reduced cardiac Ang II levels (P<0.05) and Ang 1–7 levels remained unchanged. DIZE had no effect on cardiac Ang peptides in Control rats. Circulating vs. cardiac tissue ACE and ACE2 activity In vehicle-treated STNx rats increased cardiac ACE2 activity correlated with increased cardiac ACE activity (Fig 4A), which supports the notion that in disease ACE2 increases to balance the effects of elevated cardiac ACE. The correlation between increased cardiac ACE2 and increased plasma ACE2 activity (Fig 4B) taken together with upregulation of cardiac ADAM17 mRNA suggests that increased ACE2 shedding from the diseased heart may be responsible for the increase in circulating levels of ACE2. By contrast, there was no correlation between LV ACE and plasma ACE activity. 10.1371/journal.pone.0161760.g004Fig 4 Cardiac ACE2 activity is increased to counteract elevated cardiac ACE and is shed into the circulation. Correlation analysis shows increased left ventricular (LV) ACE2 activity is associated with increased LV ACE activity (A) and plasma ACE2 activity (B), with no correlation between LV tissue and plasma ACE. Control and subtotal nephrectomy (STNx) rats without active treatment were used for the correlation analysis. Open squares represent Control rats; closed squares represent STNx rats. In vitro effect of DIZE on cardiac ACE2 activity, and ACE activity and binding The ex vivo effect of DIZE (0.1mM, 0.1μM, 0.1nM) on LV ACE2 and ACE activity was measured in Control and STNx rats. As shown in Fig 5, DIZE had no effect on cardiac ACE2 (Fig 5A) or ACE activity (Fig 5B) in either Control or STNx rats. The ex vivo effect of DIZE and ACE inhibition with ramiprilat on cardiac ACE was also examined; ramiprilat inhibited cardiac ACE but DIZE had no effect on ACE binding (Fig 5C). 10.1371/journal.pone.0161760.g005Fig 5 DIZE does no effect cardiac ACE and ACE2 under ex vivo conditions. Ex vivo DIZE had no effect on endogenous ACE2 (A) and ACE (B) activity in LV membrane preps (100μg per well) from Control (n = 4) and subtotal nephrectomy (STNx, n = 4) rats. The ACE inhibitor, ramiprilat caused a concentration-dependent displacement of specific 125I-MK351A binding from rat LV ACE and DIZE had no effect on ACE binding (C). Discussion ACE2 is highly expressed in the heart and is an important regulator of cardiac function [31, 32]. We explored the role of ACE2 in the cardiac consequences of kidney disease using a model of kidney injury due to STNx. The results confirm our previous reports that STNx leads to LVH, impaired cardiac function and elevations in cardiac ACE and ACE2 activity [19, 20]. We also report that STNx rats had increased cardiac Ang II and Ang 1–7 levels and increased cardiac gene expression of ADAM17. The results of this study were that a 2 week s.c. infusion of DIZE significantly improved diastolic function and cardiac fibrosis in STNx rats, and these benefits were achieved in the absence of a fall in blood pressure or any improvement in kidney function. DIZE reduced the gene expression of BNP, an indirect marker of cardiac injury in STNx rats, and shifted the cardiac RAS balance to a more cardioprotective profile with a reduction in both cardiac ACE and cardiac Ang II. There was no change in cardiac ACE2 activity or in cardiac Ang 1–7 levels with DIZE. The prevailing Ang II levels represent a balance between formation (due to ACE), and degradation (due to ACE2). We found no evidence that DIZE increased cardiac Ang 1–7 levels, suggesting that the major benefit of DIZE is due to reduced Ang II formation. It is unclear how DIZE is mediating its effects in STNx rats. The in vitro studies showed no direct effect of DIZE on ACE2 or ACE activity. In vivo, the effect of DIZE on cardiac ACE2 contrasts with our previous results where the cardiac benefits of ACE inhibition were associated with a reduction in cardiac ACE and ACE2 [13]. The finding that DIZE was associated with down regulation of ADAM17 mRNA led us to speculate that DIZE may have indirect effects to maintain cardiac ACE2 activity levels and thus reduce cardiac Ang II, through reduced cleavage of ACE2 from cardiac cells. This hypothesis would be consistent with reports that cardiomyocyte-specific deletion of ADAM17 prevented shedding of ACE2, whilst exogenous infusion of Ang II increased myocardial ADAM17 expression and decreased myocardial ACE2 activity [33]. Future studies should address not only ADAM17 gene expression but also ADAM17 protein content. Our results also show that the relative balance or imbalance of the ACE/Ang II and the ACE2/Ang 1–7 pathway may determine the in vivo effect of DIZE on the underlying pathophysiology. Thus, by contrast to its effects in the pathological state of kidney failure, DIZE had no effect on cardiac ACE/ACE2 activity, cardiac Ang peptides or ADAM17 in Control rats with a balanced RAS and normal blood pressure and kidney function. We have previously shown that short-term DIZE significantly reduced kidney cortical ACE activity and ameliorated the reduction in cortical and medullary ACE2 activity in STNx rats with kidney RAS imbalance, but had no effect on kidney ACE2 expression or activity in Control rats with a balanced RAS [13]. Further support for this concept comes from studies of DIZE after myocardial infarction (MI) which is also associated with imbalance of the cardiac RAS [22]. A 4-week s.c infusion of DIZE in MI in rats improved cardiac remodelling, significantly increased cardiac ACE2 mRNA and activity and reduced cardiac ACE mRNA and activity–these effects were blocked by concurrent use of the ACE2 inhibitor, C16 [22]. In hypercholesterolemic mice on a wild-type or ACE2 deficient background with Ang II induced abdominal aortic aneurysms [34], 4-weeks of intramuscular DIZE (30mg/kg) increased kidney ACE2 mRNA and activity in wild-type mice, and reduced the incidence and severity of Ang II-induced abdominal aortic aneurysms. However DIZE had no effect in ACE2-deficient mice which does suggest that it acts through an ACE2-dependent mechanism [34]. The precise mechanism of action of DIZE remains unclear [35] and there is conflicting evidence with regard its effect on ACE2 activity [13, 15, 23, 35, 36]. Kulemina et al first reported the off target effects of DIZE to activate ACE2 [15] and described a biphasic dose–response curve; DIZE activated ACE2 at low concentrations and partially inhibited ACE2 at high concentrations. Incubation of human rACE2 with DIZE (100μM) increased ACE2 activity in some studies [23], but others have shown that DIZE did not increase the enzymatic activity of mouse or human rACE2 [35]. We reported that DIZE had no effect ex vivo to alter kidney ACE2 activity, but was associated with increased kidney ACE2 activity in vivo in STNx rats [13]. The ex vivo studies in cardiac membranes of STNx and control rats in the current study showed no direct effect of DIZE to increase ACE2 activity. Taken together, the reported effects of DIZE on ACE2 activity vary according to whether the study was in vivo or ex vivo, and may depend on the underlying pathophysiology, route of administration or the tissue used to assess ACE2 activity. In order to delineate if the mechanism of action of DIZE is specifically related to ACE2 activation, future studies should examine the effect of co-administration of DIZE with specific ACE2 inhibitors such as MLN 4760 (C16) to help establish or exclude if the observed beneficial cardiac effects of DIZE are truly ACE2 dependent. With regard to ACE, the reduction in cardiac ACE activity in STNx with DIZE is likely to be an indirect effect due to reduced cardiac damage and thus less ACE activation. The degree of ACE “inhibition” with DIZE is less than that would be expected with an ACE inhibitor [14, 19], and ACE inhibition usually results in a fall in blood pressure. The ex vivo studies in cardiac membranes showed no direct effect of DIZE to inhibit ACE, which is consistent with the work of others that DIZE had no effect on the catalytic activity of ACE [15]. In summary, short-term intervention with the presumed ACE2 activator DIZE ameliorates cardiac dysfunction and cardiac fibrosis in STNx rats. The benefits of DIZE were associated with a shift in the cardiac RAS balance to a cardioprotective profile with reduced ACE and Ang II and no change in cardiac ACE2 activity. Cardiac Ang 1–7 levels were not changed with DIZE suggesting that its main effects are related to reduced Ang II. We have previously reported in STNx, that improved cardiac structure and function was associated with reduced cardiac ACE2 activity, and the finding that cardiac ACE2 remained unchanged with DIZE may reflect reduced cardiac ACE2 shedding secondary to downregulation of cardiac ADAM17. Further studies are now warranted to determine if the observed beneficial cardiac effects of DIZE are due to reduced ACE2 shedding from the cell membrane, or are specifically related to ACE2 activation. Studies are also needed to assess the long-term impact of DIZE on both cardiac and kidney function, and to investigate if combining DIZE with standard RAS blockade has incremental effects to improve the cardiac consequences of kidney disease or to prevent progression to chronic kidney disease. Although DIZE is used for the treatment of trypanosomiasis or sleeping sickness, side effects may preclude its widespread clinical use [37–39] and specific compounds that selectively amplify ACE2 activity will be needed if this approach is to find utility. Supporting Information S1 Fig ACE2 protein expression increases in STNx. Left ventricular (LV) ACE2 protein expression in Control (Cont) and subtotal nephrectomy (STNx) rats (n = 8/group). Right hand panel consists of representative photomicrographs of ACE2 immunohistochemical labelling (brown staining) (magnification x200). Data expressed as mean ± SEM. *P<0.05 disease effect (Control vs. STNx) (TIFF) Click here for additional data file. ==== Refs References 1 Levey AS , de Jong PE , Coresh J , El Nahas M , Astor BC , Matsushita K , et al The definition, classification, and prognosis of chronic kidney disease: a KDIGO Controversies Conference report . Kidney Int . 2011 ;80 (1 ):17 –28 . Epub 2010/12/15. 10.1038/ki.2010.483 .21150873 2 Schiffrin EL , Lipman ML , Mann JF . Chronic kidney disease: effects on the cardiovascular system . Circulation . 2007 ;116 (1 ):85 –97 . .17606856 3 Candido R , Burrell LM , Jandeleit-Dahm K , Cooper ME . Vasoactive peptides and the kidney In: Brenner BM , Rector FC , editors. Brenner & Rector's The Kidney . 8 ed. Philadelphia : Saunders ; 2008 p. 333 –62 . 4 Stevens PE , Levin A . Evaluation and management of chronic kidney disease: synopsis of the kidney disease: improving global outcomes 2012 clinical practice guideline . Ann Intern Med . 2013 ;158 (11 ):825 –30 . Epub 2013/06/05. 1691737 [pii] 10.7326/0003-4819-158-11-201306040-00007 .23732715 5 Donoghue M , Hsieh F , Baronas E , Godbout K , Gosselin M , Stagliano N , et al A novel angiotensin-converting enzyme-related carboxypeptidase (ACE2) converts angiotensin I to angiotensin 1–9 . Circ Res . 2000 ;87 (5 ):E1 –9 . .10969042 6 Tipnis SR , Hooper NM , Hyde R , Karran E , Christie G , Turner AJ . A human homolog of angiotensin-converting enzyme. Cloning and functional expression as a captopril-insensitive carboxypeptidase . J Biol Chem . 2000 ;275 (43 ):33238 –43 . .10924499 7 Grobe JL , Mecca AP , Lingis M , Shenoy V , Bolton TA , Machado JM , et al Prevention of angiotensin II-induced cardiac remodeling by angiotensin-(1–7) . Am J Physiol Heart Circ Physiol . 2007 ;292 (2 ):H736 –42 . .17098828 8 Grobe JL , Mecca AP , Mao H , Katovich MJ . Chronic angiotensin-(1–7) prevents cardiac fibrosis in DOCA-salt model of hypertension . Am J Physiol Heart Circ Physiol . 2006 ;290 (6 ):H2417 –23 . .16415071 9 Cao Z , Burrell LM , Tikkanen I , Bonnet F , Cooper ME , Gilbert RE . Vasopeptidase inhibition attenuates the progression of renal injury in subtotal nephrectomized rats . Kidney Int . 2001 ;60 (2 ):715 –21 . .11473654 10 Cao Z , Cooper ME . Role of angiotensin II in tubulointerstitial injury . Semin Nephrol . 2001 ;21 (6 ):554 –62 . .11709803 11 Taal MW , Brenner BM . Adaptation to nephron loss In: Brenner BM , Rector FC , editors. Brenner and Rector's The Kidney . 8 Philadelphia : Saunders ; 2008 p. 783 –819 . 12 Velkoska E , Dean RG , Burchill LJ , Levidiotis V , Burrell LM . Reduction in renal ACE2 expression in subtotal nephrectomy is ameliorated with ACE inhibition . Clin Sci (Lond) . 2010 ;118 (4 ):269 –79 . Epub 2009/08/25. CS20090318 [pii] 10.1042/CS20090318 .19698082 13 Velkoska E , Patel SK , Griggs K , Pickering RJ , Tikellis C , Burrell LM . Short-term treatment with diminazene aceturate ameliorates the reduction in kidney ACE2 activity in rats with subtotal nephrectomy . PLoS One . 2015 ;10 (3 ):e0118758 Epub 2015/03/19. 10.1371/journal.pone.0118758 .25786223 14 Burrell LM , Burchill L , Dean RG , Griggs K , Patel SK , Velkoska E . Chronic kidney disease: cardiac and renal angiotensin converting enzyme (ACE) 2 expression in rats after subtotal nephrectomy and the effect of ACE inhibition . Exp Physiol . 2012 ;97 (4 ):477 –85 . Epub 2011/12/27. 10.1113/expphysiol.2011.063156 .22198016 15 Kulemina LV , Ostrov DA . Prediction of off-target effects on angiotensin-converting enzyme 2 . J Biomol Screen . 2011 ;16 (8 ):878 –85 . Epub 2011/08/24. 1087057111413919 [pii] 10.1177/1087057111413919 .21859683 16 Zhong J , Guo D , Chen CB , Wang W , Schuster M , Loibner H , et al Prevention of angiotensin II-mediated renal oxidative stress, inflammation, and fibrosis by angiotensin-converting enzyme 2 . Hypertension . 2011 ;57 (2 ):314 –22 . Epub 2010/12/30. [pii] 10.1161/HYPERTENSIONAHA.110.164244 .21189404 17 Oudit GY , Liu GC , Zhong J , Basu R , Chow FL , Zhou J , et al Human recombinant ACE2 reduces the progression of diabetic nephropathy . Diabetes . 2010 ;59 (2 ):529 –38 . 10.2337/db09-1218 19934006 18 Malek M , Nematbakhsh M . The preventive effects of diminazene aceturate in renal ischemia/reperfusion injury in male and female rats . Adv Prev Med . 2014 ;2014 :740647 Epub 2014/12/06. 10.1155/2014/740647 25478235 19 Burchill L , Velkoska E , Dean RG , Lew RA , Smith AI , Levidiotis V , et al Acute kidney injury in the rat causes cardiac remodelling and increases angiotensin-converting enzyme 2 expression . Exp Physiol . 2008 ;93 (5 ):622 –30 . 10.1113/expphysiol.2007.040386 .18223026 20 Velkoska E , Dean RG , Griggs K , Burchill LJ , Burrell LM . Angiotensin 1–7 infusion is associated with increased blood pressure and adverse cardiac remodelling in rats with subtotal nephrectomy . Clin Sci (Lond) . 2011 ;120 (8 ):335 –45 . Epub 2010/11/26. CS20100280 [pii] 10.1042/CS20100280 .21091432 21 Lambert DW , Yarski M , Warner FJ , Thornhill P , Parkin ET , Smith AI , et al Tumor necrosis factor-alpha convertase (ADAM17) mediates regulated ectodomain shedding of the severe-acute respiratory syndrome-coronavirus (SARS-CoV) receptor, angiotensin-converting enzyme-2 (ACE2) . J Biol Chem . 2005 ;280 (34 ):30113 –9 . .15983030 22 Qi Y , Zhang J , Cole-Jeffrey CT , Shenoy V , Espejo A , Hanna M , et al Diminazene aceturate enhances angiotensin-converting enzyme 2 activity and attenuates ischemia-induced cardiac pathophysiology . Hypertension . 2013 ;62 (4 ):746 –52 . Epub 2013/08/21. [pii] 10.1161/HYPERTENSIONAHA.113.01337 .23959549 23 Shenoy V , Gjymishka A , Yagna J , Qi Y , Afzal A , Rigatto K , et al Diminazene attenuates pulmonary hypertension and improves angiogenic progenitor cell functions in experimental models . Am J Respir Crit Care Med . 2013 ;187 (6 ):648 –57 . Epub 2013/02/02. [pii] 10.1164/rccm.201205-0880OC .23370913 24 Kennedy DJ , Elkareh J , Shidyak A , Shapiro AP , Smaili S , Mutgi K , et al Partial nephrectomy as a model for uremic cardiomyopathy in the mouse . Am J Physiol Renal Physiol . 2008 ;294 (2 ):F450 –4 . .18032546 25 Burrell LM , Risvanis J , Kubota E , Dean RG , MacDonald PS , Lu S , et al Myocardial infarction increases ACE2 expression in rat and humans . Eur Heart J . 2005 ;26 (4 ):369 –75 ; discussion 22–4. .15671045 26 Dean RG , Balding LC , Candido R , Burns WC , Cao Z , Twigg SM , et al Connective tissue growth factor and cardiac fibrosis after myocardial infarction . J Histochem Cytochem . 2005 ;53 (10 ):1245 –56 . .15956033 27 Lew RA , Warner FJ , Hanchapola I , Smith AI . Characterization of angiotensin converting enzyme-2 (ACE2) in human urine . Int J Pept Res Ther . 2006 ;12 (3 ):283 –9 . 28 Tikellis C , Bialkowski K , Pete J , Sheehy K , Su Q , Johnston C , et al ACE2 deficiency modifies renoprotection afforded by ACE inhibition in experimental diabetes . Diabetes . 2008 ;57 (4 ):1018 –25 . 10.2337/db07-1212 18235039 29 Dales NA , Gould AE , Brown JA , Calderwood EF , Guan B , Minor CA , et al Substrate-based design of the first class of angiotensin-converting enzyme-related carboxypeptidase (ACE2) inhibitors . J Am Chem Soc . 2002 ;124 (40 ):11852 –3 . Epub 2002/10/03. .12358520 30 Lai ZW , Hanchapola I , Steer DL , Smith AI . Angiotensin-Converting Enzyme 2 Ecotodomain Shedding Cleavage-Site Identification: Determinants and Constraints . Biochemistry . 2011 Epub 2011/05/14. 10.1021/bi200525y .21563828 31 Burrell LM , Johnston CI , Tikellis C , Cooper ME . ACE2, a new regulator of the renin-angiotensin system . Trends Endocrinol Metab . 2004 ;15 (4 ):166 –9 . .15109615 32 Crackower MA , Sarao R , Oudit GY , Yagil C , Kozieradzki I , Scanga SE , et al Angiotensin-converting enzyme 2 is an essential regulator of heart function . Nature . 2002 ;417 (6891 ):822 –8 . .12075344 33 Patel VB , Clarke N , Wang Z , Fan D , Parajuli N , Basu R , et al Angiotensin II induced proteolytic cleavage of myocardial ACE2 is mediated by TACE/ADAM-17: A positive feedback mechanism in the RAS . J Mol Cell Cardiol . 2014 ;66C :167 –76 . Epub 2013/12/18. 10.1016/j.yjmcc.2013.11.017 .24332999 34 Thatcher SE , Zhang X , Howatt DA , Yiannikouris F , Gurley SB , Ennis T , et al Angiotensin-converting enzyme 2 decreases formation and severity of angiotensin II-induced abdominal aortic aneurysms . Arterioscler Thromb Vasc Biol . 2014 ;34 (12 ):2617 –23 . Epub 2014/10/11. 10.1161/ATVBAHA.114.304613 25301841 35 Haber PK , Ye M , Wysocki J , Maier C , Haque SK , Batlle D . Angiotensin-converting enzyme 2-independent action of presumed angiotensin-converting enzyme 2 activators: studies in vivo, ex vivo, and in vitro . Hypertension . 2014 ;63 (4 ):774 –82 . Epub 2014/01/22. 10.1161/hypertensionaha.113.02856 .24446061 36 Raffai G , Khang G , Vanhoutte PM . Angiotensin-(1–7) augments endothelium-dependent relaxations of porcine coronary arteries to bradykinin by inhibiting ACE1 . J Cardiovasc Pharmacol . 2014 ;63 (5 ):453 –60 . Epub 2014/01/07. 10.1097/fjc.0000000000000069 .24390175 37 Ruebush TK 2nd, Rubin RH , Wolpow ER , Cassaday PB , Schultz MG . Neurologic complications following the treatment of human Babesia microti infection with diminazene aceturate . Am J Trop Med Hyg . 1979 ;28 (2 ):184 –9 . Epub 1979/03/01. .572145 38 Homeida AM , El Amin EA , Adam SE , Mahmoud MM . Toxicity of diminazene aceturate (Berenil) to camels . J Comp Pathol . 1981 ;91 (3 ):355 –60 . Epub 1981/07/01. .6274924 39 Sharma SP , Bansal GC . Chemoprophylaxis with diminazene aceturate in experimental Babesia bigemina infection in cattle . Res Vet Sci . 1984 ;37 (1 ):126 –7 . Epub 1984/07/01. .6473911

PMC005xxxxxx/PMC5003361.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757135310.1371/journal.pone.0161725PONE-D-16-16226Research ArticleBiology and Life SciencesPhysiologyPhysiological ProcessesSleepMedicine and Health SciencesPhysiologyPhysiological ProcessesSleepMedicine and Health SciencesHealth CarePatientsInpatientsBiology and Life SciencesPsychologyAddictionAlcoholismSocial SciencesPsychologyAddictionAlcoholismMedicine and Health SciencesMental Health and PsychiatrySubstance-Related DisordersAlcoholismMedicine and Health SciencesPublic and Occupational HealthSubstance-Related DisordersAlcoholismBiology and Life SciencesNutritionDietAlcohol ConsumptionMedicine and Health SciencesNutritionDietAlcohol ConsumptionMedicine and Health SciencesHealth CarePatientsOutpatientsMedicine and Health SciencesNeurologySleep DisordersPeople and placesPopulation groupingsEthnicitiesAfrican AmericansBiology and Life SciencesToxicologyDetoxificationMedicine and Health SciencesPathology and Laboratory MedicineToxicologyDetoxificationCritical Transitions: A Mixed Methods Examination of Sleep from Inpatient Alcohol Rehabilitation Treatment to the Community Mixed Methods Alcohol and SleepBrooks Alyssa Todaro 12*Krumlauf Michael 1Fryer Craig S. 3Beck Kenneth H. 3Yang Li 1Ramchandani Vijay A. 2Wallen Gwenyth R. 11 National Institutes of Health Clinical Center, Bethesda, Maryland, United States of America2 National Institute on Alcohol Abuse and Alcoholism, Bethesda, Maryland, United States of America3 University of Maryland School of Public Health, Department of Behavioral and Community Health, College Park, Maryland, United States of AmericaVerdejo-García Antonio EditorUniversity of Granada, SPAINCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: ATB GRW. Data curation: LY. Formal analysis: ATB LY MK. Funding acquisition: GRW. Investigation: ATB MK GRW. Methodology: ATB GRW KHB CSF. Project administration: ATB MK GRW. Resources: VAR GRW. Supervision: KHB GRW. Visualization: ATB MK. Writing – original draft: ATB. Writing – review & editing: ATB MK CSF KHB LY VAR GRW. * E-mail: todaroad@mail.nih.gov29 8 2016 2016 11 8 e016172521 4 2016 10 8 2016 This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.Aims This prospective, repeated measures study utilized a convergent parallel mixed methods approach to assess sleep experiences among individuals who were alcohol-dependent undergoing inpatient detoxification and treatment at a clinical research facility across the transition periods associated with the rehabilitation process: the initial adjustment to becoming an inpatient and the transition from inpatient to outpatient status. Methods This study included individual semi-structured interviews and quantitative measures relating to psychological distress, sleep quality, daytime sleepiness, and sleep-related beliefs and behavior (n = 33; 66.7% male). Interviews were conducted and questionnaires were administered within one week of participants’ scheduled discharge date and again four to six weeks post-discharge when they returned for a follow-up visit (or via phone). Results Participants self-reported significant sleep disturbances at both study time points. Of those participants with valid data at both time points (n = 28), there were no significant changes in mean scores from pre- to post-discharge with the exception of self-efficacy for sleep (SE-S) being significantly higher post-discharge. Preliminary qualitative findings suggested differences between those with ongoing sleep disturbances, those whose sleep disturbances had resolved, and those with no sleep disturbances at either time point. Conclusions This analysis highlights individual variation in sleep throughout the process of inpatient treatment and transition to outpatient aftercare in individuals with alcohol dependence. Collecting quantitative and qualitative data concurrently and combining emerging themes from qualitative data with quantitative analyses allowed for a more thorough examination of this relatively novel area of research and provided information that can be utilized to inform future behavioral sleep interventions. This project has been funded in whole or in part with federal funds from the National Institutes of Health, Clinical Center intramural research program. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. government. Data AvailabilityThe dataset is available only upon request because it includes clinical data and thus potentially identifiable information. Additionally, since the project was funded through the NIH Intramural Research Program, the NIH technically owns the data. Requests for the datasets can be sent to the principal investigator of the study, Dr. Gwenyth Wallen (gwallen@cc.nih.gov).Data Availability The dataset is available only upon request because it includes clinical data and thus potentially identifiable information. Additionally, since the project was funded through the NIH Intramural Research Program, the NIH technically owns the data. Requests for the datasets can be sent to the principal investigator of the study, Dr. Gwenyth Wallen (gwallen@cc.nih.gov). ==== Body Introduction Alcoholism, a chronic and progressive disease, is often accompanied by co-morbid conditions such as sleep disturbances [1–3]. Alcohol use can negatively affect sleep via increased nightmares, snoring, and other interruptions [4]. Sleep disturbances are common during phases of drinking and recovery [5], can persist for months or years during the process of recovery [4], and are especially common among individuals who are alcohol-dependent with co-morbid depression [6]. Reduced sleep is common during withdrawal after chronic abuse of alcohol [7], even after the “acute” withdrawal period [8]. Abstinence and “moderate” drinking significantly predict a reduction of insomnia symptoms in some patients, but symptoms may persist even after achieving abstinence [9]. Interestingly, in one study, subjective sleep measures were better predictors of future drinking (relapse) than objective measures (polysomnography) [10]. Insomnia is defined at the most basic level as difficulty falling or staying asleep and is relatively common in the U.S. [11]. Sleep disturbances are related to a multitude of health problems and can greatly reduce an individual’s quality of life [12–15]. Among treatment-seeking individuals who are alcohol-dependent, insomnia symptoms may increase psychosocial consequences related to alcohol [16]. Both pharmacologic and non-pharmacologic treatment options exist for individuals who are alcohol-dependent suffering with sleep disturbances ([17]. Relationship between sleep disturbances and relapse to drinking Sleep disturbances among individuals who are alcohol-dependent may be associated with increased risk of relapse to drinking following detoxification and rehabilitation [4, 18, 19]). Baseline sleep problems upon entering inpatient treatment have predicted subsequent relapse to drinking [18, 20]. Insomnia and sleep fragmentation after a period of abstinence may be related to relapse 14 months following abstinence [21]. For some individuals, re-initiation of drinking after achieving abstinence may be an attempt to self-medicate for disrupted sleep [22], and sleep-related behaviors such as the use of alcohol to help fall asleep have been associated with relapse 12 months after discharge from a residential addiction treatment program [23]. The longer-term consequences of sleep disturbances among individuals who are alcohol-dependent remain unclear. Transition periods during the alcohol rehabilitation process For individuals who are alcohol-dependent seeking inpatient treatment, both the initial adjustment to inpatient treatment and the transition from inpatient to outpatient status are time periods accompanied by many changes. The environment and structure of an inpatient facility warrant certain adjustments on the part of the individual. Upon discharge from an inpatient facility, continuity of care with diverse programmatic support structures have been utilized to help individuals maintain healthy lifestyle changes and maximize the likelihood of sustained sobriety [24]). “Transition groups” have been used to engage individuals in the process of transitioning from inpatient facilities to discharge into outpatient care [25], but many of these programs are focused solely on preventing relapse. Beyond the intent to remain sober, access to transportation for continuing care appointments, success of inpatient treatment, and motivation levels are all factors which may influence sobriety [26]. Getting adequate, quality sleep is not only an important component of a healthy lifestyle but, as demonstrated by previous literature [4, 18, 19], might also play a role in preventing relapse. Thus, it may be beneficial for clinicians to assess sleep patterns and any changes in sleep throughout the rehabilitation process should be monitored. Potential value of mixed methods research / study design Exploring individual perspectives and experiences could provide a better understanding of sleep throughout recovery [17]. In order to understand complex phenomena such as sleep disturbances and alcoholism, qualitative research that is naturalistic and subjective in nature combined with deductive quantitative techniques moves beyond traditional approaches and could potentially increase our knowledgebase [27]. Collecting both types of data concurrently and comparing emerging themes from qualitative data with quantitative analyses allows for a more thorough understanding of the complexity of sleep and the impact of any individual differences. This prospective, repeated measures study utilized a convergent parallel mixed methods approach to assess sleep experiences among individuals who were alcohol-dependent undergoing inpatient detoxification and treatment at a clinical research facility across the transition periods associated with the rehabilitation process: the initial adjustment to becoming an inpatient and the transition from inpatient to outpatient status. Methods This study was approved by the NIH Addictions Institutional Review Board (IRB) at the National Institutes of Health (NIH; NCT # 02181659). All participants enrolled in this study were first admitted to a clinical research facility providing inpatient detoxification and treatment under a screening and assessment protocol, which enrolls adults over 18 years of age seeking treatment for alcohol dependence. All participants received continued physical evaluations, inpatient treatment of alcohol withdrawal, psychosocial management, and an educational treatment program. Participants were eligible to receive up to six or more weeks of inpatient treatment followed by 16 weeks of optional outpatient treatment. Participants were paid for the study portions they completed based on NIH guidelines [28]. All participants signed an informed consent document indicating their voluntary participation and understanding of study procedures and expectations. Inclusion and exclusion criteria Participants were eligible for this study if they were 18 years of age or older, enrolled on the screening and treatment protocol (parent study), an inpatient for 21 days or more preceding discharge, not simultaneously enrolled onto a pharmacologic intervention study, able to understand the study and provide informed consent, and willing to return to the Clinical Center four to six weeks after being discharged from inpatient treatment for a follow-up visit or complete the follow-up study visit by phone. Study timeline Specific measures already collected upon inpatient admission as part of the screening and assessment protocol were used to characterize individuals who participated in this study. Approximately one week prior to participants' scheduled discharge, a study team member approached participants to begin the first segment of data collection for the study. Interviews were conducted and questionnaires were administered within one week of participants’ scheduled discharge date and again four to six weeks post-discharge when they returned for a follow-up visit (or via phone). Qualitative measures The qualitative component of this study was based on individual semi-structured interviews. The interview questions were reviewed and pilot-tested by clinicians and investigators with extensive experience working with alcohol-dependent individuals. A second interviewer (MK or GW) was present at all interviews and introduced to the participants with an explanation that he or she would observe, take notes, and probe additional questions based on the participant’s responses. This strategy was employed to decrease the potential bias of only having one interviewer. Questions were focused on sleep patterns prior to becoming an inpatient, during the inpatient stay, and in anticipation of becoming an outpatient and were designed to complement quantitative data. The first interview covered perceptions and descriptions of sleep in the clinic to the home environment, while the second interview focused on the participants’ support system as well as barriers and facilitators to both sleep and sobriety. All interviews were conducted by the first author (AB) for consistency, with the exception of one follow-up interview conducted by the second author (MK). The qualitative phase of data collection was always conducted first to ensure that participants’ responses would not be unduly influenced by their having read the sleep-related surveys prior to being interviewed. Quantitative measures Baseline–psychological distress (collected during inpatient phase only) Specific measures collected as part of the screening protocol were used to provide descriptive data on participants. The Comprehensive Psychopathological Rating Scale (CPRS) consists of 19 items that correspond to two CPRS-based subscales for affective and anxiety syndromes [29]: 1) the Montgomery Åsberg Depression Rating Scale (MADRS) [30] and 2) the Brief Scale for Anxiety (BSA) [31]. Overall scores range from 0 to 60 with higher scores indicative of more severe symptomatology. The Structured Clinical Interview for Diagnostics and Statistics Manual-IV (DSM-IV) (SCID-I) is the standard interview to evaluate criteria for a psychiatric diagnosis, including that of alcohol dependence and disorders that are frequently co-morbid with alcohol dependence [32]. It is a structured interview consisting of 11 modules with between 35–292 items per module that takes about 120–180 minutes. Interviews are carried out by trained mental health professionals whose inter-rater reliability is continuously monitored. We assessed the number of both anxiety and mood disorders from the SCID-I. The CIWA-AR: Clinical Institute Withdrawal Assessment-Alcohol Revised is a validated tool is used to determine the severity of alcohol withdrawal based on symptoms and physical signs [33]. Sleep quality and daytime sleepiness (assessed approximately one week pre-discharge and 4–6 weeks post-discharge) The Pittsburgh Sleep Quality Index (PSQI) is a 19-item, self-rated questionnaire used to measure sleep quality and disturbances over a one-month (30 days) time interval. A global summation score higher than five is indicative of poor sleep quality or “disturbed” sleep [34]. The PSQI has been extensively validated in populations with insomnia and other sleep disorders, with psychiatric patients, and in normal populations [35, 36]. Unlike all other assessments (which were administered at one week pre-discharge and/or 4–6 weeks post-discharge), the PSQI was administered at three different time points: baseline (day 2 of inpatient treatment), one week pre-discharge, and 4–6 weeks post-discharge. Internal reliability ranged from α = 0.576 to 0.840 at the pre- and post-discharge time points. The Epworth Sleepiness Scale (ESS) is an eight-item self-administered questionnaire that provides a measure of an individual’s general level of excessive daytime sleepiness over a one week time period [37]. Individuals are asked to rate their usual chances of dozing off or falling asleep on a four-point scale in eight distinct situations or activities that most people engage in during their daily lives. Higher scores are indicative of higher levels of daytime sleepiness. A score higher than ten is indicative of “excessive” daytime sleepiness [38]. Internal reliability of the ESS was high in the current study (α = .704 to 0.794 at both time points). Sleep-related beliefs and behaviors (assessed approximately one week pre-discharge and 4–6 weeks post-discharge) The Dysfunctional Beliefs and Attitudes about Sleep Scale (brief version: DBAS-16) is a 16-item questionnaire that assesses sleep-related cognitions including faulty beliefs and appraisals, unrealistic expectations, and perceptual and attention bias [39]. Higher scores are indicative of stronger endorsement of dysfunctional beliefs. The internal reliability of the DBAS-16 was high in the current study (α = 0.831 to 0.888 at both time points). The Self-Efficacy for Sleep Scale (SE-S) includes nine items used to measure the level of confidence a person has in performing behaviors that might be helpful in initiating sleep, with higher scores indicative of greater confidence [40]. Internal reliability of the SE-S was high in the current study (α = 0.768 to 0.843 at both time points). The Sleep-Related Behaviours Questionnaire (SRBQ) assesses the use of safety behaviors that individuals may use to promote sleep and cope with tiredness [41]. Higher scores are indicative of higher frequency of engaging in safety behaviors in an effort to cope with sleeplessness or tiredness. The internal reliability of the SRBQ was high in the current study (α = 0.834 to 0.843 at both time points). Alcohol-related measures–craving and relapse (collected 4–6 weeks post-discharge) The Penn Alcohol Craving Scale (PACS) is a clinical tool for practitioners to measure alcohol craving. It is a five-item self-administered instrument that measures frequency, intensity, and duration of thoughts about drinking along with ability to resist drinking (possible range: 0–30) with demonstrably excellent internal consistency, predictive validity, construct validity, and discriminant validity [42]. The Timeline Follow-Back (TLFB) collects drinking information using personal historical events recounted over a fixed time period [43]. It is a standard assessment for measuring alcohol drinking patterns and quantification in treatment programs and was the primary measures of relapse for this study. If the TLFB was missing or invalid, we used a positive Breath Alcohol level or participants’ voluntarily self-reporting relapse during the second interview as indicators of relapse. Analyses—qualitative data Each audio-recorded interview was transcribed and quality checked prior to analysis. A codebook was developed based on emergent themes related to transitions and changes in sleep over time from the interviews. A team of two coders independently reviewed a sub-set of transcripts. Discordant coding was discussed until consensus among the coding team was achieved. NVivo (version 10.0) was utilized for further qualitative analyses and to calculate inter-rater reliability percentages. Once the iterative process of consensus building was complete, a representative from the clinical team and a mixed methods expert from the NIH Clinical Center validated the themes and codes presented herein. To ensure that the trustworthiness of qualitative data was preserved, three criteria assessing rigor were considered: creditability, auditability, and fittingness of the data [44]. Analyses—quantitative data Statistical analyses of quantitative results were conducted with the Statistical Package for Social Sciences (SPSS) software, version 22.0. All quantitative data were double-data entered, cross-checked, and reconciled where necessary. Sleep quality (PSQI) and relapse status (TLFB and other sources) were the main outcomes of interest. Paired t-tests and McNemar tests were used to compare pre- and post-discharge differences in these main outcomes. Mixed model repeated measure analyses were used to assess the sleep quality (PSQI) changes over three study time points. A p value < 0.05 was considered significant for all analyses. Missing data was assumed to be missing at random. Use of mixed methods–convergent parallel design The format for the use of mixed methods was a convergent parallel design, wherein both quantitative and qualitative data were collected simultaneously and each method of examination was given equal priority [45]. Quantitative and qualitative results were merged during analysis and interpretation [46]. Results General demographics and clinical variables are presented in Table 1. On the second day of inpatient treatment, the average Pittsburgh Sleep Quality Index (PSQI) score was indicative of sleep disturbances (mean 12.0, s.d. 4.0). All individuals who were eligible to participate in the study agreed to participate and were enrolled. Three of the 33 participants reported being diagnosed with sleep apnea, one of whom also reported being diagnosed with nightmare disorder and restless legs syndrome. One additional participant reported substance/medication-induced sleep disorder. Of the 33 participants who completed the pre-discharge study visit, 28 (84.8%) returned for the post-discharge visit. The five participants who did not return were significantly more likely to be older (p < 0.05) and African-American (p < 0.05), but did not differ significantly from those who did return based on any other demographic, clinical, or sleep-related variables. 10.1371/journal.pone.0161725.t001Table 1 Participant demographics and clinical variables (n = 33).* n (%) Gender Male 22 (66.7) Female 11 (33.3) Race/ethnicity Black/African American 15 (45.4) White 16 (48.4) Other/multiracial 2 (6.0) Relapse (post-discharge, n = 28) Relapse 7 (21.2) No relapse 7 (21.2) Missing 14 (42.4) Marital status Single 22 (66.7) Divorced 7 (21.2) Married 3 (9.1) Not provided 1 (3.0) PTSD (not mutually exclusive categories) Current 4 (12.1) Past 7 (21.2) Lifetime 9 (27.3) Mood disorders (SCID)+ 18 (54.5) Anxiety disorders (SCID)+ 17 (51.5) Other substance use disorders (SCID; excluding alcohol) 21 (63.6) Current cannabis use Abuse 2 (6.1) Dependence 2 (6.1) Current cocaine use Dependence 1 (3.0) Range Mean (s.d.) Age 25–59 years 44.42 (10.43) TLFB–number of drinking days (out of 90) 12–90 days 65.55 (26.88) TLFB–number of heavy drinking days (out of 90) 11–90 days 62.70 (28.69) TLFB–average drinks per day (Range: 4.2–33.0) 4.2–33.0 drinks 13.27 (5.95) Baseline depression (CPRS)** (n = 32) 2–37 18.0 (7.7) Baseline anxiety (CPRS)** (n = 32) 2–30 13.2 (6.6) CIWA 0–20 6.60 (5.47) PACS (post discharge, n = 28) 0–28 8.73 (8.71) * If n ≠ 33 (data were missing), it is noted in the left column. ** “Baseline” denotes day 2 of inpatient treatment. +Denotes proportion of participants with one or more mood/anxiety disorders. CPRS: Comprehensive Psychopathological Rating Scale PACS: Penn Alcohol Craving Scale CIWA: Clinical Institute Withdrawal Assessment (maximum score over first four days of inpatient admission) TLFB: Timeline Follow-Back SCID: Structured Clinical Interview for DSM Disorders Of the 28 participants who returned for the follow-up visit, 14 (50%) had data on relapse (from the Timeline Follow-Back-TLFB, breath alcohol content at the time of the follow-up visit, and/or having voluntarily admitted to drinking during their follow-up visit). Of the eight participants who had valid Timeline Follow-Back (TLFB) data at follow-up (4–6 weeks post-discharge), only one had relapsed within the 4–6 week post-discharge time frame, reporting drinking four of the 34 days with an average of 7.13 drinks per drinking day. The nine participants who did not mention drinking during their follow-up interview were treated as “missing” since verification of sobriety was not possible (denoted in Table 1). Qualitative results The goal of achieving an inter-rater agreement of over 80% was met. Qualitative themes related to transitions and changes in sleep over time are presented, along the frequency of their endorsement at each time point, in S2 Table. It is important to note is that some themes did not “emerge” on their own but were actually prompted by the interviewer. A majority of participants (72.7%) discussed some type of fear or uncertainty surrounding either the adjustment to becoming an inpatient and/or returning back home during the pre-discharge interview. Similarly, a large majority of participants (82.1%) discussed the level of difficulty or ease of their transition back home during their follow-up interviews. While only 18.2% of participants discussed a healthy lifestyle in the context of their recovery process during their inpatient admission, almost half of those who returned for the follow-up interview endorsed healthy lifestyle changes since returning home. Most participants (60.6%) identified anticipated barriers and facilitators to sobriety before leaving the inpatient facility, but fewer participants (33.3%) discussed the actual barriers and facilitators they had experienced during the follow-up interviews. Participants discussed sleep-related behaviors (i.e. bedtime routines and/or strategies for falling asleep or staying asleep) at both time points (81.8% at pre-discharge and 78.6% at post-discharge). A considerable number of participants 39.4% discussed “racing thoughts” while reflecting on their sleep or drinking patterns during the first interview. Quantitative results We examined differences in sub-scales of the PSQI by time point in Table 2. McNemar tests were used to assess differences in the distribution of “sleep disturbances vs. no sleep disturbances” (PSQI) and “excessive daytime sleepiness vs. no excessive daytime sleepiness” (Epworth Sleepiness Scale: ESS). Among those with valid data at both time points (n = 26), the proportion of those with sleep disturbances as measured by the PSQI differed significantly from pre- to post-discharge (p < 0.05). There were no significant differences in the proportion of those with excessive daytime sleepiness at pre- and post-discharge as measured by the ESS (n = 28, Fig 1). 10.1371/journal.pone.0161725.g001Fig 1 Sleep-related variables pre- and post-discharge. * *McNemar test performed only in the case of valid data at both time points (PSQI: n = 26; ESS: n = 28). Five (5) participants were lost to follow-up. p < 0.05 (change in distribution of PSQI scores). 10.1371/journal.pone.0161725.t002Table 2 Pittsburgh Sleep Quality Index (PSQI) sub-scales. Pre-discharge, mean (s.d.) Post-discharge, mean (s.d.) n = 33* n = 28** Global score 7.50 (3.53) 6.35 (4.61) Sleep quality 0.94 (0.72) 0.82 (0.86) Sleep latency 1.61 (1.03) 1.39 (0.92) Sleep duration 1.21 (1.08) 0.71 (0.98) Sleep efficiency 0.73 (1.10) 0.61 (0.88) Sleep disturbance 1.50 (0.67) 1.33 (0.62) Sleep medication 0.61 (1.20) 0.81 (1.27) Daytime sleep dysfunction 0.85 (0.62) 0.59 (0.69) PSQI raw global scale range: 0–21; sub-score scale range: 0–3 *Sleep disturbance & global score pre-discharge; n = 32 **Sleep disturbance, sleep medication, daytime sleep dysfunction, & global score post-discharge, n = 27 The mean PSQI scores approximately one week pre-discharge and 4–6 weeks post-discharge remained above the cut-off for “disturbed sleep” (7.62 ± 3.70 and 6.35 ± 4.61, respectively; Fig 1). Of those participants with valid data at both time points (n = 26 for PSQI and n = 28 for all other sleep-related variables), there were no significant changes in mean scores on sleep-related variables from pre- to post-discharge with the exception of self-efficacy for sleep (SE-S) being significantly higher post-discharge (29.25 ± 6.70 versus 31.29 ± 7.62, p = 0.048). No differences were found between males and females when considering the change in PSQI scores from pre- to post-discharge. Interestingly, non-White participants were more likely to experience improvements in PSQI scores from pre- to post-discharge (p = 0.008). In addition to the McNemar test, a repeated measures linear mixed model using 88 time points from all 33 cases found that the estimated marginal mean for baseline sleep disturbances as measured by the PSQI on day 2 of inpatient treatment (12.29, s.e. 0.73) was significantly higher than one week pre- (7.47, s.e. 0.71) and 4–6 weeks post-discharge (6.27, s.e. 0.77) PSQI scores (p < .001 for both comparisons). No significant difference was found between one week pre- and 4–6 weeks post-discharge. There were no statistically significant differences in demographic or sleep-related variables between those who relapsed, those who were sober, and those whose relapse data were missing, likely due to the small sample size. Additionally, there were no differences in baseline average number of drinks per day in the 90 days prior to admission between the same groups (p-value from Kruskal-Wallis test = 0.48). However, demographic and clinical variables are presented by relapse status (if known) and “trajectory” of sleep quality based on PSQI scores pre- and post-discharge in Table 3. In the table, “ongoing sleep disturbances” refers to individuals whose PSQI scores were above the cut-off (> 5) for sleep disturbances at both time points. “Sleep disturbances resolved” refers to individuals whose pre-discharge PSQI score was above the cut-off for sleep disturbances but their post-discharge PSQI score was below the cut-off. Finally, “no sleep disturbances at either time point” refers to individuals whose PSQI scores were below the cut-off for sleep disturbances at both time points. Only one participant went from having no sleep disturbances pre-discharge to developing sleep disturbances post-discharge based on PSQI scores, although some participants outlined negative changes in sleep from pre- to post-discharge qualitatively. Those who relapsed had higher craving scores (Penn Alcohol Craving Scale—PACS) and those with no sleep disturbances at either time point had the lowest craving scores of any group, but the differences were not statistically significant. All of the participants whose sleep disturbances resolved were non-White. 10.1371/journal.pone.0161725.t003Table 3 Participant demographics and clinical variables by relapse status and sleep quality. Relapse (n = 7) No relapse (n = 7) n (%) n (%) Male 5 (71.4) Male 4 (57.1) Non-white 2 (28.6) Non-white 4 (57.1) Mean (s.d.) Mean (s.d.) Age 43.29 (14.33) Age 47.29 (9.76) PACS 18.14 (7.90)* PACS 4.29 (3.77)* Ongoing sleep disturbances (n = 11) Sleep disturbances resolved (n = 8) No sleep disturbances at either time point (n = 6) n (%) n (%) n (%) Male 6 (54.5) Male 6 (75.0) Male 5 (83.3) Non-white 1 (9.1) Non-white 8 (100.0) Non-white 1 (16.7) Mean (s.d.) Mean (s.d.) Age 44.73 (11.56) Age 46.13 (10.55) Age 45.00 (8.92) PACS 11.59 (10.28) PACS 6.75 (6.07) PACS 3.17 (4.02) *PACS (Penn Alcohol Craving Scale) was significantly higher among those who relapsed (p = .001). Preliminary qualitative findings suggested differences between those with ongoing sleep disturbances (n = 11), those whose sleep disturbances had resolved (n = 8), and those with no sleep disturbances at either time point (n = 6). Only one of the eight individuals whose sleep disturbances resolved and two of the 11 individuals with ongoing sleep disturbances mentioned healthy lifestyles during their pre-discharge interviews. Only one of the eight individuals whose sleep disturbances had resolved mentioned “racing thoughts” during their pre-discharge interview. Lastly, to complement the quantitative results and summarize the overarching emergent themes from the interviews, we present key qualitative findings by theme, specifically those related to the transition from inpatient to outpatient, in Table 4. 10.1371/journal.pone.0161725.t004Table 4 Summary of key qualitative findings by time point and theme. Pre-discharge qualitative themes Prevalent findings Sample quotes Fear / uncertainty related to transition to becoming an inpatient or returning home • Initial adjustment period upon arriving to inpatient facility (new environment, new “rules”—for some, this was while undergoing medically-assisted detoxification and treatment for withdrawal) • This was often followed by an “adjustment” period and developing a level of comfort with the inpatient rehabilitation program routine • Anxiety, excitement, or a mixture of both feelings regarding the transition back home • Particularly among those who had not attempted sobriety before, some degree of uncertainty surrounding not knowing “triggers” to relapse, whereas those who had been sober before often focused on what they would do differently • Regardless of prior experiences with rehabilitation, participants placed a lot of emphasis on maintaining sobriety / managing stress as primary determinants of their success post-discharge • “It took me a couple of days…to get adjusted…I observe things when I’m around new things or people… just to see how comfortable I can get.” -34 year old African American female, pre-discharge • “I would be disillusioned to tell you I got this thing figured out, or–anything like that…but…I can’t live here forever…” -45 year old White male, pre-discharge Healthy lifestyle (structure, health behaviors, health information) • Appreciation of the structure associated with the inpatient facility (regular meal times, normalizing sleep schedules, making time for physical activity) • Those with concerns about their physical health appreciated the clinical / diagnostic tests and receiving information on their health • “I have more energy… I focus my energy on more positive things…now, because drinking is eliminated, I’m using that…extra energy for good things. I go to the gym, I play basketball out back, I got my bike here on campus, I go bike riding…” -37 year old White male, pre-discharge • “Definitely now at this point, I…feel that I am on the right path. And I just have a new outlook on life. So, that’s where I am now…I’m a different person that I walked in here…30 –uh, 28 days ago.” -53 year old African American female, pre-discharge Sleep-related behavior (relaxation strategies and sleep hygiene techniques) • Initiation of bed-time routines or other sleep-related behaviors during the inpatient stay (use of relaxation techniques, herbal / pharmacological remedies, attempting to implement a regular sleep schedule, etc) • Anticipating the continuation of these behaviors post-discharge • “I go to sleep with…a little bit more contentment…I’ll put it that way…when I get up in the morning, I used to dance every morning in my bedroom. And I haven’t done that in the last three to four years. So, this morning I found myself dancing before I got dressed…that was cool.” -53 year old African American female, pre-discharge • “I’m back into meditation now…and I’ve also got chamomile tea. But I’m looking forward to sleep now. And I’ve got tools to cope with now.” -53 year old African American female, pre-discharge Mind or thoughts racing • Racing thoughts / inability to stop thinking, either when trying to go to sleep or as a precursor to drinking • “I couldn’t really go to sleep ‘cause my thinking was just…here come the thoughts again…a racket going on in my mind.” -55 year old African American male, pre-discharge • “Sleep’s terrible, you know, ‘cause your mind’s racing–with all the things that are really going on…” -47 year old African American male, pre-discharge Post-discharge themes Prevalent findings Transition back home • Feeling overwhelmed with the stress of “normal” life—including job interviews, family stressors, and other aspects of their lives they had been away from for at least 3 weeks • In some cases, these stressors led to relapse • Finding a job or finding purpose / meaning in other activities (e.g. re-connecting with family or volunteering in the community) were motivators to stay sober • Those who had not relapsed at the time of the second interview were more likely to perceive other aspects of the transition back home more positively • Participants indicated there were both “ups” and “downs” (positives and negatives) related to this transition • “A lot of times, I think about drinking and things like that, and just–I knew it was wrong, but I still would just follow my impulses. Now, like–I still get cravings every now and then, but every time they go in my head I think about all the bad stuff that happened, and they’ll go away.” -27 year old African American male • “[The] transition was a little rough. I got out…I went to stay with my cousin for a little bit…and then started drinking again, um…so that fell through…I have had one slip-up since I’ve been there…besides that…[I’m] just really trying to focus on sobriety.” -27 year old White male, post-discharge Lifestyle changes (health behaviors) • Healthy lifestyle changes mentioned during the second interview included being sober, having non-alcohol methods of coping with stressful situations, losing weight, increased physical activity, drinking less coffee, re-organizing living spaces or finances • “I’m not waking up looking for the vodka bottle. That’s…the blunt way to put it…I’ve filled my life with other things, and yes, you can still have fun without drinking.” -53 year old African-American female, post-discharge • “I’m eating much better…much healthier. And I’m still exercising almost every day, whether it’s–I mean, just taking a walk–and I think those two things are kind of important, um…at least, an overall balance, and I think–it definitely helps me.” -50 year old White male, post-discharge Sleep-related behavior (relaxation strategies and sleep hygiene techniques) • Some participants continued pre-bedtime rituals they initiated as inpatients, including drinking chamomile tea, meditation, calming music, guided imagery, reading, watching TV, and other methods of relaxation • “I still do try to meditate before I go to bed… and sometimes that… just calms me down, sometimes it doesn’t. It really depends on, I think, what I went through [during] the day.” -49 year old White female, post-discharge • “And I actually went and bought one of the little tapes that they recommended…it’s just got a bunch of, like, sea sounds in it. Bird chirps…” -47 year old African American male, post-discharge Overarching changes in sleep (since leaving inpatient facility) • Drunk dreams, dreaming more frequently (sometimes attributed to a change in medication) • Improved sleep (less interruptions, comfortable environment) • Many who had maintained sobriety felt that sleep was more regular / routine • Stress of the transition back home and everyday life potentially led to increased sleep interruptions • “I will drink and then pass out–it’s not sleeping, it’s being passed out from alcohol being infused into my whole system. And it’s not a deep sleep, it’s more of a knocked out sleep. And then, after a few hours–I mean, this is not eight, nine hours–after a few hours I wake up…I can also tell in my face…when I’m drinking and I don’t sleep well, which is–always happens…I notice bags under my eyes.” -57 year old White female, post-discharge Discussion This analysis highlights individual variation in sleep throughout the process of inpatient treatment and transition to outpatient aftercare in individuals with alcohol dependence. Collecting quantitative and qualitative data concurrently and combining emerging themes from qualitative data with quantitative analyses through triangulation (simultaneously) allowed for a more thorough examination of this relatively novel area of research and provided information that can be utilized to inform future behavioral sleep interventions. As previously discussed, the rehabilitation process represents a time of transformation. Inpatient facilities may represent an environment conducive to initiating lifestyle changes, including those which could improve sleep. As with previous studies [1, 47], the individuals in this study had a wide range of co-morbid conditions. Our findings support the co-occurrence of alcohol use and sleep disturbances [5, 48], particularly in the early stage of recovery [6, 49]. Individuals undergoing inpatient alcohol treatment are in a new environment, away from their homes and communities, and may therefore become accustomed to a “schedule” for eating, sleeping, and recreation. The first few days of the inpatient stay may not be the best time to introduce an intervention, as participants may be focused on adjusting to their surroundings. As evidenced by our results, the period of transition from inpatient to outpatient treatment represents another transition period of uncertainty and change. Sustaining healthy behaviors including sleep which may have been initiated during inpatient treatment could help to maintain sobriety, a healthy lifestyle, and overall health-related quality of life. Similar to our previous work establishing the prevalence of sleep disturbances throughout the inpatient stay, there was low variability in daytime sleepiness levels (ESS) and average scores on the measure were not indicative of “excessive” daytime sleepiness. Measures of sleep-related safety behaviors (SRBQ) and dysfunctional beliefs about sleep (DBAS-16) remained relatively stable at both time points, despite participants discussing many changes occurring throughout the transition from inpatient to outpatient during their interviews. Self-efficacy for sleep (SE-S) improved significantly with inpatient standard of care which did not include specific interventions focused on sleep hygiene. However, study participants discussed bedtime routines and strategies for sleep they initiated as inpatients. In the inpatient program, it is possible that staff members could have provided suggestions for non-pharmacologic interventions. Additionally, there is a “schedule” for the unit when activities end each night and begin the next morning–and patients typically adjust accordingly. Despite these potential confounding factors, self-efficacy will be explored as an important variable in sleep behavior change in a larger sample, given its sensitivity in this small sample. The demographic and clinical variables presented in Table 3 did not differ significantly by relapse status or sleep quality, but one trend emerged that could have clinical implications and warrants further investigation: those who self-reported no sleep disturbances at either time point had the lowest craving scores. Craving may be important to examine at various points throughout the recovery process, especially when considering the possibility of relapse. Furthermore, measures of sleep quality (PSQI) and daytime sleepiness (ESS) both trended toward reduction in the prevalence from pre- to post-discharge. Additionally, several interesting qualitative differences emerged by sleep quality status: 1) those who mentioned healthy lifestyles pre-discharge were less likely to self-report no sleep disturbances at either time point and 2) those whose sleep disturbances resolved by the post-discharge time point were less likely to endorse “racing thoughts” while inpatients. These findings offer preliminary support to the idea that “stabilizing” sleep during the inpatient phase may be especially important. Strengths and limitations The mixed methods nature of this study allowed for the collection of rich and diverse information. In particular, the wealth of qualitative data accrued and the rigorous process through which they were analyzed was useful in describing the complex phenomenon of sleep throughout recovery. However, this study is not without limitations. Given a small sample size and subsequently low power levels, the quantitative analyses were strictly exploratory in nature and were meant to complement qualitative findings. The sample was non-representative; findings should not be generalized to all individuals who are alcohol-dependent, seeking treatment, and willing to return for a follow-up visit. Finally, at discharge, all but three participants were prescribed at least one medication that could have potentially altered their sleep. It is unclear how this use of sleep medications affected the nature of the qualitative narratives provided by the participants. Future directions Future efforts should include following individuals for a longer period of time post-discharge when possible to capture participants’ experiences longitudinally and explore whether lasting changes to sleep parameters occur. The quotes reflected in this paper coalesce around the overarching theme of structure; the appreciation of re-developing routines during inpatient treatment may be conducive to changing sleep habits through a tailored behavioral sleep intervention. At a minimum, sleep should be monitored in alcohol treatment programs in order to understand it both as a potential relapse trigger and an important component of a healthy lifestyle post-discharge. The most effective mechanism for assessing sleep in this particular population (i.e. objectively, subjectively, or a combination of both) still needs to be explored. Conclusion This research confirms a high prevalence of sleep disturbances in a sample of treatment seeking individuals who are alcohol-dependent throughout various stages of recovery that had not yet been explored in detail using a mixed methods approach. Patient-reported outcomes and in-depth interviews provided a clearer picture of individual experiences throughout recovery and the complexity of alcoholism and associated co-morbid conditions, compared to either data source individually. Capturing the essence of transition periods throughout the process of recovery highlights the important role they may play in sleep quality and eventually relapse. Of particular importance may be learning how to capitalize on positive lifestyle changes post-discharge from inpatient rehabilitation facilities that emerged qualitatively in this study, particularly those changes introduced during the inpatient phase of recovery and sustained. This study fills a gap in the literature by characterizing sleep throughout the rehabilitation process and ongoing maintenance of abstinence (or relapse). Supporting Information S1 Table Interview prompts. Description of interviewer prompts for both the pre-and post-discharge interviews. (DOCX) Click here for additional data file. S2 Table Qualitative themes specific to transitions and sleep changes over time. Description of themes relevant to transition from pre- to post-discharge and corresponding frequency of endorsements. (DOCX) Click here for additional data file. This paper was part of a dissertation through the University of Maryland School of Public Health, Department of Behavioral and Community Health (College Park, MD). We gratefully acknowledge Dr. Carl Lejuez and Dr. Min Qi Wang (University of Maryland), Dr. Nancy Diazgranados and Dr. Melanie Schwandt (NIAAA), Miriam Magana-Lopez, the staff of the 1SE inpatient unit and outpatient clinic, and the patients who agreed to participate in this study. ==== Refs References 1 Benca RM . Sleep in psychiatric disorders . Neurologic Clinics .1996 ; 14 : 739 –764 . 8923493 2 World Health Organization (WHO). Public health problems caused by harmful use of alcohol.2013a. Available: http://www.who.int/substance_abuse/activities/public_health_alcohol/en/index.html. 3 World Health Organization (WHO). Alcohol. 2013b. Available: http://www.who.int/substance_abuse/facts/alcohol/en/index.html. 4 Landolt HP and Gillin JC . Sleep abnormalities during abstinence in alcohol-dependent patients. Aetiology and management . CNS Drugs . 2001 ; 15 : 413 –425 . 11475945 5 Gillin JC and Drummond SP . Medication and substance abuse In: M.H. Kryger MH , Roth T , and Dement WC , editors. Principles and Practice of Sleep Medicine , 3rd edition Philadelphia, PA : WB Saunders ; 2000 . 6 Gillin JC , Smith TL , Irwin M , Kripke DF , and Schuckit M . EEG sleep studies in "pure" primary alcoholism during subacute withdrawal: relationships to normal controls, age, and other clinical variables . Biological Psychiatry . 1990 ; 27 : 477 –488 . 2310803 7 Alling C , Balldin J , Bokstrom K , Gottfries CG , Karlsson I , and Langstom G . Studies on duration of a late recovery period after chronic abuse of ethanol. A cross-sectional study of biochemical and psychiatric indicators . Acta Psychiatrica Scandinavica .1982 ; 66 : 384 –397 . 6129776 8 Mossberg D. , Liljeberg P , Borg S . Clinical conditions in alcoholics during long-term abstinence: a descriptive, longitudinal treatment study . Alcohol . 1985 ; 2 : 551 –553 . 4026978 9 Brower KJ , Krentzman A , and Robinson EA . Persistent insomnia, abstinence, and moderate drinking in alcohol-dependent individuals . Am J Addict . 2011 ; 20 : 435 –440 . 10.1111/j.1521-0391.2011.00152.x 21838842 10 Conroy DA , Todd Arnedt J , Brower KJ , Strobbe S , Consens F , Hoffmann R , and Armitage R . Perception of sleep in recovering alcohol-dependent patients with insomnia: relationship with future drinking . Alcohol Clin Exp Res . 2006 ; 30 : 1992 –1999 . 17117964 11 National Heart, Lunch, and Blood Institute (NHLBI). What Is Insomnia? 2011. Available: http://www.nhlbi.nih.gov/health/health-topics/topics/inso/. Accessed 1 March 2016. 12 Ayas NT , White DP , Al-Delaimy WK , et al A prospective study of self-reported sleep duration and incident diabetes in women . Diabetes Care . 2003 ; 26 : 380 –384 . 12547866 13 Stine TW and Chapman DP . Associations of frequent sleep insufficiency with health-related quality of life and health behaviors . Sleep Medicine . 2005 ; 6 : 23 –27 . 15680291 14 Vgontzas AN , Liao D , Bixler EO , Chrousos GP , and Vela-Bueno A . Insomnia with objective short sleep duration is associated with a high risk for hypertension . Sleep . 2009 ; 32 : 491 –497 . 19413143 15 Vgontzas AN , Liao D , Pejovic S , Calhoun S , Karataraki M , and Bixler EO Insomnia with objective short sleep duration is associated with type 2 diabetes: a population-based study . Diabetes Care . 2009 ; 32 : 1980 –1985 . 10.2337/dc09-0284 19641160 16 Chaudhary NS , Kampman KM , Kranzler HR , Grandner MA , Debbarma S , and Chakravorty S . Insomnia in alcohol dependent subjects is associated with greater psychosocial problem severity . Addict Behav . 2015 ; 50 : 165 –172 . 10.1016/j.addbeh.2015.06.021 26151580 17 Brooks AT and Wallen GR . Sleep disturbances in individuals with alcohol-related disorders: a review of cognitive-behavioral therapy for insomnia (CBT-I) and associated non-pharmacological therapies . Substance Abuse: Research and Treatment . 2014 ; 16 : 55 –62 . 18 Brower KJ , Aldrich MS , Robinson EA , Zucker RA , and Greden JF . Insomnia, self-medication, and relapse to alcoholism . American Journal of Psychiatry . 2001 ; 158 : 399 –404 . 11229980 19 Foster JH and Peters TJ . Impaired sleep in alcohol misusers and dependent alcoholics and the impact upon outcome . Alcoholism: Clinical and Experimental Research . 1999 ; 23 : 1044 –1051 . 20 Brower KJ . Alcohol's effects on sleep in alcoholics . Alcohol Research & Health . 2001 ; 25 : 110 –125 .11584550 21 Drummond SP , Gillin JC , Smith TL , and DeModena A . The sleep of abstinent pure primary alcohol patients: natural course and relationship to relapse . Alcohol Clin Exp Res . 1998 ; 22 : 1796 –1802 . 9835298 22 Vitiello MV . (1997 ) Sleep, alcohol, and alcohol abuse . Addiction Biology , 2 (2 ): 151 –158 . 10.1080/13556219772697 26735632 23 Kolla BP , Schneekloth T , Mansukhani MP , Biernacka JM , Hall-Flavin D , Karpyak V , Geske J , and Frye MA . The association between sleep disturbances and alcohol relapse: a 12-month observational cohort study . Am J Addiction . 2015 ; 24 : 362 –367 . 24 McKay JR . Continuing care in the treatment of addictive disorders . Curr Psychiatry Rep . 2006 ; 8 : 355 –362 . 16968615 25 Hanson M , Foreman L , Tomlin W , Bright Y . Facilitating problem drinking clients’ transition from inpatient to outpatient care . Health & Social Work . 1994 ; 19 : 23 –28 .8168774 26 Schaefer JA , Harris AH , Cronkite RC , and Turrubiartes P . () Treatment staff’s continuity of care practices, patients’ engagement in continuing care, and abstinence following outpatient substance-use disorder treatment . Journal of Studies on Alcohol and Drugs , 2008 ; 69 : 747 –756 . 18781250 27 Newman I , Ridenour CS , Newman C , and DeMarco GMP . A typology of research purposes and its relationship to mixed methods In: Tashakkori A , Teddlie C , editors. Handbook of Mixed Methods in Social & Behavioral Research . Thousand Oaks, CA, USA : Sage Publications ; 2003 . 28 Dominguez D , Jawara M , Martino N , Sinaii N , and Grady C . Commonly performed procedures in clinical research: a benchmark for payment . Contemporary Clinical Trials . 2012 ; 33 : 860 –868 . 10.1016/j.cct.2012.05.001 22580210 29 Åsberg M , Montgomery SA , Perris C , Schalling D , and Sedvall G . A comprehensive psychopathological rating scale . Acta Psychiatrica Scandinavica, Supplementum . 1978 ; 271 : 5 –27 . 30 Montgomery SA and Åsberg M . A new depression scale designed to be sensitive to change . The British Journal of Psychiatry . 1979 ; 134 : 382 –389 . 444788 31 Tyrer P , Owen RT , and Cicchetti DV . The brief scale for anxiety: a subdivision of the comprehensive psychopathological rating scale . Journal of Neurology, Neurosurgery, & Psychiatry . 1984 ; 47 : 970 –975 . 32 First MB , Spitzer RL , Gibbon M , Williams JBW . User's Guide for the SCID-I: Structured Clinical Interview for DSM-IV-TR Axis I Disorders- Research Version- SCID-I for DSM-IV-TR, November 2002 Revision . New York, NY : Biometrics ; 2002 . 33 Sullivan JT , Sykora K , Schneiderman J , Naranjo CA , Sellers EM . Assessment of alcohol withdrawal: the revised clinical institute withdrawal assessment for alcohol scale (CIWA-Ar) . Br. J. Addict .1989 ; 84 :1353 –1357 . 2597811 34 Buysse DJ , Reynolds CF III, Monk TH , Berman SR , and Kupfer DJ . The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and research . Psychiatry Research . 1989 ; 28 : 193 –213 . 2748771 35 Doi Y , Minowa M , Uchiyama M , Okawa M , Kim K , Shibui K , and Kamei Y . Psychometric assessment of subjective sleep quality using the Japanese version of the Pittsburgh Sleep Quality Index (PSQI-J) in psychiatric disordered and control subjects . Psychiatry Research . 2000 ; 97 : 165 –172 . 11166088 36 Backhaus J , Junghanns K , Broocks A , Riemann D , and Hohagen F . Test-retest reliability and validity of the Pittsburgh Sleep Quality Index in primary insomnia . Journal of Psychosomatic Research . 2002 ; 53 : 737 –740 . 12217446 37 Johns MW . A new method for measuring daytime sleepiness: the Epworth sleepiness scale . Sleep . 1991 ; 14 : 540 –545 . 1798888 38 Johns MW . Daytime sleepiness, snoring, and obstructive sleep apnea. The Epworth Sleepiness Scale . Chest . 1993 ; 103 : 30 –36 . 8417909 39 Morin CM , Vallieres A , and Ivers H . Dysfunctional beliefs and attitudes about sleep (DBAS): validation of a brief version (DBAS-16) . Sleep . 2007 ; 30 : 1547 –1554 . 18041487 40 Edinger JD , Wohlgemuth WK , Radtke RA , Coffman CJ , and Carney CE . Dose-response effects of cognitive behavioral insomnia therapy: a randomized clinical trial . Sleep . 2007 ; 30 : 203 –212 . 17326546 41 Ree MJ and Harvey AG . Investigating safety behaviours in insomnia: the development of the sleep-related behaviours questionnaire (SRBQ) . Behaviour Change . 2004 ; 21 : 26 –36 . 42 Flannery BA , Volpicelli JR , and Pettinati HM . Psychometric properties of the Penn Alcohol Craving Scale . Alcoholism: Clinical and Experimental Research .1999 ; 23 : 1289 –1295 . 43 Sobell LC and Sobell MB . Timeline follow-back: a technique for assessing self-reported ethanol consumption In: Allen J , Litten R , editors. Measuring Alcohol Consumption: Psychosocial and Biochemical Methods . Totowa, NJ : Humana Press ; 1992 pp. 42 –72 . 44 Barroso J . Qualitative approaches to research In LoBiondo-Wood G , Haber J , editors. Nursing Research: Methods and Critical Appraisal for Evidence-Based Practice . St. Louis, MO, USA : Elsevier ; 2010 . 45 Creswell JW and Plano Clark VL . Designing and Conducting Mixed Methods Research . SAGE Publications, Inc : Thousand Oaks, CA, USA ; 2011 . 46 Bishop FL and Holmes MM . (2013 ) Mixed methods in CAM research: a systematic review of studies published in 2012 . Evidence-Based Complementary and Alternative Medicine 2013 . 2013. 10.1155/2013/187365 47 Levola J , Aalto M , Holopainen A , Cieza A , and Pitkanen T . Health-related quality of life in alcohol dependence: a systematic literature review with a specific focus on the role of depression and other psychopathology . Nordic Journal of Psychiatry . 2014 ; 68 : 369 –384 . 10.3109/08039488.2013.852242 24228776 48 Teplin D , Raz B , Daiter J , Varenbut M , and Tyrrell M . Screening for substance use patterns among patients referred for a variety of sleep complaints . Am J Drug Alcohol Abuse . 2006 ; 32 : 111 –120 . 16450646 49 Wallen GR , Brooks AT , Whiting B , et al (2014 ) The prevalence of sleep disturbance in alcoholics admitted for treatment: a target for chronic disease management . Family & Community Health . 1997; 37 : 288 –297 .25167069

PMC005xxxxxx/PMC5003362.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757126810.1371/journal.pone.0161386PONE-D-16-00286Research ArticleMedicine and Health SciencesDiagnostic MedicineDiagnostic RadiologyMagnetic Resonance ImagingResearch and Analysis MethodsImaging TechniquesDiagnostic RadiologyMagnetic Resonance ImagingMedicine and Health SciencesRadiology and ImagingDiagnostic RadiologyMagnetic Resonance ImagingMedicine and Health SciencesOncologyCancers and NeoplasmsHematologic Cancers and Related DisordersLymphomasMedicine and Health SciencesHematologyHematologic Cancers and Related DisordersLymphomasMedicine and Health SciencesOncologyCancers and NeoplasmsCarcinomasAdenocarcinomasBiology and Life SciencesNeuroscienceBrain MappingBrain MorphometryDiffusion Weighted ImagingMedicine and Health SciencesDiagnostic MedicineDiagnostic RadiologyMagnetic Resonance ImagingBrain MorphometryDiffusion Weighted ImagingResearch and Analysis MethodsImaging TechniquesDiagnostic RadiologyMagnetic Resonance ImagingBrain MorphometryDiffusion Weighted ImagingMedicine and Health SciencesRadiology and ImagingDiagnostic RadiologyMagnetic Resonance ImagingBrain MorphometryDiffusion Weighted ImagingResearch and Analysis MethodsImaging TechniquesNeuroimagingBrain MorphometryDiffusion Weighted ImagingBiology and Life SciencesNeuroscienceNeuroimagingBrain MorphometryDiffusion Weighted ImagingMedicine and health sciencesOncologyCancers and neoplasmsHematologic cancers and related disordersLymphomasDiffuse large B-cell lymphomaMedicine and health sciencesHematologyHematologic cancers and related disordersLymphomasDiffuse large B-cell lymphomaBiology and Life SciencesAnatomyHistologyMedicine and Health SciencesAnatomyHistologyBiology and Life SciencesAnatomyNervous SystemCentral Nervous SystemMedicine and Health SciencesAnatomyNervous SystemCentral Nervous SystemMedicine and Health SciencesDiagnostic MedicineSigns and SymptomsLesionsMedicine and Health SciencesPathology and Laboratory MedicineSigns and SymptomsLesionsDiffusion-Weighted MRI Reflects Proliferative Activity in Primary CNS Lymphoma Diffusion-Weighted MRI Reflects Proliferative Activity in Primary CNS Lymphomahttp://orcid.org/0000-0003-2846-5443Schob Stefan 1*Meyer Jonas 4Gawlitza Matthias 1Frydrychowicz Clara 2Müller Wolf 2Preuss Matthias 5Bure Lionel 6Quäschling Ulf 1Hoffmann Karl-Titus 1Surov Alexey 31 Department of Neuroradiology, University Hospital Leipzig, Leipzig, Germany2 Department of Neuropathology, University Hospital Leipzig, Leipzig, Germany3 Department of Diagnostic and Interventional Radiology, University Hospital Leipzig, Leipzig, Germany4 Department of Radiology, Martin Luther University of Halle-Wittenberg, Halle-Wittenberg, Germany5 Department of Neurosurgery, University Hospital Leipzig, Leipzig, Germany6 Department of Radiology, McGill University Health Center, Montreal General Hospital, Montreal, CanadaColes Jonathan A EditorUniversity of Glasgow, UNITED KINGDOMCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: SS AS UQ KTH. Data curation: SS AS KTH MG. Formal analysis: MG JM LB CF. Investigation: SS AS UQ. Methodology: SS AS UQ KTH CF LB. Project administration: AS KTH UQ WM. Resources: WM CF. Software: JM WM MP. Supervision: KTH UQ AS WM. Validation: MG JM MP LB. Visualization: SS AS UQ. Writing – original draft: SS AS. Writing – review & editing: SS AS. * E-mail: Stefan.Schob@medizin.uni-leipzig.de29 8 2016 2016 11 8 e01613865 1 2016 4 8 2016 © 2016 Schob et al2016Schob et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Purpose To investigate if apparent diffusion coefficient (ADC) values within primary central nervous system lymphoma correlate with cellularity and proliferative activity in corresponding histological samples. Materials and Methods Echo-planar diffusion-weighted magnetic resonance images obtained from 21 patients with primary central nervous system lymphoma were reviewed retrospectively. Regions of interest were drawn on ADC maps corresponding to the contrast enhancing parts of the tumors. Biopsies from all 21 patients were histologically analyzed. Nuclei count, total nuclei area and average nuclei area were measured. The proliferation index was estimated as Ki-67 positive nuclei divided by total number of nuclei. Correlations of ADC values and histopathologic parameters were determined statistically. Results Ki-67 staining revealed a statistically significant correlation with ADCmin (r = -0.454, p = 0.038), ADCmean (r = -0.546, p = 0.010) and ADCmax (r = -0.515, p = 0.017). Furthermore, ADCmean correlated in a statistically significant manner with total nucleic area (r = -0.500, p = 0.021). Conclusion Low ADCmin, ADCmean and ADCmax values reflect a high proliferative activity of primary cental nervous system lymphoma. Low ADCmean values—in concordance with several previously published studies—indicate an increased cellularity within the tumor. The authors received no specific funding for this work. Data AvailabilityAll relevant data are within the paper.Data Availability All relevant data are within the paper. ==== Body Introduction Primary lymphomas of the central nervous system account for approximately 4% of newly diagnosed brain tumors[1]. Representing a subgroup of Non-Hodgkin Lymphoma (NHL), more than 90% of them are classified as Diffuse Large B-Cell Lymphoma (DLBCL). Occurrence of primary CNS lymphoma (PCNSL) strongly correlates with immunosuppression, either iatrogenic, infectious or congenital[2]. Although PCNSL in immunocompromised subjects show an association with Epstein-Barr Virus (EBV), the incidence rate of PCNSL in immunocompetent subjects without EBV has increased sharply over the last thirty years[3]. PCNSL rarely extends beyond the CNS. This is postulated to be due to the immune privilege of the CNS[4]. The neoplastic lymphocytes are identified and removed from the immunologically more active periphery, but persist in the CNS where the immunological surveillance is comparatively less fine-meshed, enabling them to maintain active proliferation. In clinical practice, Ki-67 is a widely used cellular marker for proliferative activity in lymphoma and other malignancies[5]. The nonhistone nuclear protein is synthesized throughout the whole cell cycle except the G0 phase, and has been shown to be responsible for cell division[6]. The prognostic value of Ki-67 in different histopathologic subtypes of lymphoma has been examined in multiple studies[5]. For example, a high level of Ki-67 expression is associated with a decreased overall survival but a good therapeutical response of subjects suffering from DLBCL following Rituximab administration. Sustained proliferation and enabled replicative immortality, both being general hallmarks of cancer[7] result in a high cellular density of the malignancy, which is a consistent finding in PCNSL[8]. Diffusion-Weighted Imaging (DWI), a technique based on phase-defocusing and phase refocusing gradients has been used to measure the motion of water molecules on the microspcopic scale. Besides its importance in stroke imaging, DWI plays a growing role in the characterization of brain tumors. Tumors with densely packed cells like astrocytoma, lymphoma and medulloblastoma[9] reveal an increased DWI signal on high b-values images and low ADC values due to the relative restriction of water caused by high cellularity in combination with high nuclear-to-cytoplasmic (N/C) ratios. The relation between low ADC values, high cellularity and poor outcome has already been shown for PCNSL[10]. However, not all studies in the field demonstrate an association between cellularity and ADC-values: Wu et al.[11] did not find a correlation between ADC values and cellularity in Diffuse Large B-Cell Lymphoma and Follicular Lymphoma. Although the relation between N/C ratio, cellularity and ADC values has been examined in PCNSL, the association between water diffusibility (ADCmin, ADCmax, ADCmean), proliferation rate (Ki-67 expression), total nuclear area, average nuclear area and cell count has not yet been elucidated. The purpose of this study was i) to investigate if ADC values in PCNSL correlate with proliferative activity in terms of Ki-67 expression and ii) to further elucidate if ADC values (ADCmin, ADCmean and ADCmax) show correlations with nuclei count, total nucleic area or average nucleic size. Materials and Methods Patient Selection All patients or caregivers gave their informed consent for the asservation of sample remnants and compiling of clinical and radiological data. Informed consent was given in writing. The study was approved by the local ethics committee (Ethikkommission Universität Leipzig, Az 330-13-18112013). Potential patients in the period of January 2006 through March 2015 were identified on the basis of the diagnosis PCNSL conducting a full-text search in the database of the institutes for pathology and radiology, respectively. The search revealed 40 patients in the radiological database, all of which were biopsied in our hospital and had a consecutive histopathological workup. Only previously untreated patients with pretreatment DWI were included. None of the studied patients were immunodeficient. Due to a lack of comparability of DWI measurements between different MR scanners (Philips Intera, Philips Achieva and Siemens Symphony, all 1.5T) we excluded 19 patients from our analysis. 21 patients (9 female, 13 male; 28–89 years, mean age 68,5 years) met the inclusion criteria. MR Imaging All images were obtained in the clinical routine workup by using a 1.5T MRI scanner (Siemens Magnetom Symphony 1,5T) with the standard Siemens head coil (CP head array, model #1P3146037). DWI was performed using a single-shot spin-echo (SE) echo planar sequence with following parameters: Echo time(TE)/Repetition time(TR) = 6000/105 ms, 90° flip angle, 57 transverse sections, slice thickness = 5 mm, field of view (FOV) = 230 mm. Diffusion-sensitizing gradients were applied sequentially in the x, y and z directions with b factors of 0 and 1000 s/mm2. ADCs were automatically calculated at the operating console of the MR scanner and displayed as corresponding ADC maps. Postcontrast T1-weighted 3D-gradient echo sequence(GRE) imaging was obtained with following parameters: TR/TE = 2150/3.93 ms, flip angle 15°, 1-mm section thickness and 230 mm FOV. A standard dose (0.1 mmol/kg body weight) of gadoteric acid (Gd-DOTA, Dotarem; Laboratoire Guerbet, Aulnay-sous-Bois, France) was injected intravenously. Routine anatomic precontrast T1/ T2_tirm_tra_dark_fluid (TR/TE = 9000/114, slice thickness 5mm, flip angle 150°, 28 transverse sections) images were also obtained. MR Image Analsysis All images were available in digital format. Image analysis was performed on a Siemens SIENET MagicView 1000 console previous to stereotactic biopsy. The lesion in question was determined by the stereotactic neurosurgeon. After stereotactic biopsy postinterventional images were compared with preoperative images to confirm that the analysed lesion corresponded to the biopsied portion of the tumor. B0, B1000 and ADC maps were coregistered with the postcontrast T1-weighted 3D-gradient echo sequence to improve visualization and correlation. Subsequently the readers reviewed the coregistered B0, B1000 and ADC maps and drew a single circular region of interest (ROI)within the enhancing part of the tumor corresponding to the biopsied portion. ROI size was determined depending on the size of the post-biopsy lesion, the lower limit being 0.2cm2, the upper limit 0.49cm2. Hemorrhage, cysts and necrosis were avoided when drawing the ROIs. ADC min, max and mean were measured on the corresponding maps. Additionally, deltaADC (deltaADC = ADCmax—ADCmin) was calculated for each tumor. The used technique for measuring ADC values was chosen because of its robustness and clinical feasibility. Histopathological Analysis In every case the diagnosis of cerebral lymphoma was confirmed by surgical biopsy. Ki-67 antigen stained specimens (MIB-1 monoclonal antibody, DakoCytomation, Denmark) were re-analyzed in this study. All images were digitalized by using a research microscope Jenalumar with camera Diagnostic instruments 4.2 (Jena, Germany). The tumor proliferation index was estimated as following relation: number of specifically stained nuclei divided by all nuclei. The area showing the highest number of positive cell nuclei was selected in each case. Cell count was calculated as total number of nuclei per two high power fields (x400). In addition, the histological samples were analyzed for estimation of total and average nucleic areas by using ImageJ software 1.48v (NIHS) as described in previous reports [12]. Tumor cell nuclei were easily identified by their color intensities. Total and average nucleic areas were automatically calculated in two high power fields (x400). Statistical analysis Statistical analysis and graphics creation were performed using SPSS version 22. Collected data was evaluated by means of descriptive statistics. Spearman-Rho correlation coefficient was used to analyze the association between ADC values and histological parameters. P-values < 0.05 were taken to indicate statistical significance. Results DWI measurements and histopathological analysis All investigated PCNSL were located supratentorially and showed close proximity to the lateral ventricles. The size of the lesions ranged from 1.2cm to 7.8cm. 6 out of 21 patients had multifocal supratentorial lesions. As described above only the stereotactically biopsied part of the tumor was analysed with MRI. Fig 1a exemplarily shows a representative T1 weighted spin echo image of a CNS lymphoma after gadolinium administration. Fig 1b shows the corresponding ADC map. 10.1371/journal.pone.0161386.g001Fig 1 MRI and histopathological findings in a patient with diagnosed cerebral lymphoma. a. T1w image documenting a lesion with marked homogenous enhancement. b. ADC map. The ADCmin value of the lesion is 64x10-4mm2s-1, the ADCmean value of the lesion is 80×10−4 mm2s-1 and the ADCmax value is 133×10−4 mm2s-1. c. Immunohistochemical stain (MIB-1 monoclonal antibody). Ki 67-index = 50%. d. “Particles tool” analysis image (ImageJ converted MIB-1 image). Cell count (total number of nuclei in the image), total nucleic area (total area of all nuclei in the image) and average nucleic area (total area of all nuclei in the image divided by the number of nuclei in the image) were calculated via the particles tool in ImageJ. For the image shown in fig 1d the following values were determined by the “Particles tool” analysis; cell count = 1424, total nucleic area = 105678.14 μm2 and average nucleic area = 74.37 μm2. Table 1 displays gender, age and ADC fractions. Table 2 summarizes ADC values, cell count, Ki-67 expression, total nucei area and average nuclei area of all PCNSL patients. 10.1371/journal.pone.0161386.t001Table 1 Summary of age, sex and ADC values (ADCmin, ADCmean, ADCmax and deltaADC) of each lesion corresponding to the respective individuum measured as described above. ADCmin, ADCmean, ADCmax and deltaADCvalues displayed as × 10−4 mm2s-1. Sex Age (years) ADCmin (10−4 mm2s-1) ADCmax (10−4 mm2s-1) ADCmean (10−4 mm2s-1) deltaADC (10−4 mm2s-1) F 64 64 152 87 51 M 73 53 158 98 108 M 78 72 123 88 127 M 59 65 174 115 50 F 74 85 224 150 82 F 28 51 123 71 69 F 89 64 133 80 68 F 80 60 110 85 109 M 79 64 133 87 73 M 86 58 126 73 67 M 67 70 141 91 71 M 64 77 190 140 88 M 55 49 170 74 121 M 82 66 139 87 72 F 77 77 157 107 69 M 73 54 121 87 80 M 32 65 147 89 139 M 64 35 120 80 85 F 78 70 178 118 113 F 55 60 121 81 105 M 65 63 190 114 61 10.1371/journal.pone.0161386.t002Table 2 Synopsis of ADC Fractions and histopathological parameters. N Minimum Maximum Average Value Standard Deviation ADCmin (10−4 mm2s-1) 21 35.00 85.00 62.95 11.01 ADCmax (10−4 mm2s-1) 21 107.00 224.00 146.67 30.92 ADCmean (10−4 mm2s-1) 21 71.31 157.00 97.83 25.12 deltaADC (10−4 mm2s-1) 21 50.00 139.00 86.10 25.37 Nuclei count 21 319.00 1922.00 1288.62 366.69 Total nuclei area (μm2) 21 19988.01 216517.76 106617.71 44549.13 Average nuclei area (μm2) 21 53.20 267.91 86.52 46.41 In brief, the ranges of ADC values were as follows: ADCmin = 35.0–85.0 10−4 mm2s-1 (mean value 62.95x10-4 mm2s-1), ADCmean = 71.31–157.00x10-4 mm2s-1 (mean value 97.82x10-4 mm2s-1), ADC max = 107.00–224.00x10-4 mm2s-1 (mean value 147.66x10-4 mm2s-1) and deltaADC = 50.00–139.00x10-4 mm2s-1 (mean value 86.10x10-4 mm2s-1). Most lesions exhibited high proliferation rates ranging from 50% to 95% with a mean value of 76.19% Ki67 expressing nuclei. Fig 1c exemplarily shows Ki-67 staining of a primary CNS lymphoma. The immunostained section corresponds to the lymphoma shown in fig 1a and 1b. Total nuclei area was ranging widely from 19988.01 to 216517.76 μm2 with a mean value of 106617.71 μm2. Interestingly, average nuclei areas had a large range as well, varying from 53.20 to 267.91μm2 with a mean value of 86.52 μm2. Fig 1d exemplarily shows the particles tool analysis image corresponding to the lymphoma shown in fig 1a, 1b and 1c. Correlation analysis Table 3 summarizes the statistical analysis of the relation between histopathological parameters and ADC fractions, displaying the following findings: i) ADCmin showed a statistically significant correlation with Ki-67 expression (r = -0.454, p = 0.038). No correlations were found for ADCmin with nuclei count (r = -0.127, p = 0.584), total nuclei area (r = -0.228, p = 0.312) or average nuclei size (r = -0.173, p = 0.455). ii) ADCmean showed statistically significant correlations with Ki-67 expression (r = -0.546, p = 0.010) and total nuclei area (r = -0.500, p = 0.021). No correlations were found between ADCmean and nuclei count (r = -0.340, p = 0.132) or average nuclei size (r = -0.289, p = 0.204). iii) ADCmax showed a statistically significant correlation with Ki-67 expression (r = -0.515, p = 0.017). No correlations were found for ADCmax with nuclei count (r = -0.012, p = 0.960), total nuclei area (r = -0.155, p = 0.501) or average nuclei area (r = -0.144, p = 0.534). DeltaADC showed no statistically significant correlations with the investigated histopathological features (deltaADC and nuclei count: r = -0.223, p = 0.332; deltaADC and total nuclei area: r = -0.310, p = 0.172; deltaADC and average nuclei area: r = -0.212, p = 0.357), although a trend was delineable for deltaADC and Ki-67 (r = -0.428, p = 0.053). Fig 2 displays scatterplots graphically demonstrating the statistically significant relationships between DWI fractions and histopathological parameters. 10.1371/journal.pone.0161386.t003Table 3 Identified correlations between diffusion and histopathological parameters, statistically significant findings (p<0.05) are displayed in bold italics. Parameter Cell count Ki-67 (%) Total Nuclei Area (μm2) Average Nuclei Area *μm2) r = p = r = p = r = p = r = p = ADCmin (10-4mm2s-1) -0.127 0.584 -0.454 0.038 -0.228 0.321 -0.173 0.455 ADCmean (10-4mm2s-1) -0.340 0.132 - 0.546 0.010 -0.500 0.021 -0.289 0.204 ADCmax (10-4mm2s-1) 0.012 0.960 -0.515 0.017 -0.155 0.501 -.0144 0.534 deltaADC (104mm2s-1) -0.223 0.332 -0.428 0.053 -0.310 0.172 -0.212 0.357 10.1371/journal.pone.0161386.g002Fig 2 Significant correlations between ADC fractions (displayed as 10−4 mm2s-1) and histopathological parameters (Ki-67: positively stained nuclei displayed in % and total nuclei area: displayed in μm2). Statistical analysis documenting significant correlations between ADCmin and Ki-67 level (r = -0.454; p = 0.038) (a), ADCmean and Ki-67 level (r = -0.546; p = 0.010) (b), ADCmax and Ki-67 level (r = -0.515; p = 0.017). Significant correlations between ADCmean and total nuclei area (r = -0.500; p = 0.021) (d) are also shown. Discussion To the best of our knowledge the inverse correlation between different ADC fractions and proliferative activity of PCNSL has not been elucidated yet. The present study identified significant correlations between ADCmin, ADCmean and ADCmax measurements and cellular Ki-67 expression in biopsied portions of PCNSL, thus reflecting tumor biology in terms of proliferative activity. Furthermore, the study validated in vivo measured, low ADCmean values as a surrogate for tumor portions with increased nuclei area, reflecting increased cellularity. The significance of DWI for evaluation of PCNSL and differentiation from other brain lesions has grown considerably over the years. An early study by Cotton and colleagues investigated 9 biopsy-proven cases of cerebral lymphoma and found that the lesions were hyperintense in DWI with normal or decreased ADC values[13]. A later study by Zacharia and coworkers found that restricted diffusion with decreased ADC values is a consistent finding of PCNSL in immunocompetent patients prior treatment[8]. Haldorsen et al. discussed that DWI provides a potentially valuable tool to distinguish between glioma, metastases and PCNSL based on cellularity-related diffusion restriction[14]. Kickingereder and colleagues found that all ADC fractions (min, max and mean) were lower in PCNSL compared to glioblastoma[15]. Mabray and coworkers demonstrated that the combined use of ADC and CSF biomarkers (CXC chemokine ligand 13 and interleukin 10) increased the diagnostic performance for the diagnosis of PCNSL [16]. Although diffusion restriction is a consistent imaging finding in lymphoma, the question whether in vivo measured ADC values reflect cellularity in lymphoma has been answered differently in previous studies. For example, Wu and colleagues[9,11,17] showed that follicular lymphoma and diffuse large B-cell lymphoma did not exhibit statistically different ADC values, whereas the corresponding histopathological investigation revealed that the cellularity of both entities varied greatly in a statistically significant manner. Contrary to this, Guo et al.[9] showed that cellularity and ADC values correlated inversely in a statistically significant manner after investigation of high grade astrocytomas and PCNSL. The results of our study concur with the findings of Guo and colleagues, validating the link between decreased water diffusibility in lymphomas measurable in MRI based on high cellularity in histology. Values of ADC fractions in our study are approximately ten times larger than those given by Guo et al.[9] and Cotton et al.[13]. We attribute this difference to the fact that different MRI scanners and coils were used in these studies. Guo and colleagues used a 1.5T GE platform, Cotton and colleagues used a 1.5T Philips MRI scanner and we used a 1.5T Siemens scanner. As shown by Sasaki et al.[18] and Kivrak et al.[19], absolute ADC values can substantially vary among different coil systems, imagers, vendors, and magnetic field strengths. Although numerous studies provide evidence for the relationship of low ADC values and high cellularity, the question which ADC parameter—ADCmin, ADCmean or ADCmax best reflects the actual cellular density in a malignant tumor has not been answered conclusively. Chen and coworkers[17,20] conducted a meta analysis investigating the relationship between DWI and cellularity in different types of malignant tumors and found a strong, inverse correlation between ADC and cellularity especially in brain tumors including PCNSL. According to Chen et al. an overall of 14 studies examined the correlation of cellularity and restricted water diffusibility in brain tumors. Three of these studies showed that ADCmin only reflected increased cellularity. Ten out of fourteen studies showed that ADCmean only reflected increased cellularity. Only one study showed that both, ADCmin and ADCmean indicated high tumor cellularity. The fact that in few studies ADCmin alone reflected high cellularity, whereas in the majority of the studies ADCmean (but not ADCmin) indicated high cellularity raises the question as to how this discrepancy can be explained. Some studies did not reveal the exact method for ADC measurement, generally ADC values were measured based on ROIs corresponding to the whole contrast agent enhancing part of the malignancy. These values were correlated to the histologically derived data (e.g. cell count, nucleus-cytoplasm-ratio), which represented only the very small, biopsied portion of the tumor. Since most of the malignant tumors are composed of more aggressively infiltrating, rapidly dividing portions exhibiting high cellularity and less biologically active portions, potentially hemorrhage and necrosis as well, it is of great importance to only correlate the local apparent diffusion coefficient that corresponds to the biopsied, histologically examined part of the malignancy. As a consequence of the aforementioned we assume that the varying correlations being found between ADCmin and cellularity or ADCmean and cellularity, respectively, express differences in tumor heterogeneity between the different studies. Since we assume that ROIs, which extend to the whole contrast agent enhancing part of the tumor (including significantly more than the histologically investigated part) may result in rather unspecific ADC measurements, we identified the biopsied part of each PCNSL, measured the corresponding ADCmin, ADCmean and ADCmax values and correlated them with nuclei count, total nuclei area and average nuclei area. We found a conclusive, statistically significant correlation of ADCmean and total nuclei area, which concurs with the majority of the studies investigated by Chen and colleagues, indicating ADCmean as a good in vivo measurable surrogate for increased cellularity. In the second step this study revealed that ADCmin, ADCmean and ADCmax exhibit statistically significant inverse correlations with Ki-67 expression, thus providing evidence of low ADC values as in vivo surrogates for increased proliferative activity of PCNSL. This is in accordance with the work of Guzmán-De-Villoria et colleagues[21], who found a trend of correlation between low ADCmean values and Ki-67expression in glioma and other brain tumors. We assume that the relationship between ADC values and Ki-67 expression in gliomas is more difficult to prove, since the tissue architecture of glioma is more heterogeneous and the resulting ADC values exhibit a greater degree of variation compared to PCNSL. Other investigators were able to determine that increasing ADC values indicate positive therapeutic responses of tumors to antiproliferative treatment[22]. Further studies were able to show that low ADC values were good indicators for the efficacy of specific, proliferation-inhibitory drugs[23,24]. With reference to our findings we hypothesize that the antiproliferative effect of chemotherapeutic drugs results in a decrease of Ki-67 positive, proliferating cells and finally in a reduction of overall tumor cellularity—thus leading to a measurable increase of water diffusibility in MRI. This assumption is based on the results of Ho et al.[25] who were able to demonstrate that tumor areas with low ADC values exhibit increased glucose-uptake, thus being more metabolically active. As reported by Surov et al., different histopathological features are associated with different DWI parameters in meningioma[12]. Based on these findings and the results of our present study we hypothesize that—due to the comparatively homogeneous cellular architecture of CNS lymphoma—all ADC parameters in CNS lymphoma are influenced by the proliferative activity of the tumor. In a recently published study we were able to show that expression of the water channel Aquaporin 4 in meningioma correlates well with ADCmax values[26] We thereupon assume that in different entities (lymphoma, meningioma, glioma etc.), diverse histopathological and cyto-architectural features influence ADC parameters to a varying degree. Our study had several limitations. The main limitation of our study is the use of perfusion-sensitive ADC values. In the standardized clinical setting, DWI of the brain is conventionally performed with a low (b = 0mm2/s) and a high b-value (b = 1000mm2/s). Diffusion-weighted data acquired over a range of b values that includes low b values (< 100–150 s/mm2) are sensitive to signal attenuation from capillary perfusion[27]. Perfusion-insensitive diffusion weighted data can be obtained by using only higher b-values (>150 mm2/s)[27]. As shown by Priola et al., the use of perfusion-insensitive ADC measurements significantly improves diagnostic accuracy of DW-MRI[28]. Based on the fact that contrast enhancement was used as reference for ROI positioning, the ADC values obtained in our study may be overestimated due to perfusion effects[29,30]. Furthermore, the results of our study were possibly influenced by the small sample size and the retrospective design. Additionally, only a small portion of the contrast enhancing part of the mass was biopsied and investigated histopathologically, which does not reflect probable tumor heterogeneity. The applied technique for measuring ADC values was chosen because of its robustness and clinical feasibility. Conclusion This is the first study to show that ADCmin, ADCmean and ADCmax values are associated with high proliferation rates of PCNSL, which indicates a promising potential for ADC measurements as in vivo markers for treatment response or disease relapse. Furthermore, this study confirmed ADCmean as best in vivo surrogate of DWI for cellular density in PCNSL. ==== Refs References 1 Surawicz TS , McCarthy BJ , Kupelian V , Jukich PJ , Bruner JM , Davis FG . Descriptive epidemiology of primary brain and CNS tumors: results from the Central Brain Tumor Registry of the United States, 1990–1994 . Neuro-Oncology . 1999 ;1 : 14 –25 . 11554386 2 Batchelor T , Loeffler JS . Primary CNS lymphoma . J Clin Oncol . 2006 ;24 : 1281 –1288 . 10.1200/JCO.2005.04.8819 16525183 3 Olson JE , Janney CA , Rao RD , Cerhan JR , Kurtin PJ , Schiff D , et al The continuing increase in the incidence of primary central nervous system non-Hodgkin lymphoma . Cancer . 2002 ;95 : 1504 –1510 . 10.1002/cncr.10851 12237919 4 Galea I , Bechmann I , Perry VH . What is immune privilege (not)? Trends in Immunology . 2007 ;28 : 12 –18 . 10.1016/j.it.2006.11.004 17129764 5 He X , Chen Z , Fu T , Jin X , Yu T , Liang Y , et al Ki-67 is a valuable prognostic predictor of lymphoma but its utility varies in lymphoma subtypes: evidence from a systematic meta-analysis . BMC Cancer . BMC Cancer; 2014 ;14 : 1 –13 . 10.1186/1471-2407-14-153 24383403 6 Schlüter C , Duchrow M , Wohlenberg C . The cell proliferation-associated antigen of antibody Ki-67: a very large, ubiquitous nuclear protein with numerous repeated elements, representing a new kind of cell …. The Journal of cell …. 1993 ;: 1 –10 . 7 Hanahan D , Weinberg RA . Hallmarks of Cancer: The Next Generation . Cell . Elsevier Inc ; 2011 ;144 : 646 –674 . 10.1016/j.cell.2011.02.013 21376230 8 Zacharia TT , Law M , Naidich TP , Leeds NE . Central Nervous System Lymphoma Characterization by Diffusion-Weighted Imaging and MR Spectroscopy . Journal of Neuroimaging . 2008 ;18 : 411 –417 . 10.1111/j.1552-6569.2007.00231.x 18494774 9 Guo AC , Cummings TJ , Dash RC , Provenzale JM . Lymphomas and High-Grade Astrocytomas: Comparison of Water Diffusibility and Histologic Characteristics1 . Radiology . 2002 ;224 : 177 –183 . 10.1148/radiol.2241010637 12091680 10 Barajas RF , Rubenstein JL , Chang JS , Hwang J , Cha S . Diffusion-Weighted MR Imaging Derived Apparent Diffusion Coefficient Is Predictive of Clinical Outcome in Primary Central Nervous System Lymphoma . AJNR Am J Neuroradiol . 2010 ;31 : 60 –66 . 10.3174/ajnr.A1750 19729544 11 Wu X , Pertovaara H , Dastidar P , Vornanen M , Paavolainen L , Marjomäki V , et al European Journal of Radiology . European Journal of Radiology. Elsevier Ireland Ltd ; 2013 ;82 : e158 –e164 . 10.1016/j.ejrad.2012.11.021 12 Surov A , Caysa H , Wienke A , Spielmann RP , Fiedler E . Correlation Between Different ADC Fractions, Cell Count, Ki-67, Total Nucleic Areas and Average Nucleic Areas in Meningothelial Meningiomas . Anticancer Res . 2015 ;35 : 6841 –6846 .26637905 13 Cotton F , Ongolo-Zogo P , Louis-Tisserand G , Streichenberger N , Hermier M , Jouvet A , et al Diffusion and perfusion MR imaging in cerebral lymphomas . J Neuroradiol . 2006 ;33 : 220 –228 . 17041526 14 Haldorsen IS , Espeland A , Larsson EM . Central Nervous System Lymphoma: Characteristic Findings on Traditional and Advanced Imaging . American Journal of Neuroradiology . 2011 ;32 : 984 –992 . 10.3174/ajnr.A2171 20616176 15 Kickingereder P , Wiestler B , Sahm F , Heiland S , Roethke M , Schlemmer H-P , et al Primary Central Nervous System Lymphoma and Atypical Glioblastoma: Multiparametric Differentiation by Using Diffusion-, Perfusion-, and Susceptibility-weighted MR Imaging . Radiology . 2014 ;272 : 843 –850 . 10.1148/radiol.14132740 24814181 16 Mabray MC , Barajas RF , Villanueva-Meyer JE , Zhang CA , Valles FE , Rubenstein JL , et al The Combined Performance of ADC, CSF CXC Chemokine Ligand 13, and CSF Interleukin 10 in the Diagnosis of Central Nervous System Lymphoma . AJNR Am J Neuroradiol . 2015 ;37 : 74 –79 . 10.3174/ajnr.A4450 26381553 17 Chen L , Liu M , Bao J , Xia Y , Zhang J , Zhang L , et al The Correlation between Apparent Diffusion Coefficient and Tumor Cellularity in Patients: A Meta-Analysis . Hess CP , editor. PLoS ONE . 2013 ;8 : e79008 10.1371/journal.pone.0079008.s001 24244402 18 Sasaki M , Yamada K , Watanabe Y , Matsui M , Ida M , Fujiwara S , et al Variability in absolute apparent diffusion coefficient values across different platforms may be substantial: a multivendor, multi-institutional comparison study . Radiology . 2008 ;249 : 624 –630 . 10.1148/radiol.2492071681 18936317 19 Kıvrak AS , Paksoy Y , Erol C , Koplay M , Özbek S , Kara F . Comparison of apparent diffusion coefficient values among different MRI platforms: a multicenter phantom study . Diagn Interv Radiol . 2012 ;19 : 433 –437 . 10.5152/dir.2013.13034 20 Chen L , Zhang J , Chen Y , Wang W , Zhou X , Yan X , et al Relationship between apparent diffusion coefficient and tumour cellularity in lung cancer . PLoS ONE . 2014 ;9 : e99865 10.1371/journal.pone.0099865 24919055 21 GuzmŁn-De-Villoria JA . Added value of advanced over conventional magnetic resonance imaging in grading gliomas and other primary brain tumors . 2015 ;: 1 –10 . 10.1186/s40644-014-0035-8 22 Chen Y , Zhong J , Wu H , Chen N . The clinical application of whole-body diffusion-weighted imaging in theearly assessment of chemotherapeutic effects in lymphoma:the initial experience . Magnetic Resonance Imaging . Elsevier Inc ; 2012 ;30 : 165 –170 . 10.1016/j.mri.2011.09.019 22133288 23 Jeon TY , Kim CK , Kim J-H , Im GH , Park BK , Lee JH . Assessment of early therapeutic response to sorafenib in renal cell carcinoma xenografts by dynamic contrast-enhanced and diffusion-weighted MR imaging . BJR . 2015 ;88 : 20150163 10.1259/bjr.20150163 26133222 24 Demir S , Altinkaya N , Kocer NE , Erbay A , Oguzkurt P . Variations in apparent diffusion coefficient values following chemotherapy in pediatric neuroblastoma . Diagn Interv Radiol . 2015 ;21 : 184 –188 . 10.5152/dir.2014.14187 25519453 25 Ho K-C , Lin G , Wang J-J , Lai C-H , Chang C-J , Yen T-C . Correlation of apparent diffusion coefficients measured by 3T diffusion-weighted MRI and SUV from FDG PET/CT in primary cervical cancer . Eur J Nucl Med Mol Imaging . 2008 ;36 : 200 –208 . 10.1007/s00259-008-0936-5 18779960 26 Schob S , Surov A , Wienke A , Meyer HJ , Spielmann RP , Fiedler E . Correlation Between Aquaporin 4 Expression and Different DWI Parameters in Grade I Meningioma . Mol Imaging Biol . 2016 10.1007/s11307-016-0978-1 27 Koh D-M , Collins DJ . Diffusion-Weighted MRI in the Body: Applications and Challenges in Oncology . American Journal of Roentgenology . American Roentgen Ray Society ; 2007 188 : 1622 –1635 . 10.2214/AJR.06.1403 17515386 28 Priola AM , Priola SM , Gned D , Piacibello E , Sardo D , Parvis G , et al Diffusion-weighted quantitative MRI to diagnose benign conditions from malignancies of the anterior mediastinum: Improvement of diagnostic accuracy by comparing perfusion-free to perfusion-sensitive measurements of the apparent diffusion coefficient . J Magn Reson Imaging . 2016 10.1002/jmri.25203 29 Padhani AR , Liu G , Mu-Koh D , Chenevert TL , Thoeny HC , Takahara T , et al Diffusion-Weighted Magnetic Resonance Imaging as a Cancer Biomarker: Consensus and Recommendations . Neoplasia . 2009 ;11 : 102 –125 . 10.1593/neo.81328 19186405 30 Priola AM , Gned D , Veltri A , Priola SM . Chemical shift and diffusion-weighted magnetic resonance imaging of the anterior mediastinum in oncology: Current clinical applications in qualitative and quantitative assessment . Crit Rev Oncol Hematol . 2016 ;98 : 335 –357 . 10.1016/j.critrevonc.2015.11.012 26653574

PMC005xxxxxx/PMC5003363.txt

==== Front PLoS PathogPLoS PathogplosplospathPLoS Pathogens1553-73661553-7374Public Library of Science San Francisco, CA USA 2757142210.1371/journal.ppat.1005856PPATHOGENS-D-16-01025Research ArticleBiology and Life SciencesMicrobiologyMicrobial MutationBiology and Life SciencesGeneticsGenomicsMicrobial GenomicsViral GenomicsBiology and Life SciencesMicrobiologyMicrobial GenomicsViral GenomicsBiology and Life SciencesMicrobiologyVirologyViral GenomicsBiology and Life SciencesGeneticsMutationPoint MutationBiology and Life SciencesComputational BiologyGenome AnalysisGenomic LibrariesBiology and Life SciencesGeneticsGenomicsGenome AnalysisGenomic LibrariesBiology and life sciencesOrganismsVirusesRNA virusesOrthomyxovirusesInfluenza virusesInfluenza A virusBiology and life sciencesMicrobiologyMedical microbiologyMicrobial pathogensViral pathogensOrthomyxovirusesInfluenza virusesInfluenza A virusMedicine and health sciencesPathology and laboratory medicinePathogensMicrobial pathogensViral pathogensOrthomyxovirusesInfluenza virusesInfluenza A virusBiology and life sciencesOrganismsVirusesViral pathogensOrthomyxovirusesInfluenza virusesInfluenza A virusBiology and life sciencesOrganismsVirusesRNA virusesOrthomyxovirusesInfluenza VirusesBiology and Life SciencesMicrobiologyMedical MicrobiologyMicrobial PathogensViral PathogensOrthomyxovirusesInfluenza VirusesMedicine and Health SciencesPathology and Laboratory MedicinePathogensMicrobial PathogensViral PathogensOrthomyxovirusesInfluenza VirusesBiology and Life SciencesOrganismsVirusesViral PathogensOrthomyxovirusesInfluenza VirusesBiology and Life SciencesGeneticsMutationSubstitution MutationBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesTransfectionResearch and Analysis MethodsMolecular Biology TechniquesTransfectionThe Mutational Robustness of Influenza A Virus Influenza Mutational Fitness Effectshttp://orcid.org/0000-0003-3984-4748Visher Elisa 1 http://orcid.org/0000-0002-3274-9230Whitefield Shawn E. 2 McCrone John T. 2 http://orcid.org/0000-0002-0997-1108Fitzsimmons William 1 Lauring Adam S. 1 2 * 1 Division of Infectious Diseases, Department of Internal Medicine, University of Michigan, Ann Arbor, Michigan, United States of America 2 Department of Microbiology and Immunology, University of Michigan, Ann Arbor, Michigan, United States of America Ferguson Neil M. Editor Imperial College London, UNITED KINGDOM The authors have declared that no competing interests exist. Conceived and designed the experiments: EV SEW JTM WF ASL. Performed the experiments: EV SEW JTM WF ASL. Analyzed the data: EV SEW JTM WF ASL. Contributed reagents/materials/analysis tools: EV SEW JTM WF ASL. Wrote the paper: EV ASL. * E-mail: alauring@med.umich.edu29 8 2016 8 2016 12 8 e10058569 5 2016 10 8 2016 © 2016 Visher et al2016Visher et alThis is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.A virus’ mutational robustness is described in terms of the strength and distribution of the mutational fitness effects, or MFE. The distribution of MFE is central to many questions in evolutionary theory and is a key parameter in models of molecular evolution. Here we define the mutational fitness effects in influenza A virus by generating 128 viruses, each with a single nucleotide mutation. In contrast to mutational scanning approaches, this strategy allowed us to unambiguously assign fitness values to individual mutations. The presence of each desired mutation and the absence of additional mutations were verified by next generation sequencing of each stock. A mutation was considered lethal only after we failed to rescue virus in three independent transfections. We measured the fitness of each viable mutant relative to the wild type by quantitative RT-PCR following direct competition on A549 cells. We found that 31.6% of the mutations in the genome-wide dataset were lethal and that the lethal fraction did not differ appreciably between the HA- and NA-encoding segments and the rest of the genome. Of the viable mutants, the fitness mean and standard deviation were 0.80 and 0.22 in the genome-wide dataset and best modeled as a beta distribution. The fitness impact of mutation was marginally lower in the segments coding for HA and NA (0.88 ± 0.16) than in the other 6 segments (0.78 ± 0.24), and their respective beta distributions had slightly different shape parameters. The results for influenza A virus are remarkably similar to our own analysis of CirSeq-derived fitness values from poliovirus and previously published data from other small, single stranded DNA and RNA viruses. These data suggest that genome size, and not nucleic acid type or mode of replication, is the main determinant of viral mutational fitness effects. Author Summary Like other RNA viruses, influenza virus has a very high mutation rate. While high mutation rates may increase the rate at which influenza virus will adapt to a new host, acquire a new route of transmission, or escape from host immune surveillance, data from model systems suggest that most new viral mutations are either lethal or highly detrimental. Mutational robustness refers to the ability of a virus to tolerate, or buffer, these mutations. The mutational robustness of a virus will determine which mutations are maintained in a population and may have a greater impact on viral evolution than mutation rate. We defined the mutational robustness of influenza A virus by measuring the fitness of a large number of viruses, each with a single point mutation. We found that the overall robustness of influenza was similar to that of poliovirus and other viruses of similar size. Interestingly, mutations appeared to be more easily accommodated in hemagglutinin and neuraminidase than elsewhere in the genome. This work will inform models of influenza evolution at the global and molecular scale. Doris Duke Charitable Foundation (US)CSDA 2013105Lauring Adam S. http://dx.doi.org/10.13039/100000060National Institute of Allergy and Infectious DiseasesAI118886Lauring Adam S. http://dx.doi.org/10.13039/100000057National Institute of General Medical SciencesT32GM007544McCrone John T. This work was supported by a Clinician Scientist Development Award from the Doris Duke Charitable Foundation (CSDA 2013105) and NIH R01 AI118886, both to ASL. JTM was supported by the Michigan Predoctoral Training Program in Genetics (T32GM007544). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction The predictable burden of seasonal influenza and the unpredictability of the next pandemic are attributable in large part to the rapid evolution of influenza virus [1–4]. Like other RNA viruses, influenza viruses replicate with extremely low fidelity, with a mutation rate of roughly 2 x 10−5 substitutions per nucleotide copied per cellular infection [5–7]. Influenza viruses also undergo reassortment of their genomic segments, a combinatorial exchange of genetic material analogous to recombination in other RNA viruses [8,9]. Together, low replicative fidelity and frequent reassortment allow influenza virus populations to generate significant diversity. This capacity may allow influenza viruses to maintain, or to quickly generate, the requisite mutations that mediate cross species transmission, escape from neutralizing antibody, or drug resistance [10]. The focus on mutation as a driving force in viral evolution has tended to downplay the tremendous fitness cost of mutation [11,12]. Here, we define viral fitness as the capacity of an individual, or population, to generate infectious progeny. Most mutations have deleterious effects on fitness, which suggests that mutational tolerance may play a significant role in determining the genetic diversity that can be maintained within a population [13]. Mutational robustness refers to phenotypic stability in the face of mutation [14–16]. High mutation rates select for increased mutational robustness [17], and a more robust population can increase its genetic diversity without a dramatic alteration in mean fitness. A virus’ intrinsic robustness may influence its fitness in vitro and virulence in vivo [18]. The impact of individual mutations on viral fitness is the mutational fitness effect (MFE). A virus’ mutational robustness is described in terms of the strength and distribution of the MFE [19]. Together with mutation rate, the MFE governs many aspects of evolution including: the relative importance of selection vs. genetic drift, the efficiency of selection and adaptation, the impact of recombination (or reassortment), and the role of epistasis in fixing new and beneficial mutations. It is therefore essential for accurate models of molecular evolution [20,21]. A virus’ sensitivity to mutation may also determine the effectiveness of lethal mutagenesis [22–24]. Early studies of mutational fitness effects relied on mutation accumulation (MA) experiments, where the imposition of extreme bottlenecks propagates and fixes newly generated mutations by drift as opposed to selection [25–29]. An alternative approach commonly used in RNA viruses is to accelerate mutation accumulation by passaging virus in the presence of mutagenic drugs [30–32]. While both methods provide valuable information about the average fitness impact of random mutation, uncertainty about the number of mutations per clone and the fitness effect of each individual mutation makes it difficult to accurately model a distribution. These issues also complicate more recent, high throughput methods based on next generation sequencing [33–35]. Furthermore, none of these approaches provide accurate estimates of the fraction of mutations that are lethal to the virus. A less exhaustive, but more controlled assay for MFE is to measure the fitness of a set of viral clones, each with a randomly selected single nucleotide substitution. This approach unambiguously assigns fitness values, or selection coefficients, to individual mutations. It is also highly quantitative and allows for estimation of the lethal fraction. The first such study, of vesicular stomatitis virus (VSV), found that over 90% of random single nucleotide mutations reduce replicative fitness and 40% are lethal in this negative sense RNA virus [13]. Subsequent work by Sanjuan, Elena, and colleagues found somewhat similar distributions of MFE in f1 (ssDNA phage), phiX 174 (ssDNA phage), QB (+ssRNA phage), and tobacco etch virus (+ssRNA virus) [19,36–40]. Together these data suggest that despite their differences in genome organization and replication strategy, ssRNA and ssDNA viruses are equally sensitive to mutation. Despite their importance to pathogen evolution, all available genome-wide studies of viral MFE have been performed in evolutionary model systems. Here we characterize the distribution of MFE in influenza A virus, a segmented negative sense RNA virus whose evolutionary dynamics are important to global health. We used site directed mutagenesis to generate a library of plasmids encoding 128 influenza A viruses, each with a single point mutation in the A/WSN33/H1N1 genetic background. The large number of mutants allowed us to define the MFE across the genome and to compare the lethal fraction and MFE between segments coding for the surface proteins to those coding for the internal proteins. We find that the MFE of influenza A are remarkably similar to those of other viruses with varying genome structure. While similar proportions of mutations in the surface proteins and internal proteins were lethal, the average impact of mutation appeared to be less deleterious in HA and NA. Our results suggest that the size and compactness of a virus’ genome, and not necessarily the genomic nucleic acid or mode of replication, are major determinants of its intrinsic mutational robustness. Results Our primary goal was to determine the distribution of mutational fitness effects across the influenza A genome. We generated all single nucleotide mutations in the commonly used laboratory strain, A/WSN/33/H1N1, hereafter referred to as WSN33 or wild type (WT) [41]. In WSN33, the 8 genomic segments range in size from 0.9 to 2.3 kb. We planned to make 149 mutants and grouped them into two libraries–“genome-wide” and “comparison”–of pre-specified size and composition. Of the 149 total mutations we attempted (S1 Table), we successfully generated 128 (86%). For our “genome-wide” library, we reasoned that in order to achieve an unbiased distribution of mutations throughout the genome, our library should contain a number of mutations on each segment that is proportional to the size of each segment—PB2 17.2%, PB1 17.2%, PA 16.4%, HA 13.1%, NP 11.5%, NA 10.4%, M 7.5%, NS 6.5%, of the total genome respectively. We used a custom R script to choose randomly the nucleotide position and substitution type in accordance with this distribution. For our unbiased genome-wide analysis (n = 95), we generated 14 (14.7%) PB2, 16 (16.8%) PB1, 16 (16.8%) PA, 14 (14.7%) HA, 12 (12.6%) NP, 10 (10.5%) NA, 8 (8.4%) M and 5 (5.3%) NS mutations. Though we failed to generate 14% of attempted mutations, this did not significantly alter the distribution of the mutations across the eight segments (S1A Fig). Viral surface proteins that are targeted by the immune system often exhibit greater sequence diversity than internal structural and enzymatic proteins. In many cases, the relationship of this diversity to the intrinsic mutational robustness of the genes encoding these surface proteins is unknown. Therefore, a secondary goal of our study was to compare the distribution of mutational fitness effects for the HA- and NA-encoding segments to the other 6 segments, which code for internal proteins. We improved our power to detect a difference by generating an additional 18 HA and 15 NA mutations. In this aggregate “comparison” library (n = 128), 45% (n = 57) of our library is contained on segments coding for HA and NA and 55% (n = 71) on the other 6 segments. (S1B Fig). The entire data set includes 38 transition mutations and 90 transversion mutations, which is in the range of what one would expect by chance (Fisher test, p = 0.75). Identification of lethal mutations Previous work indicates that a substantial, but varying proportion, of mutations in RNA viruses are lethal. Accurate assessment of this lethal fraction is an essential, and non-trivial, task. In a reverse genetic approach, the efficiency of transfection and viral recovery are key parameters, since a failed transfection and a lethal mutation will both give supernatants with undetectable titers. We used two approaches to address this problem. In the first, we estimated the transfection failure rate and calculated the probability of miscalling a viable mutant as lethal. Over the course of this study, we transfected the WT virus 19 individual times, and never failed to recover virus in our P0 supernatants. As in [13], we quantified the expected probability of a transfection failure as if we had observed a single WT transfection failure. This conservative approach gave a per transfection failure rate of < 5.26%. Because we attempted three independent transfections for each candidate lethal, the likelihood of repeated transfection failure is 0.05263, or 1.46 x10-4. In a set of 128 viruses, we would expect to falsely identify fewer than one (128 * 1.46 x 10−4 = 0.019) virus as lethal. This probabilistic model assumes equal transfection efficiency for WT and mutant viruses. We also tested one of our mutants, PB2_14 (PB2 C532A), which had moderately reduced fitness (0.81, see below). Here too, we successfully recovered virus in 19 individual transfections. We also observed several cases of mutants with undetectable titers at P0 and a moderate P1 titer after blind passage of the supernatant from transfected cells. We characterized these P1 stocks by next generation sequencing, and in all cases, found either reversion of the introduced mutation or contamination of the culture by other viruses that were transfected or passaged on the same day. We therefore considered a mutation to be lethal if we had undetectable virus in three independent transfections or if we were unable to recover the mutation in a P1 stock after blind passage of the P0 stock. Using these criteria, there were 30 lethal mutations (31.6%) in the genome-wide dataset and 38 lethal mutations in the combined dataset (29.7%, Table 1). The lethal fraction did not differ appreciably between either the HA/NA encoding segments and the rest of the genome (16/57 vs. 22/71; Fisher exact test, p = 0.85) or the HA encoding segment and the rest of the genome (7/32 vs. 31/96; Fisher exact test, p = 0.37). We had 62% to detect a two fold difference in the lethal fraction for HA/NA and 52% power to detect a two fold difference in the lethal fraction for HA. No synonymous mutations were lethal, but 2 non-coding mutations and a stop codon loss were lethal. 10.1371/journal.ppat.1005856.t001Table 1 Lethal mutations. Segment Mutation Amino Acid Dataset Clone ID 1 (PB2) C839A T271K Genomewide PB2-10 U1559G V511G Genomewide PB2-2 U2305C Stop reversion Genomewide PB2-1 2 (PB1) U675A Y217stop Genomewide PB1-3 A728U K235M Genomewide PB1-15 C1123U Q367stop Genomewide PB1-14 U1232G L403W Genomewide PB1-9 U1268G L415stop Genomewide PB1-16 G1581U E519D Genomewide PB1-7 3 (PA) A263U E80V Genomewide PA-12 G529C A169P Genomewide PA-11 C1324A P434T Genomewide PA-14 G1634C W537S Genomewide PA-5 C1682A A553D Genomewide PA-4 4 (HA) U55G L8R Genomewide HA-11 U799G I256R Genomewide HA-12 U915C C295R HA/NA HA-25 C1137A H369N HA/NA HA-32 A1143U N371Y Genomewide HA-5 G1153A G374E Genomewide HA-16 A1345C N438T HA/NA HA-24 5 (NP) A151U I36F Genomewide NP-6 U521C M159T Genomewide NP-2 C1194A S383R Genomewide NP-4 A1505G Y487C Genomewide NP-10 6 (NA) U334G F105L HA/NA NA-24 G422A D135N HA/NA NA-13 A449C S144R HA/NA NA-25 U668A C217S HA/NA NA-31 U693C I225T Genomewide NA-8 G968U G317C Genomewide NA-6 G1313U V432L Genomewide NA-7 U1340G W441G Genomewide NA-2 U1400C Non-Coding HA/NA NA-28 7 (M) U506C M1 S161P Genomewide M-8 G551A M1 E176K Genomewide M-6 G936U M2 E75stop Genomewide M-3 A1009G Non-Coding Genomewide M-5 Clonality of viable mutants As in other studies of viral mutational fitness effects, we measured the fitness of P1 viral stocks rather than the initial transfection supernatant. Given the error prone replication of RNA viruses, we considered it possible that second-site mutations might accumulate in the short time between transfection and completion of the first passage. This might be exaggerated in the less fit viruses, since there would be strong positive selection of a compensatory mutation. If a compensatory mutation swept through the population, it would confound assignment of a fitness value to the initial mutation. Therefore, we sequenced the entire genome of the P1 stocks for 78 of our mutants on the Illumina platform to identify any second site mutations and their frequency. In all but 3 cases, the desired mutation was present at >94%, and was present at >99% in nearly all. Very few additional mutations were identified at >2% (Table 2, S2 Table). We relied on Sanger sequencing alone to confirm the desired mutation in replicate stocks of the remaining 13 viable viruses. We conclude that these stocks are essentially clonal and that the fitness of these populations will reflect the impact of each individual mutation. 10.1371/journal.ppat.1005856.t002Table 2 Next generation sequencing of viral stocks. Mutant Mutation Amino Acid Mutation > 95% Secondary Mutations (frequency) Dataset HA-10 T1599A S523T Yes PA-A95G (5.0) Genomewide HA-21 A334T N101I Yes * - * Genomewide HA-30 A648C S206R 94.7 HA-T1583G (4.6); HA-A534G (2.7) HA/NA HA-46 C231G L67V Yes * - * HA/NA M-1 G661C M212I Yes M-T117G (3.0) Genomewide M-7 C174G C174G Yes * - * Genomewide NA-14 G98A G27R Yes PB2-G2086T (2.5) HA/NA NA-20 A909C K297T Yes PB2-T2156A (2.8) HA/NA NA-9 G1355A G1355A 0.75 NS-C684T (2.7) Genomewide NP-8 A454C A454C Yes PB2-A233G (2.5) Genomewide NS-2 G227T R76R Yes PB2-A235G (2.7) Genomewide PA-1 C500T A159V 92.6 PA-C500G (5.2); PB2-A913G (3.3); PB2-C1919T (2.9) Genomewide PA-6 G240A L72L 92.5 - Genomewide PA-7 A1358T Y445F Yes PB2-T1627C (3.7) Genomewide PB1-11 A2187T R721S Yes PA-T1469C (2.0) Genomewide PB2-8 C839A T271K Yes * - * Genomewide * Sanger sequence only Measurement of viral fitness In prior studies of MFE, fitness has been measured as the difference in exponential growth rates for WT and mutant strains, measured either in parallel or in direct competition[19,36,37,39,42]. Given the relative imprecision of one step growth curves for quantification of growth parameters in influenza and many other viruses, we measured the relative fitness of each mutant in direct competition with the WT over serial passage [18]. In this assay, the mutant is competed with a tagged WT reference, which has a cluster of synonymous mutations in the PB1 open reading frame. These mutations allowed us to distinguish the barcoded WT from a non-barcoded mutant in a mixed infection using quantitative reverse transcription PCR with primers specific for their respective sequences. Importantly, this tagged WT virus competed equally well with the untagged WT virus over 6 passages, demonstrating the selective neutrality of the marker (Fig 1A). In our serial passage competition assay, the change in relative frequency of the WT and mutant over time is the difference in growth rate, or the selection rate constant (Fig 1B). The exponent of this value is the relative fitness. We have shown previously that this assay can provide precise measurements of relative fitness with as few as 3 technical replicates [18], although it is less sensitive for weakly deleterious or beneficial mutations with a relative fitness close to 1. 10.1371/journal.ppat.1005856.g001Fig 1 Direct competition assay for relative fitness. (A) Equal infectious units of a barcoded version of the WT were competed against WT at an moi of 0.01, and the amount of each virus at each passage was compared to the input by RT-qPCR as described in the methods. The slope of the regression of the difference in the log10 change in ratio for each virus over time is the fitness. The assay was performed in triplicate and the slopes of the three lines are 0.007, 0.129, and 0.007, which corresponds to a fitness of 1.02 ± 0.008 for the barcoded virus relative to WT. (B) Sample data for two single nucleotide mutants. Each was competed against the barcoded WT as in (A) and relative fitness measured as calculated in the methods. One replicate each of NA-18 (circles, fitness = 0.64) and HA-40 (squares, fitness = 1.02) is shown. Note that we fit our regressions through passages 1–4 and excluded P0 as slight deviations from a 1:1 ratio of the two viruses in the inoculum can skew the slope when fit through this data point. We performed our competition assays in A549, a cancerous human lung epithelial cell line, which supported efficient replication of WSN33 and may be more physiologically relevant to influenza virus replication than other commonly used lines. Passages were carried out at a multiplicity of 0.01 infectious units (TCID50) per cell. Given that the burst size of influenza A in A549 cells is <100 per cell, most passages will represent 2 cellular infection cycles with minimal co-infection in the second cycle. We note that this moi corresponded to just 10,000 infectious units at each passage, and we cannot exclude that genetic drift could lead to some variability in the fitness measurements with a transfer population of this size [43]. Distribution of mutational fitness effects We were able to measure the fitness of 89 out of the 90 viable mutants. Despite repeated attempts with multiple stocks, we were unable to obtain data for HA-8. In competitions with this mutant, we consistently saw large reductions in titer of both HA-8 and the wild type after a single passage, perhaps due to dominant negative effects or the impact of defective interfering particles. With the exception of PB1-6, which was measured twice, all fitness values are based on 3 replicate competition assays, and the mean fitness and standard deviation are reported in Table 3 (see also Figs 2 and 3). Seventy-one of the mutations in our library were nonsynonymous, and these viruses had relative fitness values ranging from 0.26 to 1.12 (mean 0.80, standard deviation 0.20). Nearly all of the 18 viruses with synonymous mutations exhibited a fitness close to 1 (mean 0.93, standard deviation 0.21). The sole exception was PA-6/G240A, which is synonymous in the canonical PA open reading frame (Leu72). Neither this mutation, nor any of the other PA mutations were within the alternate PA-X reading frame [44]. Three mutations affected proteins in two different reading frames: PB1 A251C (fitness 0.45), NS G648U (fitness 0.92), and NS G650C (fitness 0.78). Three out of the 4 nonsense mutations were lethal (see Table 1), the exception was NS-5/G648T (NS1 E208stop, NEP M50I), which had a fitness of 0.92. Consistent with the important role of 5’ and 3’ translated regions in RNA synthesis and packaging, 2 out of the 3 non-coding mutations were lethal [45]. The 3rd was HA-8 (see above). 10.1371/journal.ppat.1005856.t003Table 3 Fitness values of viable mutants. Segment Mutation Amino Acid Change Fitness Mean Fitness SD Dataset Clone ID 1 (PB2) U306A P93 (SYN) 0.90 0.13 Genomewide PB2-16 A440U Q138L 0.78 0.18 Genomewide PB2-12 C532A P169T 0.81 0.04 Genomewide PB2-14 C880U H285Y 0.96 0.07 Genomewide PB2-4 A1167U R380S 0.94 0.06 Genomewide PB2-11 U1251C D408 (SYN) 0.89 0.05 Genomewide PB2-3 A1495C S490R 0.96 0.18 Genomewide PB2-8 U1527A R500 (SYN) 0.90 0.04 Genomewide PB2-5 G1660C V545L 0.94 0.04 Genomewide PB2-7 A1854G T609 (SYN) 1.03 0.05 Genomewide PB2-15 A2113C I696L 0.73 0.14 Genomewide PB2-13 2 (PB1) A251C PB1 D76A, PB1-F2 T45P 0.45 0.01* Genomewide PB1-6 C522U F166 (SYN) 1.10 0.15 Genomewide PB1-2 C549U N175 (SYN) 1.01 0.24 Genomewide PB1-1 A581C Q186P 0.30 0.01 Genomewide PB1-10 G599A R192K 0.99 0.09 Genomewide PB1-5 U1288A S422T 0.95 0.03 Genomewide PB1-13 A1322C K433T 0.48 0.10 Genomewide PB1-12 G1764U W580C 0.26 0.00 Genomewide PB1-4 A2187U R721S 0.34 0.09 Genomewide PB1-11 A2277C E751D 0.65 0.09 Genomewide PB1-8 3 (PA) A88G K22E 0.81 0.12 Genomewide PA-8 A92G E23G 0.58 0.05 Genomewide PA-10 U237A L71 (SYN) 1.11 0.13 Genomewide PA-9 G240A L72 (SYN) 0.17 0.07 Genomewide PA-6 C500U A159V 0.85 0.11 Genomewide PA-1 U878G M285R 0.62 0.14 Genomewide PA-3 U964G F314V 0.36 0.13 Genomewide PA-13 G1041C K339N 0.90 0.04 Genomewide PA-2 A1358U Y445F 0.82 0.15 Genomewide PA-7 U1685C I554T 0.89 0.08 Genomewide PA-15 U2123C V700A 0.89 0.10 Genomewide PA-16 4 (HA) C231G L67V 0.71 0.03 HA/NA HA-46 A334U N101I 0.97 0.06 Genomewide HA-21 C368A L112 (SYN) 0.93 0.08 Genomewide HA-13 U408A S126T 0.95 0.16 HA/NA HA-37 G542U K170N 1.01 0.12 HA/NA HA-36 A648C S206R 0.72 0.14 HA/NA HA-30 A699U N223Y 1.03 0.10 HA/NA HA-38 A784G E251G 0.94 0.04 Genomewide HA-22 C822U L264 (SYN) 1.00 0.06 HA/NA HA-26 A939U N303Y 0.80 0.13 HA/NA HA-41 G1006U S325I 0.83 0.08 HA/NA HA-45 A1050C I340L 0.87 0.08 HA/NA HA-43 A1057U Y342F 0.87 0.19 Genomewide HA-1 A1174U Q381L 0.93 0.06 HA/NA HA-27 C1229G I399M 0.85 0.10 Genomewide HA-18 A1264C K411T 0.65 0.20 HA/NA HA-31 G1292A M420I 1.12 0.07 HA/NA HA-33 U1299C L423 (SYN) 1.13 0.27 HA/NA HA-39 U1466A N478K 0.64 0.14 Genomewide HA-3 A1512G S494G 0.87 0.04 HA/NA HA-42 U1583G D517E 1.00 0.02 HA/NA HA-40 G1587C V519L 1.06 0.24 Genomewide HA-15 U1599A S523T 0.88 0.13 Genomewide HA-10 C1696A S555Y 0.73 0.01 Genomewide HA-20 A1749C noncoding ND ND Genomewide HA-8 5 (NP) U198G D51E 0.88 0.06 Genomewide NP-1 A425G D127G 1.01 0.07 Genomewide NP-9 G436U A131S 0.76 0.11 Genomewide NP-5 A454C M137L 0.60 0.05 Genomewide NP-8 C485U T147I 1.00 0.28 Genomewide NP-7 A1160U E372V 0.77 0.10 Genomewide NP-11 A1229U N395I 0.99 0.09 Genomewide NP-12 C1485A D480E 0.99 0.04 Genomewide NP-3 6 (NA) G98A G27R 0.73 0.03 HA/NA NA-14 C109A I30 (SYN) 1.04 0.24 HA/NA NA-30 G158A G47R 0.94 0.20 HA/NA NA-22 A176C S53R 0.90 0.11 HA/NA NA-21 G201C G61A 0.98 0.03 Genomewide NA-4 C454U C145 (SYN) 1.08 0.11 HA/NA NA-26 C700U T227 (SYN) 0.84 0.04 Genomewide NA-10 G758C V247L 0.54 0.06 Genomewide NA-3 A909C K297T 1.09 0.14 HA/NA NA-20 A1026U K336M 0.39 0.08 HA/NA NA-29 G1040C V341L 0.74 0.09 HA/NA NA-18 U1070G S351A 0.90 0.19 Genomewide NA-5 U1130G F371V 0.89 0.22 HA/NA NA-16 U1162C T381 (SYN) 0.98 0.01 Genomewide NA-1 G1168U R383 (SYN) 0.93 0.05 HA/NA NA-19 G1355A E446K 0.71 0.06 Genomewide NA-9 7 (M) C174G M1 P50R 0.40 0.11 Genomewide M-7 A541C M1 L172 (SYN) 0.93 0.09 Genomewide M-4 G661C M1 M212I 0.48 0.01 Genomewide M-1 U861G M2 C50G 0.74 0.07 Genomewide M-2 8 (NS) U51A NS1 F9I, NEP F9I 0.86 0.05 Genomewide NS-1 G227U NS1 R67 (SYN) 0.85 0.13 Genomewide NS-2 G648U NS1 E208stop, NEP M50I 0.92 0.04 Genomewide NS-5 G650C NS1 E208D, NEP R51T 0.78 0.04 Genomewide NS-4 A809G NEP Q104R 0.91 0.06 Genomewide NS-3 ND, no data * range of 2 replicates 10.1371/journal.ppat.1005856.g002Fig 2 Location and fitness for all mutations. Each mutation in Tables 1 and 3 is shown in its reading frame(s) with substitution type (nonsynonymous, synonymous, or noncoding) and fitness (see legend). 10.1371/journal.ppat.1005856.g003Fig 3 Histograms and cumulative distribution functions of influenza A virus mutational fitness effects. Data are shown for all single nucleotide mutants (A, n = 128), the randomly selected genome-wide dataset (B, n = 95), the HA and NA dataset (C, n = 57), and the “internal” dataset (D, n = 71). Relative fitness values, bin width 0.1, are shown on the x-axis, and number of mutations in each histogram bar (left) and percent in cumulative distribution (right) are shown on the y-axes. The cumulative distribution functions show only the viable mutations (fitness > 0). HA mutational fitness effects in head and stem regions We mapped our HA mutants to the protein structure in order to observe structural patterns in HA fitness effects (Fig 4). Similar to previous studies on mutational fitness effects in influenza A hemagglutinin [35,46], we observed that the HA head seemed to tolerate mutations better than the stem domain. The mutants on the HA1 head domain had an average fitness of 0.77 and a lethal fraction of 2/14. Mutants on the HA2 stem region had an average fitness of 0.56 and a lethal fraction of 4/11. While suggestive, these differences in mean fitness did not achieve statistical significance (p = 0.258, Mann Whitney U test). Two mutations were located in one of the four H1N1 antigenic sites [47], K170N (Sa, fitness 1.01 ± 0.12) and S206R (Sb, fitness 0.72 ± 0.14). 10.1371/journal.ppat.1005856.g004Fig 4 Fitness impact of mutations on HA. Mutations were placed onto a structural model of the hemagglutinin protein (PBD 1RVX). Shown are mutations on the head and stem regions, HA1 and HA2. Non-coding mutations (HA-8) and mutations on the signal peptide (HA-11), splice site (HA-1), and transmembrane domain (HA-40, HA-15, HA-10, HA-20) are not shown. Mutations are color coded as follows according to their relative fitness: lethal mutations are red, 0.6–0.8 are orange, 0.8–1.0 are yellow, and 1.0–1.2 are green. We found no HA mutations with a fitness between 0–0.6. Fitting mutational fitness effects to probability density functions We compared the distributions of fitness effects by plotting the fitness of all mutants as histograms and overlaying fitness values for the viable mutants as cumulative distribution functions. We show distributions for the total dataset (n = 128, Fig 3A), the randomly selected genome-wide dataset (n = 95, Fig 3B), the expanded HA/NA dataset (n = 57, Fig 3C), and the comparison “internal” 6 segment dataset (n = 71, Fig 3D). As above, the lethal fraction in each group was ~30%. Of the viable mutants, 47 out of 90 were weakly deleterious or neutral (fitness 0.85–1.05). Seven were weakly beneficial (fitness 1.05–1.15). We are conservative in classifying weakly deleterious and weakly beneficial mutations, since the fitness values for many of these mutants were not statistically different from 1. The mutations in 35 of the 90 viable mutants were more deleterious (fitness < 0.85). The mean and standard deviation for the viable mutants were: 0.82 ± 0.21 in the total, 0.80 ± 0.22 in the genome-wide, 0.88 ± 0.16 in the HA/NA, and 0.78 ± 0.24 in the internal datasets, respectively. While the HA/NA and internal datasets had a similar lethal fraction (Table 1 and associated text), the fitness impact in the viable fraction tended to be lower in the HA/NA-encoding segments relative to the internal segment group, but did not achieve statistical significance (p = 0.08, Mann Whitney U test). The fitness values of the viable mutants were not distributed normally, and we therefore determined which type of distribution best fit the data. We modeled our empiric data on the deleterious mutations against exponential, gamma, beta, Weibull, and lognormal predictions and determined the best fit based on their Akaike information criteria (AIC, Table 4). For each dataset, the beta model was the best at capturing the distribution of fitness effects. We also fit distributions from a large, recently published dataset of poliovirus mutants [48]. Here, the beta model also provided the best fit. These beta models are described by two shape parameters, α and β, in which the expected value of the beta distribution is α/(α+β). The α and β parameters for each of the random, surface, and internal influenza MFE datasets as well as the poliovirus MFE dataset are similar (Table 4). Some have suggested that deleterious mutational fitness effects may be better explained by more complex models [49]. We therefore attempted to combine the above distributions with uniform distributions, in which we allowed the latter to describe a proportion of our mutants’ fitness effects. However, in the cases tested, adding this uniform distribution did not improve the fit of our models. While the influenza A and poliovirus mutational fitness effects were both best described by a beta distribution, there was no similar consensus among previously characterized viruses [19]. VSV was best described by a lognormal + uniform distribution, TEV by a beta distribution, phiX174 by an exponential distribution, QB by a gamma distribution, and F1 by a log normal distribution. In most of these fits, the AIC of the beta distribution did not differ appreciably from the AIC of the best-fit probability density function. 10.1371/journal.ppat.1005856.t004Table 4 Model fit for MFE distributions. Total Genome-wide HA/NA Internal Poliovirus Exponential 109.824436 84.814139 49.11733 61.536311 6970.5079 Gamma -2.633794 5.372353 -23.9851 9.75584 1338.4441 Weibull -33.906937 -17.061899 -35.7708 -4.538429 -499.8882 Lognormal 12.427892 17.069145 -21.1132 18.0786 3174.4454 Beta -68.630982 -48.693444 -42.91293 -28.713227 -2519.1795 Shape Parameters (95% CI) alpha 3.258 (2.277–4.487) 2.833 (1.882–4.062) 6.306 (3.531–10.216) 2.47 (1.544–3.715) 2.166 (2.086–2.247) beta 0.985 (0.731–1.297) 0.893 (0.638–1.216) 1.398 (0.857–2.151) 0.88 (0.595–1.252) 1.022 (0.989–1.056) Fitness in alternate genetic backgrounds All of our mutations were generated in the context of WSN33, a lab-adapted H1N1 strain. Recent work suggests that epistasis is quite prevalent across the influenza virus genome, and we therefore considered it likely that the same mutations would have distinct fitness effects in other genetic backgrounds [50–53]. To understand the degree to which our fitness measurements for specific WSN33 mutations are generalizable, we compared our data to those obtained for analogous mutants in WSN33 and other genetic backgrounds. Bloom and colleagues have used deep mutational scanning (DMS) to broadly sample most of the possible amino acid substitutions in the HA and nucleoprotein open reading frames [35,46,54,55]. Deep sequencing of mutant libraries before and after passage was used to infer site preferences for every amino acid at every site and to calculate site entropy, or inherent tolerance of substitution at a given site. Site entropy will tend to capture the general mutability of a position, while site preference is likely to be more specific for a mutation in a given genetic background. Importantly for our comparison, many of the plasmids had multiple mutations and the site preferences represent the average effect of a mutation in the background of very similar, but distinct, sequences. We first compared our 43 HA point mutants to DMS data on HA from WSN33 [35,46]. Here, we observed a statistically significant, and reasonably strong, correlation between our fitness values and site entropy for nonsynonymous substitutions (Fig 5A, Spearman r = 0.56, p = 0.0018). Consistent with the fact that both datasets were generated in the WSN33 background, the correlation between the fitness of viruses with our substitutions (nonsynonymous and synonymous) and the site preference for the same substitution was similar (Spearman r = 0.66, p < 0.001). It isn’t clear the degree to which the modest correlations for site entropy and site preference are attributable to differences in experimental set-up or the scale and error of measurement for two very different datasets. Of note, Thyagarajan and Bloom observed a similarly modest correlation (Pearson r = 0.48, p < 10−10) between their DMS data on WSN33 and a DMS separate study by Wu et al. [33], see discussion in [35]. 10.1371/journal.ppat.1005856.g005Fig 5 Correlation of fitness values with site entropy and preference. (A) Fitness of nonsynonymous HA mutants vs. site entropy (top) and fitness of all HA mutants vs. site preference (bottom) as reported in [46]. Unscaled values are shown. Correlations were similar for scaled values, see S3 Table. (B) Fitness of nonsynonymous NP mutants vs. site entropy (top) and fitness of all NP mutants vs. site preference (bottom) as reported for PR8 in [55]. (C) Fitness of nonsynonymous NP mutants vs. site entropy (top) and fitness of all NP mutants vs. site preference (bottom) as reported for A/Aichi/1968 (H3N2) in [55]. Note that scale in (A) for site preference is different as there were synonymous mutants in this dataset, but not in the NP datasets. We next compared the fitness of our 12 NP mutants to DMS data on NP from influenza A/Puerto Rico/8/1934 (H1N1, PR8) and A/Aichi/1968 (H3N2) [54,55]. In PR8, a closely related lab strain, we found a stronger correlation between fitness and site entropy (Fig 5B, Spearman r = 0.79, p = 0.0037). The correlation between WSN33 fitness and PR8 site preference was weaker (Spearman r = 0.60) but did achieve statistical significance (p = 0.0450). Interestingly, our WSN33 fitness values were correlated with site entropy in the more distantly related Aichi H3N2 strain (Fig 5C, Spearman r = 0.59, p = 0.0465). The correlation with site preference was weaker for the WSN33-Aichi comparison than the WSN33-PR8 comparison (Spearman r = 0.26) and did not achieve significance (p = 0.4031). Together, our fitness values are reasonably correlated with site entropy in distinct strains and subtypes, and these data suggest that a sizeable portion of mutational fitness effects are attributable to the general mutational tolerance at a given site. The observation that fitness was correlated with site preference for HA in WSN33 and NP in PR8 (H1N1), but not in Aichi (H3N2) suggests that the fitness impact of these mutations will vary with genetic background due to intragenic or intergenic epistasis. We are cautious in this interpretation given the differences in the underlying experimental design, sample size, and nature of the datasets, as outlined above. Given that Doud and Bloom found general conservation in site preference between PR8 and A/Aichi/1968, it is possible that our differences could be due to the type and number of amino acid substitutions in our smaller NP dataset. We further explored the generalizability of our MFE data by querying the influenza research database for H1N1 sequences containing any of our nonsynonymous mutations. We reasoned that of our WSN33 amino acid substitutions, only those with weakly deleterious, neutral, or beneficial fitness effects would be observed in circulating strains. Indeed, we found only two mutations, both in HA, that were present at reasonably high frequencies in more recent, circulating strains (S4 Table, S2 Fig). One was weakly deleterious, Q381L (fitness 0.93 ± 0.06, present in 78.7%) and the other was beneficial, HA M420I (fitness 1.12 ± 0.07, present at 79.7%). Consistent with what one would expect, our deleterious mutations had extremely low prevalence in the database. The prevalence of only two other non-synonymous mutations was above 0.1%, HA N101I (fitness 0.97 ± 0.06, present in 0.247%) and NA D135N (lethal, present in 0.173%). The significance of these mutants in circulating strains is unclear as they represent a relatively small number of sequences in the entire database (54/21,805 and 28/16187, respectively). Comparison to other viruses Sanjuan, Elena, and colleagues have used a similar approach to characterize the mutational fitness effects for 5 viruses: VSV, TEV, QB, phiX174, and F1 [13,19,37,40,56]. This set of viruses includes viruses with both DNA and RNA genomes (among which the authors found similar distribution patterns), and influenza is the first segmented virus to be described by this method. We compared the lethal fraction, mean, and variance of our mutational fitness effects to these previously characterized viruses. Because relative fitness values in VSV, QB, phiX174, and F1 were measured as differences in exponential growth rate, we first transformed our serial passage fitness values based on the wild type exponential growth rate in our experimental conditions (S5 Table). A similar transformation has been described for TEV to allow comparison of exponential growth rates across systems (see Methods and [19]). Overall, we found that the lethal fraction and our scaled fitness values closely matched those for these other small DNA and RNA viruses. We also calculated the skewness and kurtosis for each data set to quantitatively describe the shape of the distribution (Table 5). Following the general property that mutations are more likely to be deleterious than beneficial, we found negative values for skewness in each of our data sets. Kurtosis, measuring the “peakness” of a distribution, was above 5 for each of our data sets. A positive kurtosis value indicates that a probability density function has a heavier tail and more values near the mean than predicted by a Gaussian distribution. 10.1371/journal.ppat.1005856.t005Table 5 Comparison to MFE in other viruses. Genome Sample Lethal Fraction Arithmetic Mean Variance Skewness Kurtosis VSV ss (-) RNA 48 0.396 -0.132 0.036 -1.795 3.007 TEV ss (+) RNA 66 0.409 -0.112 0.041 0.285 -0.382 QB ss (+) RNA 42 0.286 -0.103 0.018 -1.167 0.238 phiX 174 ss DNA 45 0.200 -0.126 0.047 -1.957 4.022 F1 ss DNA 100 0.210 -0.107 0.037 -1.909 3.165 Influenza genomewide seg -ssRNA 95 0.316 -0.124 0.027 -1.970 6.866 Influenza HA/NA seg -ssRNA 57 0.281 -0.059 0.010 -1.790 7.101 Influenza internal seg -ssRNA 71 0.310 -0.139 0.033 -1.728 5.557 Discussion We report the first genome-wide study of the mutational fitness effects of single nucleotide mutations in influenza A virus. Unlike other studies of mutational tolerance in influenza, we took great pains to define the lethal fraction, a key parameter in the distribution of MFE. We ensured the relative clonality of nearly all of our stocks by next generation sequencing and used a highly quantitative assay for fitness measurements. Both the lethal fraction and the overall distribution of MFE of influenza A are quite similar to what has been found for ssDNA and other RNA viruses. Consistent with what has been assumed, but to our knowledge never shown, the surface proteins of influenza virus appear to be slightly more tolerant of point mutation than the internal viral proteins. This finding did not achieve statistical significance. These results have important implications for quantitative models of influenza evolution and our general understanding of the mutational robustness of RNA viruses. The MFE of many viruses were initially explored by mutation accumulation (MA) experiments involving serial plaque to plaque transfers [25–29]. While these studies suggested that many mutations were deleterious, MA experiments are generally unable to assign fitness values to individual mutations. Similarly, mutagen sensitivity has been used to define the global impact of mutation and the relative robustness of viral strains or even multiple viral species [23,30,31,57,58]. Neither gives a reliable estimate of the lethal fraction and comparisons across viral families can be confounded by differences in host cells, basal viral mutation rates, and the pleiotropic effects of commonly used mutagens on cellular metabolism. More recently, several groups have used higher throughput assays to measure viral mutational tolerance [33,35,48]. While these approaches allow for impressively large datasets, they are similarly imprecise in assigning fitness values to individual mutations. Because lethal and strongly deleterious mutations will be present at extremely low frequency, they are easily lost from populations with serial passage and variably ascertained by even the best deep sequencing approaches. For example, the impressive characterization of poliovirus MFE by CirSeq could only quantify the average fitness of a mutation across a range of haplotypes, given that each mutation can potentially arise in the setting of a genome that already bears a different mutation [48]. Many mutations were also present at very low frequency, varied significantly across passages, and were presumably subject to clonal interference. Measurement of fitness effects from these data also required extensive modeling of genetic drift. In influenza A virus, two different groups have used random mutagenesis of plasmids and next generation sequencing of recovered viruses to infer the mutational robustness of hemagglutinin [33,35] and nucleoprotein [54,55]. Because many of the plasmids had multiple mutations and the sequencing assays were unable to accurately measure fitness, these important works were nevertheless unable to model the distribution of MFE. The only genome-wide study of influenza A MFE, of which we are aware, utilized transposon-based insertional mutagenesis, and is less informative about the impact of single nucleotide mutation [59]. We chose a lower throughput, and in our opinion more reliable, approach to define the distribution of MFE in influenza A virus. Sanjuan, Moya, and Elena pioneered the use of small libraries of viruses with single nucleotide substitutions to model the MFE of vesicular stomatitis virus [13]. Since this initial study, Sanjuan has used a similar approach to define the MFE of the phages QB, f1, and phiX174 [36,37], and Elena has applied it to tobacco etch virus [40]. The advantage of this approach is that it links specific mutations to fitness values, provides precise fitness measurements, and accurately estimates the lethal fraction. We improved on this method by generating a larger genome-wide library of mutants and by also including sub-libraries to compare the impact of mutations on different viral proteins. Our exhaustive characterization of the lethal fraction by repeated transfection, documentation of the relative clonality of each stock by deep sequencing, and use of a highly quantitative competition assay also ensure the quality of our model of influenza A MFE. An obvious drawback of this approach is that it was only feasible to analyze 128 single nucleotide mutations. While we took great pains to ensure random selection of our mutations, it is possible that our random sample of mutations is biased in some way from the overall set of roughly 39,000 possible SNV in influenza A viruses. We are encouraged that our distribution is similar to what has been found in both high and low throughput studies [19,48]. Fitness values are specific to the environmental conditions in which they are measured [60]. For example, in cell culture they can vary with host cell type, temperature, and multiplicity of infection. We chose to use A549 cells, as they are derived from a respiratory epithelium and support relatively efficient replication of influenza virus. The variability and inefficiency of primary cells present difficulties for a large-scale comparative study such as ours. We chose a relatively low multiplicity to reduce bias in fitness measurements due to complementation and reassortment in multiply infected cells. We note, however, that our relatively low transfer populations in the serial passage experiment could contribute to the observed variability in replicate fitness measurements through genetic drift [43]. While our cell-based assay will capture elements of fitness related to virus interactions with the host cell machinery and evasion of the innate immune response, it does not model the impact of adaptive immunity. While we would argue that many of the deleterious mutations are likely to be so in many environments, some mutations that are only weakly deleterious in cell culture may be subject to stronger purifying selection in nature. Finally, our mutations in WSN33 may not have the same fitness impact in H3N2, or other subtypes, due to intragenic and intergenic epistasis. The observed correlation between site entropy and fitness on a set of a HA and NP mutants suggests that the overall distribution of MFE will be conserved. Across viral systems, increased genetic variability is often observed in surface or structural proteins relative to internally located, non-structural proteins. This bias could be attributable to local differences in mutation rate or mutational robustness and may have implications for evasion of host antibody [61]. It is therefore interesting that the fitness impact of mutations in the segments coding for hemagglutinin and neuraminidase appear to be slightly less than those in proteins encoded by the other six genomic segments. Host-specific evolutionary rates are also higher for HA and NA relative to the other 6 segments [62]. The 6 internal segments have similar rates, and we were underpowered to distinguish differences in robustness among them. Our finding of a trend toward increased mutational tolerance in the HA protein, and the head in particular, is consistent with the tolerance of HA to transposon insertion [59] and deep mutational scanning experiments [35]. These data suggest that HA is more robust to mutation, possibly as a consequence of this history of strong and repeated selection for antigenic escape. A byproduct of this robustness could be the greater exploration of antigenic sequence space and subsequent immune escape [63]. Here, mutational robustness would increase evolvability, or the capacity for adaptive evolution. This concept is attractive in view of the proposed epochal evolution of influenza A H3N2 on neutral networks [64]. We note that our study was of an H1N1 virus, and there may be important differences in the mutational robustness of H3N2 and H1N1 surface proteins. This is especially important given their different evolutionary rates [65]. As we and others have pointed out, the relationship between mutational robustness and evolvability is complex, and there are clearly situations in which increased robustness will either increase or decrease evolvability [16,66–68]. The genome-wide distribution of mutational fitness effects of influenza A virus is similar to those of poliovirus ([48] and this work), VSV, tobacco etch virus, and phages F1, QB, and phiX174 [19,36–40]. Because we relied on an assumption of exponential growth over the course of the competition assay to transform our fitness values into exponential growth rates (see Methods), it is possible that we have either over- or underestimated the similarities. Given that this collection includes a negative sense ssRNA virus (VSV), a segmented negative sense RNA virus (influenza A), three positive sense ssRNA viruses, and two ssDNA viruses, it is clear that the type and polarity of the genomic nucleic acid are not major determinants of robustness. Their similarity likely reflects shared constraints due to small and compact genomes [19]. Small genome viruses often have overlapping reading frames or regulatory elements that are more sensitive to genetic disruption. Larger genomes allow for more genetic redundancy and modularity, which could limit the impact of point mutations. Interestingly, the hypersensitivity of these viruses to mutation at an individual level may actually promote robustness at a population level [69]. In large populations under strong purifying selection, deleterious variants are rapidly purged and the wild type sequence preserved. We note that recombination and reassortment in RNA viruses are often considered to be adaptive as they could also increase robustness by purging deleterious mutations from populations [70–72]. Given the association between sex and mutation fitness effects in evolutionary theory, it is interesting that we find similar robustness of an asexual virus (VSV), a recombining virus (poliovirus) and a reassorting virus (influenza). While our assay conditions minimized reassortment, a long-term life history of reassortment does not appear to have altered the intrinsic mutational fitness effects of influenza virus. The evolutionary dynamics of influenza A virus have been intensively studied at the global and molecular level. Our data on mutational fitness effects will be useful in modeling these processes across scales. We expect that further comparative studies in this area might elucidate important similarities and differences between different influenza subtypes and between influenza and other viruses. Materials and Methods Viruses and cells Madin Darby canine kidney cells (MDCK) were provided by Dr. Arnold Monto (University of Michigan) and A549 human lung epithelial cells were provided by Dr. Michael Bachman (University of Michigan). Human embryonic kidney 293T fibroblasts were provided by Dr. Raul Andino (UCSF). Both cell lines were maintained in Dulbecco’s Modified Eagle Medium (DMEM, Invitrogen) supplemented with 10% fetal bovine serum (Gibco and HyClone), 25mM HEPES (Invitrogen), and 0.1875% bovine serum albumin (Life Technologies). Viral infections were performed in DMEM supplemented with 25mM HEPES, 0.1875% bovine serum albumin, and 2μg/ml TPCK-trypsin (Worthington). The 8 plasmid reverse genetic system containing the genomic segments for influenza A/WSN33 (pHW181-PB2, pHW182-PB1, pHW-183-PA, pHW184-HA, pHW185-NP, pHW186-NA, pHW187-M, pHW188-NS) was a kind gift of Robert Webster ([41], St. Jude’s Children’s Research Hospital) Site directed mutagenesis An R script was used to select randomly the nucleotide position and base change for all mutants in this study and to design optimal oligonucleotides for mutagenesis by polymerase chain reaction (PCR). Noncoding and promoter regions were included. Due to sequence context and GC content, several primers had to be designed manually (S1 Table). One hundred and eight of the single nucleotide mutants were generated using the QuickChange site directed mutagenesis kit (Agilent Technologies) according to the manufacturer’s protocol. The remaining 21 single nucleotide mutants were generated by overlap extension PCR [73]. Here, the outer primers contained 5’ BsmBI or BsaI restriction sites for subsequent cloning into pHW2000 [41,74]. The barcoded PB1 segment (555) was made using the Quickchange protocol and primers pPolPB1_555f 5’ GATCACAACTCATTTCCAACGGAAACGGAGGGTGAGAGACAAT 3’ and pPolPB1-555r ATTGTCTCTCACCCTCCGTTTCCGTTGGAAATGAGTTGTGATC. In each plasmid clone, the presence of the desired mutation(s) and the absence of second site mutations were verified by sequencing of the entire influenza segment. Transfection and viral stocks Equal quantities of MDCK and 293T cells were seeded at a total density of 1,000,000 cells per well of a 6 well plate 24 hours prior to transfection in complete DMEM (see above). Each transfection mixture contained 1μg of the mutant plasmid, 1μg of each of the other 7 wild type plasmids, 16μl of TransIT-LT1 (Mirus) and 250μl Optimem (Gibco). These reagents were incubated together for 45 minutes at room temperature and applied dropwise to the cellular monolayer. After 24 hours, the media was changed to viral infection media (see above). Recombinant passage 0 virus was harvested 48 hours post-transfection, clarified by centrifugation at 200 x g for 3 minutes, and stored in aliquots with 0.5% glycerol at minus 70°C. Passage 1 (P1) stocks were generated by a single passage on MDCK cells at a multiplicity of infection (moi) ≤ 0.001. Virus was applied to cells for one hour and aspirated, and then viral media was added. Passage 1 stocks were harvested at 48 hours post-infection. All stocks were titered by tissue culture infectious dose (TCID50, [75]). We subjected all transfection supernatants with undetectable P0 titers to blind passaging. One or 0.333 ml of virus was applied to a confluent T75 or T25 flask, respectively. These cultures were monitored for cytopathic effect and titered at 48 hours. Next generation sequencing Viral RNA was harvested from 200μl of each P1 supernatant using either Purelink Viral RNA (Invitrogen) or Qiamp Viral RNA (Qiagen) kits. Multiplex reverse transcription-PCR amplification of all 8 influenza virus genome segments was performed on RNA samples using Superscript III with HiFi platinum Taq (Invitrogen 12574) and primers Uni12/Inf1 (5’-GGGGGGAGCAAAAGCAGG-3’), Uni12/Inf3 (5’-GGGGGAGCGAAAGCAGG-3’), and Uni13/Inf1 (5’-CGGGTTATTAGTAGAAACAAGG-3’) [76]. Seven hundred fifty nanograms of the each amplified cDNA were sheared to an average size of 300 to 400 bp using a Covaris S220 focused ultrasonicator. Sequencing libraries were prepared using the NEBNext Ultra DNA library prep kit (NEB E7370L), Agencourt AMPure XP beads (Beckman Coulter A63881), and NEBNext multiplex oligonucleotides for Illumina (NEB E7600S). Indexed samples were pooled in equal quantities and sequenced on an Illumina MiSeq instrument with 2 x 250-base paired end reads. Sequencing reads that passed standard Illumina quality control filters were binned by index and aligned to the reference genome using bowtie [77]. Single nucleotide variants (SNV) were identified and analyzed using DeepSNV [78]. The DeepSNV algorithm relies on a clonal control to estimate the local error rate within a given sequence context and to identify strand bias in base calling. It then applies a hierarchical binomial model based on mutation calls for test and control at each base and position to identify true-positive SNV. The clonal control was a library prepared in an identical fashion from 8 plasmids containing the A/WSN33/H1N1 genome and sequenced in the same flow cell. Code for our implementation of DeepSNV can be found at https://github.com/lauringlab/variant_pipeline, and can be accessed anonymously. True positive SNV were identified from the raw output tables (S2 Table) by applying the following filtering criteria in R: (i) Bonferonni corrected p value < 0.01, (ii) average MapQ score on variant reads > 30, (iii) average phred score on variant positions > 35, (iv) average position of variant call on a read > 50 and < 200, (v) variant frequency > 0.02. Application of these criteria to a viral spike-in dataset with equivalent genome copy number inputs resulted in > 95% sensitivity and > 99.99% specificity for SNV present at > 1% of the population [79]. Competition assays Competitions were performed on A549 cells in 12 well plates, plated at a density of 2.6 x 105 per well 24 hours prior to infection. Cells were infected at a total MOI of 0.01 with an equal TCID50 of WT and a given mutant. Three replicate wells were infected with each pair of viruses. Passage 1 virus was harvested after 48 hours and one replicate was titered by TCID50. This titer was used to calculate the dilution factor necessary to maintain an MOI of 0.01 for subsequent passages. Four passages were performed for each competition. RNA was harvested from each passage using PureLink 96 well RNA mini kits (Invitrogen). Random hexamers were used to prime cDNA synthesis with 1/10 of the RNA. Each cDNA was analyzed by real time PCR using three different primer sets with duplicate PCR reactions for each sample/primer set. The first set, PB1_149f 5’ CAGAAAGGGGAAGATGGACA 3’ and PB1_360r 5’ GTCCACTCGTGTTTGCTGAA 3’, were used to quantify total viral genomic RNA. The second set, pPol1PB1_555f 5’ TCAGAGAAAGAGACGAGTGAG 3’ and pPol1PB1_555r 5’ AAACCCCCTTATTTGCATCC 3’, were used to quantify the amount of mutant viral RNA. The third set, pPol1PB1_555fm 5’ ATTTCCAACGGAAACGGAGGG 3’ and pPol1PB1_555r 5’ AAACCCCCTTATTTGCATCC3’, were used to quantify the amount of barcoded WT viral RNA. We verified that the levels of RNA (Cycle threshold, Ct) were well correlated with the infectious titer (TCID50/ml) for the mutants shown in Fig 1B. Duplicate wells were averaged and values were excluded from subsequent analysis if duplicates wells differed by > 0.5 Ct or any of the Ct were out of the empirically determined linear range for that primer pair. Relative amounts of WT and mutant RNA were determined by normalizing the cycle thresholds for each to those of the common (PB1_149f and PB1_360r) primer set (DCt = CtVirus-CtPB1 total). The normalized values for each virus passages 1–4 were then compared to passage 0 to obtain a ratio relative to P1 (ΔΔCt = CtPX-CtP0). This relative Ct value was converted to reflect the fold change (Δratio = 2-logΔΔCt). The change in ratio of the mutant relative to the change in ratio of the WT as a function of passage is the fitness ([ΔratioMut-ΔratioWT]/time). Transformation of fitness into relative growth rate Except for poliovirus, the data on other viruses were extracted from [19]. Here, the relative fitness values for VSV, QB, phiX174, and F1 were calculated as differences in the exponential growth rate. The data on TEV were originally derived by qPCR as in the present study. In this comparative analysis, Sanjuan applied a correction to the TEV data to enable direct comparisons of growth rate. We applied this same correction to our data to transform the values to something resembling an exponential growth rate. The exponential growth rate, r, of influenza A on A549 in a 12 well dish at moi 0.01, is derived from the equation: Nt = N0ert, where t = 2 days (48 hour passage), N0 is 10,000 infectious units (the amount at time 0), and Nt is the titer after time t, or 48 hours. For Nt, we used the mean of three replicates for HA-40 (9.3e5), which has a fitness of 1 (see Fig 1B). This is similar to the scaled output of WSN33 that we have observed at different moi and different well sizes. Solving for r gives a value of 2.26 per day. We then applied the following correction from [19], y = (lnx + r)/r, where y is the relative exponential growth rate of a given mutant, x is the relative fitness measured by qPCR and reported in Tables 1 and 3, and r is the estimate of the WT exponential growth rate under our experimental conditions (2.26 per day). For example, a fitness of 0.8 in our assay is transformed to a relative exponential growth rate of 0.90. The relative exponential growth rates, as calculated, are reported in S5 Table. We note that using a value 2e6 for Nt (which would be roughly consistent with the average dilution factor applied at each passage in many of the competition experiments, gives a growth rate of 2.65 per day and relative exponential growth rate of 0.916 for a virus with a fitness of 0.8. Statistics All secondary analysis and statistical tests were performed in R and GraphPad Prism6. R and Python scripts for analysis of Influenza Research Database data (as of July 8, 2016) and other analyses are available at https://github.com/lauringlab/MFE_paper. Supporting Information S1 Fig Distribution of single nucleotide variants in the influenza A genome. (A) A randomly selected genome-wide dataset (n = 95), in which the number of mutants generated per segment (grey bars) closely matches the expected distribution based on percentage of the genome contained on each segment (black bars). (B) Extra HA and NA mutations were generated to compare larger numbers of mutations on these segments, n = 57, to those on the other 6 (internal) segments, n = 71. (PDF) Click here for additional data file. S2 Fig Frequency of mutations in the Influenza Research Database. Shown are the frequencies of nonsynonymous amino acid substitutions in our mutant dataset (y-axis) and their fitness values (x-axis). (PDF) Click here for additional data file. S1 Table Primers used in this study. (XLSX) Click here for additional data file. S2 Table Secondary mutations as determined by Illumina sequencing. (XLSX) Click here for additional data file. S3 Table Site entropy and site preference values for HA and NP mutants. (XLSX) Click here for additional data file. S4 Table Raw data on mutation frequency from the Influenza Research Database. (XLS) Click here for additional data file. S5 Table Relative exponential growth rates for all mutants. (XLSX) Click here for additional data file. We thank Raul Andino, Ashley Acevedo, and Leonid Brodsky for providing the raw data on the poliovirus mutational fitness effects. We thank Matt Pauly for creating the mutation selection script, Anna Berezovsky and Keqin Gong for technical assistance, and Robert Woods for helpful discussion and a critical reading of the manuscript. ==== Refs References 1 Webster RG , Bean WJ , Gorman OT , Chambers TM , Kawaoka Y . Evolution and ecology of influenza A viruses . Microbiol Rev . 1992 ;56 : 152 –179 . 1579108 2 Nelson MI , Holmes EC . The evolution of epidemic influenza . Nat Rev Genet . 2007 ;8 : 196 –205 . 10.1038/nrg2053 17262054 3 Ghedin E , Sengamalay NA , Shumway M , Zaborsky J , Feldblyum T , Subbu V , et al Large-scale sequencing of human influenza reveals the dynamic nature of viral genome evolution . Nature . 2005 ;437 : 1162 –1166 . 10.1038/nature04239 16208317 4 Rambaut A , Pybus OG , Nelson MI , Viboud C , Taubenberger JK , Holmes EC . The genomic and epidemiological dynamics of human influenza A virus . Nature . 2008 ;453 : 615 –619 . 10.1038/nature06945 18418375 5 Sanjuán R , Nebot MR , Chirico N , Mansky LM , Belshaw R . Viral mutation rates . J Virol . 2010 ;84 : 9733 –9748 . 10.1128/JVI.00694-10 20660197 6 Nobusawa E , Sato K . Comparison of the mutation rates of human influenza A and B viruses . J Virol . 2006 ;80 : 3675 –3678 . 10.1128/JVI.80.7.3675-3678.2006 16537638 7 Parvin JD , Moscona A , Pan WT , Leider JM , Palese P . Measurement of the mutation rates of animal viruses: influenza A virus and poliovirus type 1 . J Virol . 1986 ;59 : 377 –383 . 3016304 8 Marshall N , Priyamvada L , Ende Z , Steel J , Lowen AC . Influenza Virus Reassortment Occurs with High Frequency in the Absence of Segment Mismatch . Basler CF , editor. PLoS Pathog . 2013 ;9 : e1003421 23785286 9 Tao H , Steel J , Lowen AC . Intrahost Dynamics of Influenza Virus Reassortment . J Virol . 2014 ;88 : 7485 –7492 . 10.1128/JVI.00715-14 24741099 10 Russell CA , Fonville JM , Brown AEX , Burke DF , Smith DL , James SL , et al The potential for respiratory droplet-transmissible A/H5N1 influenza virus to evolve in a mammalian host . Science . 2012 ;336 : 1541 –1547 . 10.1126/science.1222526 22723414 11 Elena SF , Sanjuán R . Adaptive value of high mutation rates of RNA viruses: separating causes from consequences . J Virol . 2005 ;79 : 11555 –11558 . 10.1128/JVI.79.18.11555-11558.2005 16140732 12 Belshaw R , Gardner A , Rambaut A , Pybus OG . Pacing a small cage: mutation and RNA viruses . Trends Ecol Evol (Amst) . 2008 ;23 : 188 –193 . 10.1016/j.tree.2007.11.010 18295930 13 Sanjuán R , Moya A , Elena SF . The distribution of fitness effects caused by single-nucleotide substitutions in an RNA virus . Proc Natl Acad Sci USA . 2004 ;101 : 8396 –8401 . 10.1073/pnas.0400146101 15159545 14 de Visser JAGM , Hermisson J , Wagner GP , Ancel Meyers L , Bagheri-Chaichian H , Blanchard JL , et al Perspective: Evolution and detection of genetic robustness . Evolution . 2003 ;57 : 1959 –1972 . 14575319 15 Elena SF , Carrasco P , Darós J-A , Sanjuán R . Mechanisms of genetic robustness in RNA viruses . EMBO Rep . 2006 ;7 : 168 –173 . 10.1038/sj.embor.7400636 16452927 16 Lauring AS , Frydman J , Andino R . The role of mutational robustness in RNA virus evolution . Nat Rev Micro. Nature Publishing Group ; 2013 ;11 : 327 –336 . 10.1038/nrmicro3003 17 Wilke CO , Adami C . Evolution of mutational robustness . Mutat Res . 2003 ;522 : 3 –11 . 12517406 18 Lauring AS , Acevedo A , Cooper SB , Andino R . Codon Usage Determines the Mutational Robustness, Evolutionary Capacity, and Virulence of an RNA Virus. Cell Host and Microbe. Elsevier Inc ; 2012 ;12 : 623 –632 . 10.1016/j.chom.2012.10.008 19 Sanjuán R . Mutational fitness effects in RNA and single-stranded DNA viruses: common patterns revealed by site-directed mutagenesis studies . Philos Trans R Soc Lond, B, Biol Sci . 2010 ;365 : 1975 –1982 . 10.1098/rstb.2010.0063 20478892 20 Fonville JM . Expected Effect of Deleterious Mutations on Within-Host Adaptation of Pathogens . Kirkegaard K , editor. J Virol . 2015 ;89 : 9242 –9251 . 10.1128/JVI.00832-15 26109724 21 Koelle K , Rasmussen DA . The effects of a deleterious mutation load on patterns of influenza A/H3N2's antigenic evolution in humans . Traulsen A , editor. eLife. eLife Sciences Publications Limited ; 2015 ;4 : e07361 10.7554/eLife.07361 22 Bull JJ , Sanjuán R , Wilke CO . Theory of lethal mutagenesis for viruses . J Virol . 2007 ;81 : 2930 –2939 . 10.1128/JVI.01624-06 17202214 23 Graci JD , Gnädig NF , Galarraga JE , Castro C , Vignuzzi M , Cameron CE . Mutational robustness of an RNA virus influences sensitivity to lethal mutagenesis . J Virol . 2012 ;86 : 2869 –2873 . 10.1128/JVI.05712-11 22190724 24 O'Dea EB , Keller TE , Wilke CO . Does Mutational Robustness Inhibit Extinction by Lethal Mutagenesis in Viral Populations? Wahl L , editor. PLoS Comput Biol . 2010 ;6 : e1000811 10.1371/journal.pcbi.1000811.g007 20548958 25 Lázaro E , Escarmís C , Domingo E , Manrubia SC . Modeling viral genome fitness evolution associated with serial bottleneck events: evidence of stationary states of fitness . J Virol . 2002 ;76 : 8675 –8681 . 10.1128/JVI.76.17.8675-8681.2002 12163587 26 Duarte E , Clarke D , Moya A , Domingo E , Holland J . Rapid fitness losses in mammalian RNA virus clones due to Muller's ratchet . Proc Natl Acad Sci USA. National Academy of Sciences ; 1992 ;89 : 6015 –6019 . 27 Clarke DK , Duarte EA , Moya A , Elena SF , Domingo E , Holland J . Genetic bottlenecks and population passages cause profound fitness differences in RNA viruses . J Virol . 1993 ;67 : 222 –228 . 8380072 28 Eyre-Walker A , Keightley PD . The distribution of fitness effects of new mutations . Nat Rev Genet . 2007 ;8 : 610 –618 . 10.1038/nrg2146 17637733 29 Montville R , Froissart R , Remold SK , Tenaillon O , Turner PE . Evolution of mutational robustness in an RNA virus . PLoS Biol . 2005 ;3 : e381 10.1371/journal.pbio.0030381 16248678 30 Lee CH , Gilbertson DL , Novella IS , Huerta R , Domingo E , Holland JJ . Negative effects of chemical mutagenesis on the adaptive behavior of vesicular stomatitis virus . J Virol . 1997 ;71 : 3636 –3640 . 9094637 31 Crotty S , Cameron CE , Andino R . RNA virus error catastrophe: direct molecular test by using ribavirin . Proc Natl Acad Sci USA . 2001 ;98 : 6895 –6900 . 10.1073/pnas.111085598 11371613 32 Novella IS , Presloid JB , Beech C , Wilke CO . Congruent Evolution of Fitness and Genetic Robustness in Vesicular Stomatitis Virus . J Virol . 2013 ;87 : 4923 –4928 . 10.1128/JVI.02796-12 23408631 33 Wu NC , Young AP , Al-Mawsawi LQ , Olson CA , Feng J , Qi H , et al High-throughput profiling of influenza A virus hemagglutinin gene at single-nucleotide resolution . Sci Rep . 2014 ;4 10.1038/srep04942 34 Qi H , Olson CA , Wu NC , Ke R , Loverdo C , Chu V , et al A quantitative high-resolution genetic profile rapidly identifies sequence determinants of hepatitis C viral fitness and drug sensitivity . PLoS Pathog. Public Library of Science ; 2014 ;10 : e1004064 10.1371/journal.ppat.1004064 35 Thyagarajan B , Bloom JD . The inherent mutational tolerance and antigenic evolvability of influenza hemagglutinin . eLife . 2014 ;3 : e03300 –e03300 . 10.7554/eLife.03300 36 Peris JB , Davis P , Cuevas JM , Nebot MR , Sanjuán R . Distribution of fitness effects caused by single-nucleotide substitutions in bacteriophage f1 . Genetics. Genetics Society of America ; 2010 ;185 : 603 –609 . 10.1534/genetics.110.115162 37 Domingo-Calap P , Cuevas JM , Sanjuán R . The fitness effects of random mutations in single-stranded DNA and RNA bacteriophages . PLoS Genet . 2009 ;5 : e1000742 10.1371/journal.pgen.1000742 19956760 38 Cuevas JM , Domingo-Calap P , Sanjuán R . The fitness effects of synonymous mutations in DNA and RNA viruses . Mol Biol Evol . 2012 ;29 : 17 –20 . 10.1093/molbev/msr179 21771719 39 Cuevas JM , Duffy S , Sanjuán R . Point mutation rate of bacteriophage PhiX174 . Genetics . 2009 ;183 : 747 –749 . 10.1534/genetics.109.106005 19652180 40 Carrasco P , la Iglesia de F , Elena SF . Distribution of Fitness and Virulence Effects Caused by Single-Nucleotide Substitutions in Tobacco Etch Virus . J Virol. American Society for Microbiology ; 2007 ;81 : 12979 –12984 . 10.1128/JVI.00524-07 41 Hoffmann E , Neumann G , Kawaoka Y , Hobom G , Webster RG . A DNA transfection system for generation of influenza A virus from eight plasmids . Proc Natl Acad Sci USA . 2000 ;97 : 6108 –6113 . 10.1073/pnas.100133697 10801978 42 Cuevas JM , Domingo-Calap P , Sanjuán R . The fitness effects of synonymous mutations in DNA and RNA viruses . Mol Biol Evol. Oxford University Press ; 2012 ;29 : 17 –20 . 10.1093/molbev/msr179 43 Voronin Y , Holte S , Overbaugh J , Emerman M . Genetic drift of HIV populations in culture . PLoS Genet . 2009 ;5 : e1000431 10.1371/journal.pgen.1000431 19300501 44 Jagger BW , Wise HM , Kash JC , Walters K-A , Wills NM , Xiao Y-L , et al An overlapping protein-coding region in influenza A virus segment 3 modulates the host response . Science. American Association for the Advancement of Science ; 2012 ;337 : 199 –204 . 10.1126/science.1222213 45 Watanabe T , Watanabe S , Noda T , Fujii Y , Kawaoka Y . Exploitation of Nucleic Acid Packaging Signals To Generate a Novel Influenza Virus-Based Vector Stably Expressing Two Foreign Genes . J Virol . 2003 ;77 : 10575 –10583 . 10.1128/JVI.77.19.10575-10583.2003 12970442 46 Doud MB , Bloom JD . Accurate Measurement of the Effects of All Amino-Acid Mutations on Influenza Hemagglutinin . Viruses . 2016 ;8 10.3390/v8060155 47 Xu R , Ekiert DC , Krause JC , Hai R , Crowe JE , Wilson IA . Structural basis of preexisting immunity to the 2009 H1N1 pandemic influenza virus . Science. American Association for the Advancement of Science ; 2010 ;328 : 357 –360 . 10.1126/science.1186430 48 Acevedo A , Brodsky L , Andino R . Mutational and fitness landscapes of an RNA virus revealed through population sequencing . Nature . 2014 ;505 : 686 –690 . 10.1038/nature12861 24284629 49 Elena SF , Ekunwe L , Hajela N , Oden SA , Lenski RE . Distribution of fitness effects caused by random insertion mutations in Escherichia coli . Genetica . 1998 ;102–103 : 349 –358 . 9720287 50 Bloom JD , Gong LI , Baltimore D . Permissive secondary mutations enable the evolution of influenza oseltamivir resistance . Science . 2010 ;328 : 1272 –1275 . 10.1126/science.1187816 20522774 51 Gong LI , Suchard MA , Bloom JD . Stability-mediated epistasis constrains the evolution of an influenza protein . eLife . 2013 ;2 : e00631 10.7554/eLife.00631 23682315 52 Gong LI , Bloom JD . Epistatically Interacting Substitutions Are Enriched during Adaptive Protein Evolution . Weinreich DM , editor. PLoS Genet . 2014 ;10 : e1004328 10.1371/journal.pgen.1004328 24811236 53 Kryazhimskiy S , Dushoff J , Bazykin GA , Plotkin JB . Prevalence of epistasis in the evolution of influenza A surface proteins . PLoS Genet . 2011 ;7 : e1001301 10.1371/journal.pgen.1001301 21390205 54 Bloom JD . An experimentally determined evolutionary model dramatically improves phylogenetic fit . Mol Biol Evol. Oxford University Press ; 2014 ;31 : 1956 –1978 . 10.1093/molbev/msu173 55 Doud MB , Ashenberg O , Bloom JD . Site-Specific Amino Acid Preferences Are Mostly Conserved in Two Closely Related Protein Homologs . Mol Biol Evol . 2015 ;32 : 2944 –2960 . 10.1093/molbev/msv167 26226986 56 Peris JB , Davis P , Cuevas JM , Nebot MR , Sanjuán R . Distribution of fitness effects caused by single-nucleotide substitutions in bacteriophage f1 . Genetics. Genetics ; 2010 ;185 : 603 –609 . 10.1534/genetics.110.115162 20382832 57 Pauly MD , Lauring AS . Effective lethal mutagenesis of influenza virus by three nucleoside analogs . J Virol . 2015 ;89 : 3584 –3597 . 10.1128/JVI.03483-14 25589650 58 Smith EC , Blanc H , Vignuzzi M , Denison MR . Coronaviruses Lacking Exoribonuclease Activity Are Susceptible to Lethal Mutagenesis: Evidence for Proofreading and Potential Therapeutics . Diamond MS , editor. PLoS Pathog . 2013 ;9 : e1003565 10.1371/journal.ppat.1003565.t001 23966862 59 Heaton NS , Sachs D , Chen C-J , Hai R , Palese P . Genome-wide mutagenesis of influenza virus reveals unique plasticity of the hemagglutinin and NS1 proteins . Proceedings of the National Academy of Sciences. National Acad Sciences ; 2013 ;110 : 20248 –20253 . 10.1073/pnas.1320524110 60 Orr HA . Fitness and its role in evolutionary genetics . Nat Rev Genet . 2009 ;10 : 531 –539 . 10.1038/nrg2603 19546856 61 Cuevas JM , Geller R , Garijo R , López-Aldeguer J , Sanjuán R . Extremely High Mutation Rate of HIV-1 In Vivo . Rowland-Jones SL , editor. PLoS Biol . 2015 ;13 : e1002251 –19 . 10.1371/journal.pbio.1002251 26375597 62 Worobey M , Han G-Z , Rambaut A . A synchronized global sweep of the internal genes of modern avian influenza virus . Nature. Nature Publishing Group ; 2014 ;: 1 –16 . 10.1038/nature13016 63 Plotkin JB , Dushoff J . Codon bias and frequency-dependent selection on the hemagglutinin epitopes of influenza A virus . Proc Natl Acad Sci USA . 2003 ;100 : 7152 –7157 . 10.1073/pnas.1132114100 12748378 64 Koelle K , Cobey S , Grenfell B , Pascual M . Epochal evolution shapes the phylodynamics of interpandemic influenza A (H3N2) in humans . Science . 2006 ;314 : 1898 –1903 . 10.1126/science.1132745 17185596 65 Bedford T , Riley S , Barr IG , Broor S , Chadha M , Cox NJ , et al Global circulation patterns of seasonal influenza viruses vary with antigenic drift . Nature . 2015 ;523 : 217 –220 . 10.1038/nature14460 26053121 66 McBride RC , Ogbunugafor CB , Turner PE . Robustness promotes evolvability of thermotolerance in an RNA virus . BMC Evol Biol . 2008 ;8 : 231 10.1186/1471-2148-8-231 18694497 67 Cuevas JM , Moya A , Sanjuán R . A genetic background with low mutational robustness is associated with increased adaptability to a novel host in an RNA virus . J Evol Biol . 2009 ;22 : 2041 –2048 . 10.1111/j.1420-9101.2009.01817.x 19702844 68 Elena SF . RNA virus genetic robustness: possible causes and some consequences . Curr Opin Virol . 2012 10.1016/j.coviro.2012.06.008 69 Krakauer DC , Plotkin JB . Redundancy, antiredundancy, and the robustness of genomes . Proc Natl Acad Sci USA . 2002 ;99 : 1405 –1409 . 10.1073/pnas.032668599 11818563 70 Fonville JM , Marshall N , Tao H , Steel J , Lowen AC . Influenza Virus Reassortment Is Enhanced by Semi-infectious Particles but Can Be Suppressed by Defective Interfering Particles . Lauring A , editor. PLoS Pathog . 2015 ;11 : e1005204 –30 . 10.1371/journal.ppat.1005204 26440404 71 Xiao Y , Rouzine IM , Bianco S , Acevedo A , Goldstein EF , Farkov M , et al RNA Recombination Enhances Adaptability and Is Required for Virus Spread and Virulence . Cell Host and Microbe. Elsevier Inc ; 2016 ;19 : 493 –503 . 10.1016/j.chom.2016.03.009 72 Simon-Loriere E , Holmes EC . Why do RNA viruses recombine? Nat Rev Micro . 2011 ;9 : 617 –626 . 10.1038/nrmicro2614 73 Ho SN , Hunt HD , Horton RM , Pullen JK , Pease LR . Site-directed mutagenesis by overlap extension using the polymerase chain reaction . Gene . 1989 ;77 : 51 –59 . 2744487 74 Hoffmann E , Stech J , Guan Y , Webster RG , Perez DR . Universal primer set for the full-length amplification of all influenza A viruses . Arch Virol . 2001 ;146 : 2275 –2289 . 11811679 75 REED LJ , MUENCH H . A SIMPLE METHOD OF ESTIMATING FIFTY PER CENT ENDPOINTS . 1938 ;27 : 493 –497 . Available: http://aje.oxfordjournals.org/content/27/3/493.short 76 Zhou B , Donnelly ME , Scholes DT , St George K , Hatta M , Kawaoka Y , et al Single-reaction genomic amplification accelerates sequencing and vaccine production for classical and Swine origin human influenza a viruses . J Virol . 2009 ;83 : 10309 –10313 . 10.1128/JVI.01109-09 19605485 77 Langmead B , Trapnell C , Pop M , Salzberg SL . Ultrafast and memory-efficient alignment of short DNA sequences to the human genome . Genome Biol . 2009 ;10 : R25 10.1186/gb-2009-10-3-r25 19261174 78 Gerstung M , Beisel C , Rechsteiner M , Wild P , Schraml P , Moch H , et al Reliable detection of subclonal single-nucleotide variants in tumour cell populations . Nature Communications. Nature Publishing Group ; 1AD;3 : 811 –8 . 10.1038/ncomms1814 79 McCrone JT , Lauring AS . Measurements of intrahost viral diversity are extremely sensitive to systematic errors in variant calling . J Virol. American Society for Microbiology ; 2016 ;: JVI.00667–16. 10.1128/JVI.00667-16

PMC005xxxxxx/PMC5003364.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757136310.1371/journal.pone.0162042PONE-D-16-24899Research ArticleBiology and Life SciencesAnatomyBrainCerebral CortexCerebellumMedicine and Health SciencesAnatomyBrainCerebral CortexCerebellumBiology and Life SciencesNeuroscienceCognitive ScienceCognitive PsychologyLearningBiology and Life SciencesPsychologyCognitive PsychologyLearningSocial SciencesPsychologyCognitive PsychologyLearningBiology and Life SciencesNeuroscienceLearning and MemoryLearningBiology and Life SciencesNeuroscienceCognitive ScienceCognitive NeuroscienceMotor ReactionsBiology and Life SciencesNeuroscienceCognitive NeuroscienceMotor ReactionsBiology and Life SciencesNeuroscienceCognitive ScienceCognitive NeuroscienceReaction TimeBiology and Life SciencesNeuroscienceCognitive NeuroscienceReaction TimeMedicine and Health SciencesNeurologyNeurodegenerative DiseasesMovement DisordersAtaxiaBiology and Life SciencesAnatomyMusculoskeletal SystemMedicine and Health SciencesAnatomyMusculoskeletal SystemMedicine and Health SciencesOphthalmologyVisual ImpairmentsResearch and Analysis MethodsBioassays and Physiological AnalysisElectrophysiological TechniquesBrain ElectrophysiologyTranscranial StimulationTranscranial Magnetic StimulationBiology and Life SciencesPhysiologyElectrophysiologyNeurophysiologyBrain ElectrophysiologyTranscranial StimulationTranscranial Magnetic StimulationMedicine and Health SciencesPhysiologyElectrophysiologyNeurophysiologyBrain ElectrophysiologyTranscranial StimulationTranscranial Magnetic StimulationBiology and Life SciencesNeuroscienceNeurophysiologyBrain ElectrophysiologyTranscranial StimulationTranscranial Magnetic StimulationBiology and Life SciencesNeuroscienceBrain MappingTranscranial StimulationTranscranial Magnetic StimulationImpaired Spatio-Temporal Predictive Motor Timing Associated with Spinocerebellar Ataxia Type 6 Spatio-Temporal Prediction and SCA6http://orcid.org/0000-0003-2555-9719Broersen Robin 12Onuki Yoshiyuki 1Abdelgabar Abdel R. 1Owens Cullen B. 2Picard Samuel 12Willems Jessica 2Boele Henk-Jan 2Gazzola Valeria 13Van der Werf Ysbrand D. 14De Zeeuw Chris I. 12*1 Netherlands Institute for Neuroscience, Royal Dutch Academy of Arts & Sciences, Amsterdam, The Netherlands2 Department of Neuroscience, Erasmus Medical Center, Rotterdam, The Netherlands3 Department of Psychology, University of Amsterdam, Amsterdam, The Netherlands4 Department of Anatomy and Neurosciences, VU University Medical Center, Amsterdam, The NetherlandsMeck Warren H EditorDuke University, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: YO SP YDvdW CIDZ. Data curation: RB YO SP CIDZ. Formal analysis: RB YO. Funding acquisition: VG YDvdW CIDZ. Investigation: RB YO ARA CBO SP JW HJB. Methodology: YO ARA SP YDvdW CIDZ. Project administration: RB SP JW. Resources: CIDZ. Software: YO SP. Supervision: VG YDvdW CIDZ. Validation: RB ARA YO SP YDvdW CIDZ. Visualization: RB YO. Writing – original draft: RB YO. Writing – review & editing: RB YO ARA CBO VG YDvdW CIDZ. * E-mail: c.dezeeuw@erasmusmc.nl29 8 2016 2016 11 8 e016204221 6 2016 16 8 2016 © 2016 Broersen et al2016Broersen et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Many daily life activities demand precise integration of spatial and temporal information of sensory inputs followed by appropriate motor actions. This type of integration is carried out in part by the cerebellum, which has been postulated to play a central role in learning and timing of movements. Cerebellar damage due to atrophy or lesions may compromise forward-model processing, in which both spatial and temporal cues are used to achieve prediction for future motor states. In the present study we sought to further investigate the cerebellar contribution to predictive and reactive motor timing, as well as to learning of sequential order and temporal intervals in these tasks. We tested patients with spinocerebellar ataxia type 6 (SCA6) and healthy controls for two related motor tasks; one requiring spatio-temporal prediction of dynamic visual stimuli and another one requiring reactive timing only. We found that healthy controls established spatio-temporal prediction in their responses with high temporal precision, which was absent in the cerebellar patients. SCA6 patients showed lower predictive motor timing, coinciding with a reduced number of correct responses during the ‘anticipatory’ period on the task. Moreover, on the task utilizing reactive motor timing functions, control participants showed both sequence order and temporal interval learning, whereas patients only showed sequence order learning. These results suggest that SCA6 affects predictive motor timing and temporal interval learning. Our results support and highlight cerebellar contribution to timing and argue for cerebellar engagement during spatio-temporal prediction of upcoming events. http://dx.doi.org/10.13039/501100003246Nederlandse Organisatie voor Wetenschappelijk Onderzoek452-14-015Gazzola Valeria http://dx.doi.org/10.13039/501100003246Nederlandse Organisatie voor Wetenschappelijk Onderzoek433-09-245Van der Werf Ysbrand D. http://dx.doi.org/10.13039/501100003298Nederlands Instituut voor Onderzoek van de GezondheidszorgDe Zeeuw Chris I. http://dx.doi.org/10.13039/501100003246Nederlandse Organisatie voor Wetenschappelijk OnderzoekDe Zeeuw Chris I. http://dx.doi.org/10.13039/501100000781European Research CouncilDe Zeeuw Chris I. This study was supported by the Dutch Organization for Medical Sciences and Life Sciences (C.I.D.Z.); Programmes for Excellence ’Brain & Cognition: an Integrated Approach’ (433-09-245) (Y.D.v.d.W and C.I.D.Z.) and VIDI (452-14-015) (V.G.) of the Netherlands Organization for Scientific Research; and the ERC-advanced (CCC), ERC-PoC (BrainFrame), CEREBNET and C7 programs of the European Community (C.I.D.Z.). Data AvailabilityAll relevant data are within the paper.Data Availability All relevant data are within the paper. ==== Body Introduction Many daily life activities demanding immediate timed motor responses require integration of spatial and temporal information at the millisecond range (e.g. hitting a ball when playing tennis). Our ability to continuously create predictions based on spatial and temporal cues (‘spatio-temporal prediction’) aids us to make timed, coordinated movements and perceptual judgments in relation to changes in our environment. Using functional magnetic resonance imaging (fMRI), we have shown that co-activation of cerebellum and hippocampus occurred during spatio-temporal prediction when healthy participants were asked to make precisely timed finger movements based on moving visual cues, whereas no co-activation occurred during reactive timing or motor imagery tasks [1]. Interestingly, the participants tended to press buttons slightly prior to the optimal timing of visual cues as learning progressed, indicating anticipation in their responses. Those findings suggest that the cerebellum and hippocampus cooperate to plan ahead and execute precise spatial and temporal motor responses. The exact cerebellar contribution to temporal aspects in this task however, remains uncertain. A previous study has shown that patients with spinocerebellar ataxia (SCA) are impaired at intercepting a moving target by performing a timed button press to launch another moving object for collision, a task also requiring spatio-temporal prediction [2]. Only patients with disorders that affect the cerebellum such as SCA type 6 and 8 and head essential tremor (ET), scored significantly worse on this predictive timing task when compared to those with other neurological disorders, such as Parkinson’s disease [3]. During the same task, higher blood oxygenation level-dependent (BOLD) signals in several cerebellar regions as well as thalamus and multiple cortical areas were associated with successful performance. In fact, the activity change during successful trials in several cerebellar regions was found to be higher in controls compared to SCA patients, indicating a role for the cerebellum in spatio-temporal prediction during target interception [4]. In a velocity judgement task that required purely perceptual spatio-temporal prediction, the posterior cerebellar lobule VII crus I was engaged when both temporal and spatial cues, but not when only spatial cues were used. Psychophysiological interactions (PPI) analysis further revealed that connectivity between the posterior cerebellum and several other upstream areas in the brain increased during spatio-temporal prediction in the perceptual domain [5]. This suggests that the cerebellum is important when integration of temporal information is required to complete the task goal. Pioneering work to determine cerebellar contribution to timing processes has been done by Ivry and colleagues [6–9]. Their work has received support from multiple studies in which timing elements of tasks were used explicitly or implicitly, depending on whether participants were required to use an overt estimation of time or to use time as a by-product to reach the task goal, respectively [10]. Deficits at both explicit timing tasks (e.g. temporal discrimination, temporal reproduction and synchronized repetitive finger-tapping tasks) and implicit timing tasks (e.g. spatio-temporal trajectory prediction and serial prediction tasks) have been found in patients with cerebellar disorders [2,3,6,11–16]. These findings complement evidence from neuroimaging studies showing cerebellar activation associated with temporal processing [4,17–19]. Disrupted cerebellar activity would likely be accountable for the observed deficits in these patients. Tasks employing repetitive transcranial magnetic stimulation (rTMS) furthermore show that temporal processing can be disrupted by interfering with cerebellar activity [20,21]. Together, these studies strongly implicate the cerebellum as part of a wider neuronal network involved in timing functions [9]. In the present study, we investigated the role of the cerebellum in spatio-temporal prediction of moving visual stimuli. We tested patients with spinocerebellar ataxia type 6 (SCA6) and healthy age-matched controls on two related motor tasks: a predictive and a reactive motor timing task. This allowed us to study cerebellar contribution to predictive and reactive motor timing functions separately. We further employed an experimental design in which we randomized the sequence order of markers and temporal intervals between them so as to investigate cerebellar involvement in sequence order and temporal interval learning. We hypothesized that cerebellar patients would show a reduced temporal precision in responses, coinciding with reduced performance on the task requiring spatio-temporal prediction, but not on the reactive timing task. Furthermore we expected that cerebellar patients would be impaired at both sequence order and temporal interval learning. SCA6 patients form an adequate subject group to test these hypotheses, since the atrophy specifically affects the cerebellum in adults [22,23]. SCA6 is an autosomal dominant genetic disorder caused by a CAG repeat expansion in exon 47 of the CACNA1A gene, which encodes the α1A (Cav2.1) subunit of neuronal P/Q-type voltage-gated calcium channel. This causes essentially pure cerebellar atrophy with a profound loss of Purkinje cells [24] and manifests with a range of symptoms including imbalance, dysarthria, gait ataxia, upper limb incoordination and tremor [25,26]. Our main finding is that SCA6 patients failed to establish spatio-temporal prediction in timing of their responses, reflecting a deficit in motor anticipation to upcoming events. Whereas healthy participants timed their button presses on average before the optimal timing with better temporal precision, this anticipation was absent in SCA6 patients. Performance of patients was reduced during the predictive timing task, but this deficit was limited to early correct responses that were made during the anticipatory period of the task. On the reactive timing task, randomization of sequence order of stimuli and temporal intervals between them resulted in reduced performance of the healthy controls. In contrast, randomization of temporal intervals did not further reduce performance in SCA6 patients. In summary, our findings suggest that cerebellar dysfunction may impair spatio-temporal prediction and prevent temporal interval learning to occur. These findings lend support to leading views on cerebellar involvement in timing functions. Subjects and Methods Participants We tested 19 patients with spinocerebellar ataxia type 6 (SCA6) (mean age 61.8 ± 7.2, standard deviation (SD), range 49–80) and 14 healthy control participants (mean age 64.5 ± 4.7, SD, range 58–74) for this experiment. We attempted to achieve homogeneity across participants by using several exclusion criteria, including suffering from fatigue, neuropsychological disorders or expertise with musical instruments, since our task requires skills often used in, for example, piano and guitar playing. We excluded 7 patients from our analysis based on the occurrence of severe tremors (unilateral or bilateral) and/or excessive fatigue and 2 control participants based on technical issues of the behavioral task during the experiment. We included the remaining participants: 12 SCA6 patients (3 males; 2 left-handed) and 12 healthy control participants (7 males; 2 left-handed) in our data analysis. The average age of these SCA6 patients was 60.3 ± 6.4, SD (range 49–67) and 64.8 ± 4.8, SD (range 58–74) of control participants [t22 = 1.985, P = 0.06, two-sample t-test]. We did not detect significant correlations between age and performance on the predictive timing task (r = 0.139, P = 0.517, Pearson correlation) or performance on the reactive timing task (r = -0.213, P = 0.318, Pearson correlation). All participants met the criteria of normal or corrected-to-normal vision, no excessive computer gaming or playing musical instruments for more than five hours a week and no history of other neuropsychological or motor disorders at the time of measurements or in the past. Participants were instructed prior to measurements and a written informed consent was obtained from all participants. A neurological exam was performed by a certified neurologist prior to the experiment to obtain a measurement of severity of cerebellar ataxia, according to the guidelines of the Scale for the Assessment and Rating of Ataxia (SARA) [27]. The average SARA scores in SCA6 patients was 9.5 ± 5.2, SD on a scale of 0–40. This score is the average of scores on eight different subtasks, comprising an assessment of gait, stance, sitting, speech, finger and hand movements, and leg coordination. A score of 0 indicates that no neurological manifestations could be observed, whereas a score of 40 reflects the occurrence of severe neurological manifestations [27]. This study was approved by the medical ethical committee of Erasmus MC, Rotterdam (MEC-2013-095). Behavioral task All participants completed the entire experiment, which included two modified versions of the Serial Interception Sequence Learning task [1,28]: a predictive timing task and a reactive timing task. In both tasks (Fig 1A and 1B), four empty white circles were placed on top of a gray rectangular background screen. Each of the white circles corresponded to four keys on a standard Qwerty-keyboard, which in turn were associated to a pre-determined finger: the ‘e’ key corresponded to the left middle finger, the ‘f’ key to the left index, the ‘j’ key to the right index and the ‘o’ key to the right middle. In the predictive timing task, participants were instructed to press the corresponding button on the keyboard at the moment upward moving black circles (moving markers) completely overlapped with the corresponding static white circles (target markers) at the top of the screen (Fig 1A). Only one moving marker could overlap with a target marker at a given moment. The diameter of the moving markers was 20 pixels smaller than that of the target marker to fit completely inside the target markers and not to interfere with the moment of the complete overlap between markers as observed by the participant. Moving markers scrolled from the bottom of the screen to the top at a constant velocity (2.78 milliseconds per pixel). In the reactive timing task, participants were instructed to press the corresponding button on the keyboard as quickly as possible when the white target marker changed (flashed) to red for 200 milliseconds (Fig 1B). Only one target marker could flash at a given moment. Between each trial, a short task instruction appeared on the screen for 3 seconds: ‘Press on the corresponding button with your finger as soon as the circles entirely overlap’ and ‘Press on the corresponding button with your finger as soon as a circle flashes’ (translated from Dutch), during the predictive and reactive task, respectively. The duration of this task instruction formed the only inter-trial interval time. Between each session, which lasted on average 7.9 ± 0.43 (SD) minutes participants took a short break (1–2 minutes) to reduce physical and mental fatigue. No feedback to indicate correct responses was given during the task. Both tasks were programmed in MATLAB R2010a (MathWorks, Massachusetts, USA) using Psychtoolbox-3 [29,30]. 10.1371/journal.pone.0162042.g001Fig 1 Behavioral task and experimental conditions. (A) During the predictive timing task black markers move from the bottom towards the four white target markers at the top of the screen. Participants are instructed to press the corresponding key at the moment that the black moving marker fully overlaps with the target marker (red square highlights the corresponding key). Target markers (from left to right) correspond to the ‘e’, ‘f’, ‘j’ and ‘o’ keys of a keyboard. (B) In the reactive timing task the white target markers change color to red for 200ms and participants are instructed to press the corresponding key as fast as possible. (C) Trials have one of four possible task conditions based on the sequence of and temporal interval between markers (fixed or randomized). (D) Both in the reactive timing task and predictive timing task, participants first complete a training session containing 8 trials of the FIX-FIX condition (i.e. fixed sequence and fixed interval), followed by two experimental sessions in which 2 trials of each condition are presented in a pseudo-randomized order. Task conditions We studied the role of the cerebellum in sequence and temporal interval learning by employing in both tasks a 2-by-2 factorial task design with two factors: the sequential order of stimuli and temporal interval between them. The sequences of moving markers (or marker flashes) and the temporal intervals between stimuli were either according to a pre-defined order (FIX) or shuffled based on the pre-defined order (RANDOM) (Fig 1C). The combination of these resulted in four task conditions: FIX-FIX (i.e. fixed sequence and fixed intervals), FIX-RANDOM, RANDOM-FIX and RANDOM-RANDOM. For each condition, a sequence of 12 circles was presented 4 consecutive times, which resulted in 48 moving or flashing markers that were presented during each trial. In FIX-FIX condition of the predictive timing task, the sequence of moving markers followed the same order 4 consecutive times: F1386ms-E718ms-J852ms-E852ms-O852ms-E702ms-J969ms-F952ms-J1887ms-O668ms-F1420ms-O668ms (E: left-most trajectory, F: second trajectory, J: third trajectory, O: right-most trajectory, subscripted rates: temporal differences between moving markers in milliseconds). In FIX-FIX condition of the reactive timing task, the sequence of the marker flashes was E707ms-F1345ms-E1078ms-O849ms-J1200ms-F703ms-O1490ms-E844ms-J1824ms-O1180ms-F811ms-J1075ms. The average temporal interval between appearances was 1001 milliseconds (range 668–1887 milliseconds) in the predictive timing task and 1092 milliseconds (range 703–1824 milliseconds) in the reactive timing task. Experimental protocol All experiments were conducted in the participants’ homes. A 13 or 15 inch computer screen was used to present the visual stimuli with a resolution of 1280 x 800 pixels (13 inch screen: all SCA6 patients and 4 out of 12 control participants). To rule out that screen size had an effect on performance, we compared performance on both tasks of controls tested using a 13 inch screen (N = 4) versus controls tested using a 15 inch screen (N = 8). We did not find significant differences between these controls on both the predictive timing task [t10 = 0.981, P = 0.35, two-sample t-test] and on the reactive timing task [t10 = 0.869, P = 0.405]. Participants were instructed prior to the experiment by an information letter mailed in advance. The tasks were also verbally explained to the participants right before starting the tasks. A summary of the instructions was finally displayed on the screen before the task began. When participants confirmed they fully understood the tasks, we started the experiment. Each task consisted of one training session and two experimental sessions. During the training session, eight trials with the FIX-FIX condition were presented. In each experimental session, each task condition was presented twice in a pseudo-randomized order using randomization algorithms in MATLAB (Fig 1D). All SCA6 patients and 5 out of 12 control participants started with the predictive timing task, followed by the reactive timing task as part of an experimental sequence. The remaining control participants (N = 7) first started with the reactive timing task, directly followed by the predictive timing task. To rule out that the order of tasks had an effect on performance, we compared performance on both tasks of controls tested first on the predictive task (N = 5) versus controls tested first on the reactive task (N = 7). We did not find significant differences between these controls on both the predictive timing task [t10 = 0.157, P = 0.878, two-sample t-test] and on the reactive timing task [t10 = 0.727, P = 0.484, two-sample t-test]. Analysis of behavioral data Classification of correct responses depended on two factors, pressing the corresponding button and pressing the button within a given time window. To maximize the sensitivity of analyzing responses within a set time window, we calculated performance as the area under the curve of the correct ratio (AUCCR). The correct ratio was defined as the number of corresponding (correct) button presses within a time window, divided by the total number of markers within a trial. The minimum temporal interval between markers was 668 milliseconds in the predictive timing task and 703 milliseconds in the reactive timing task. Based on these minimum times, we used a time window between -334ms to +334ms relative to perfect marker overlap in the predictive timing task and 0 to 703ms after marker flash in the reactive timing task throughout our analysis. To assess spatio-temporal prediction in the predictive timing task, we calculated the AUCCR before marker overlap (-334ms to < 0ms) representing early correct responses and after marker overlap (> 0ms to +334ms), representing late correct responses. We ascertained normality of data distributions using the Shapiro-Wilk test. Unless indicated otherwise, we used one-way repeated measures analyses of variance (ANOVA) with trial number (for training sessions) or trial condition (for experimental sessions) as within-subject variable and group as between-subject variable to test AUCCR on both reactive and predictive tasks. We used the Mauchly’s test to test the assumption of sphericity and in case this assumption was violated, we applied the Greenhouse-Geisser correction. Levene’s test of equality of error variances was used to test whether the assumption of equal error variances was met. To analyze pooled response and reaction times, we used the non-parametric Wilcoxon rank sum test and Wilcoxon signed rank test with Bonferroni correction for multiple comparisons, since these distributions violated the assumption of normality. We compared variances of these distributions using the Brown-Forsythe test. Comparison of the first principal component (PC1) of the AUCCR of both tasks was done using independent-samples t-tests and a correlation analysis was performed using the Pearson correlation. These performance measures were also used in the methods section to test for effects of age, screen size and task order. Statistics were conducted in SPSS version 22 (IBM, New York, USA) and MATLAB R2011b (MathWorks, Massachusetts, USA). Results are reported as mean ± standard error of the mean (SEM), unless indicated otherwise. Results SCA6 patients show lower predictive motor timing performance, localized to early correct responses We used area under the curve of the correct ratio (AUCCR) to analyze performance of responses made within a temporal window around complete marker overlap (-334ms to +334ms; all correct responses). Over the course of the training session, both SCA6 patients and controls learned the sequence and timings of the markers (Fig 2A). Controls showed a significantly higher AUCCR compared to SCA6 patients during the training session, indicating that controls made more correct button presses within the assigned time window [main effect of trials: F7,154 = 7.44, P < 0.001; main effect of group: F1,22 = 10.23, P < 0.01; interaction trials x group: F7,154 = 0.87, P = 0.53]. To assess the effect of spatio-temporal prediction in our task, we segregated the data based on the time of button press. Early and late correct responses are corresponding button responses made before or after the moving marker completely overlapped with the target marker, respectively (Fig 2B and 2C). When analyzing early correct responses we found that controls showed a significantly higher AUCCR compared to SCA6 patients, although both groups showed a learning effect [main effect of trials: F7,154 = 10.67, P < 0.001; main effect of group: F1,22 = 15.56, P = 0.001; interaction trials x group: F7,154 = 1.25, P = 0.28]. In contrast, for late correct responses we found no significant differences in AUCCR between groups and no learning effect [main effect of trials: F3.313, 72.883 = 0.71, P = 0.56; main effect of group: F1,22 = 0.02, P = 0.89; interaction trials x group: F3.313, 72.883 = 2.4, P = 0.23, with Greenhouse-Geisser correction]. A third comparison including all factors revealed a significant interaction between groups and type of correct response (early versus late), confirming that the observed group difference was localized to early correct responses only [F1, 22 = 10.09, P = 0.004, with Greenhouse-Geisser correction; two-way repeated measures ANOVA]. In short, our data shows a performance difference between SCA6 patients and healthy controls during training on the predictive timing task, which is localized to (early) correct responses made during the anticipatory period of the paradigm. 10.1371/journal.pone.0162042.g002Fig 2 Results training session of predictive timing task. (A) Task performance showing all correct responses, expressed as area under the curve of the correct ratio (AUCCR) for both controls and SCA6 patients, showing a learning effect and a difference between groups. (B) AUCCR of early correct responses only, where the learning effect and group differences are present. (C) AUCCR of late correct responses only, where no learning effect or group differences are present. (D) Response times distributions of correct button presses of controls. Average response times are significantly lower than 0ms on all trials (asterisks). (E) Response times distributions of correct button presses of SCA6 patients. Average response times are not significantly different from 0ms on any of the trials. (F) Distribution of pooled response times for all training trials, showing that controls time their button press more precisely as observed by a smaller variance compared to SCA6 patients (asterisks). (A, B, C) Error bars represent SEM, ** indicates P < 0.01 for main effects of trials (asterisks horizontal line) and main effects of group (asterisks vertical line). (D, E) White dots represent the mean and black bars represent the SD. (F) Histogram shows 16.7ms bins. * indicates P < 0.05, ** indicates P < 0.01. SCA6 patients show impaired anticipation in motor timing To gain insight in the characteristics of motor timing in the predictive task, we visualized response time distributions of correct button presses of both controls and SCA6 patients (Fig 2D and 2E). Negative and positive response times correspond to button presses before and after perfect marker overlap, respectively. We pooled response times of correct button presses from all participants per group, since some SCA6 patients had only few correct responses per trial. This resulted in analyses based on a fixed template design. We calculated average response times for each trial of the training session: trial 1 [SCA6: 2.7 ± 154.8ms, control: -20.2 ± 120.3ms; mean ± SD], trial 2 [SCA6: -15.9 ± 148.3ms, control -51.9 ± 113.3ms], trial 3 [SCA6: -14.8 ± 150.3ms, control: -33.9 ± 122.2ms], trial 4 [SCA6: 12.1 ± 157.4ms, control: -49.1 ± 110ms], trial 5 [SCA6: 4.5 ± 152.8ms, control: -53.2 ± 116.6ms], trial 6 [SCA6: 6.9 ± 151.2ms, control: -55.9 ± 117.8ms], trial 7 [SCA6: 1.7 ± 147ms, control: -71.1 ± 116.3ms] and trial 8 [SCA6: -1.4 ± 147ms, control: -65.8 ± 113.5ms]. We found that response times of control participants were significantly lower than those of SCA6 participants on all trials, except on trial 1 and 3 [trial 1: mean difference (MD) = 22.9ms, Z = 2.42, P = 0.12; trial 2: MD = 36ms, Z = 4.26, P < 0.001; trial 3: MD = 19.1ms, Z = 2.53, P = 0.09; trial 4: MD = 61.2ms, Z = 7.0, P < 0.001; trial 5: MD = 57.7ms, Z = 6.22, P < 0.001; trial 6: MD = 62.8ms, Z = 7.24, P < 0.001; trial 7: MD = 72.8ms, Z = 8.37, P < 0.001; trial 8: MD = 64.4ms, Z = 7.31, P < 0.001; Wilcoxon rank sum test, with Bonferroni correction]. Previous findings have indicated that healthy participants establish spatio-temporal prediction during the training session of this predictive timing task [1], as indicated by average response times lower than 0ms. Indeed, this effect was also present in our control group, since we found that response times were significantly lower than 0ms on all trials of the training session [trial 1: Z = -2.84, P = 0.04; trial 2: Z = -9.51, P < 0.001; trial 3: Z = -6.5, P < 0.001; trial 4: Z = -9.46, P < 0.001; trial 5: Z = -9.85, P < 0.001; trial 6: Z = -10.16, P < 0.001; trial 7: Z = -12.13, P < 0.001; trial 8: Z = -11.65, P < 0.001; Wilcoxon signed rank test, with Bonferroni correction] (Fig 2D). In contrast, we did not find this effect in our SCA6 patient group, given that average response times of SCA6 patients were not different from 0ms on any of the training trials [trial 1: Z = -0.51, P = 1; trial 2: Z = -1.87, P = 1; trial 3: Z = -1.93, P = 1; trial 4: Z = -1.84, P = 1; trial 5: Z = -0.27, P = 1; trial 6: Z = -0.66, P = 1; trial 7: Z = -0.37, P = 1; trial 8: Z = -0.24, P = 1; Wilcoxon signed rank test, with Bonferroni correction] (Fig 2E). These findings indicate that spatio-temporal prediction is established during the training session in control participants, but not in SCA6 patients. SCA6 patients show reduced temporal precision in predictive motor timing Since participants were instructed to time their button presses as closely as possible to perfect marker overlap, we investigated whether both groups were able to correctly time their button press. Therefore, we assessed whether there were differences in response time distributions between groups with variance as a measure of precision. A lower variance of response times would indicate a higher precision, whereas a higher variance would suggest a more temporally distributed response pattern, indicating a lower precision. To address this question, we compared the variance of pooled response time distributions of the training session between groups using the Brown-Forsythe test of variances, since this test is relatively robust and insensitive to deviations from normality [31]. We found that SCA6 patients showed a significantly higher variance of response times compared to controls [SCA6: 2.284x104 ms2, control: 1.371x104 ms2, F1,5808.03 = 239.98, P < 0.001] (Fig 2F). To verify that this difference was also present during the experimental sessions, we compared the variance of pooled response times of correct responses during the experimental sessions and we found that also here controls had a significantly higher variance [SCA6: 2.195x104 ms2, control: 1.352x104 ms2, F1, 14580.35 = 239.98, P < 0.001]. These results indicate that correct button presses were more temporally distributed in the SCA6 group, corresponding to a lower temporal precision of responses during both training and experimental sessions. Together, these findings suggest that healthy participants anticipate the upcoming event of a marker overlap with better temporal precision, whereas this anticipation of motor timing from the spatial and temporal cues is impaired in SCA6 patients. Sequential or temporal order randomization of markers does not affect predictive motor timing performance We investigated sequential order and temporal interval learning by randomizing the order and temporal intervals of markers, resulting in four conditions that were presented during experimental sessions (Fig 1C and 1D). We calculated the AUCCR for both groups during different trial conditions: FIX-FIX condition [SCA6: 0.18 ± 0.01, control: 0.21 ± 0.013], FIX-RANDOM condition [SCA6: 0.178 ± 0.011, control: 0.208 ± 0.012], RANDOM-FIX condition [SCA6: 0.183 ± 0.011, control: 0.208 ± 0.011] and RANDOM-RANDOM condition [SCA6: 0.177 ± 0.01, control: 0.204 ± 0.011]. No significant differences in AUCCR between trial conditions or between groups were found, although the main effect of group approached the significance level and could be considered a trend (Fig 3A). Moreover, there was no interaction effect between conditions and groups [main effect of condition: F3,66 = 0.82, P = 0.49; main effect of group: F1,22 = 3.46, P = 0.08; interaction group x condition: F3,66 = 0.31, P = 0.82]. Therefore we could conclude from these data that randomizing sequences or temporal interval of markers did not lead to differences in performance or overall group differences when considering all correct responses. 10.1371/journal.pone.0162042.g003Fig 3 Results experimental sessions of predictive timing task. (A) Task performance on different trial conditions is visualized for both SCA6 patients and controls. No differences between groups or between conditions are observed in all correct responses. (B) AUCCR calculated for early correct responses shows a significant group difference between SCA6 patients and controls, but not between conditions. (C) AUCCR calculated for late correct responses does not show any differences between conditions or between groups. Error bars represent SEM, ** indicates P < 0.01. Since we had found group differences in predictive motor timing for early correct responses during training (Fig 2B), we hypothesized that specifically these ‘anticipatory’ responses could be affected by randomization of sequence or temporal interval of markers. Therefore, we analyzed the data based on the timing of button presses and compared performance between conditions and groups (Fig 3B and 3C). We calculated the AUCCR of early correct responses on the FIX-FIX condition [SCA6: 0.091 ± 0.012, control: 0.139 ± 0.011], FIX-RANDOM condition [SCA6: 0.096 ± 0.012, control: 0.141 ± 0.009], RANDOM-FIX condition [SCA6: 0.099 ± 0.01, control: 0.144 ± 0.009], and RANDOM-RANDOM condition [SCA6: 0.091 ± 0.008, control: 0.137 ± 0.007] and we found that AUCCR differed significantly between groups. However, randomizing the sequence or temporal interval of markers did not lead to performance differences [main effect of condition: F3,66 = 1.94, P = 0.132; main effect of group: F1,22 = 11.95, P = 0.002; interaction group x condition: F3,66 = 0.38, P = 0.99] (Fig 3B). In addition, we investigated whether randomization had an effect on performance for late correct responses (Fig 3C). We calculated the AUCCR during the FIX-FIX condition [SCA6: 0.084 ± 0.007, control: 0.069 ± 0.01], FIX-RANDOM condition [SCA6: 0.079 ± 0.009, control: 0.064 ± 0.009], RANDOM-FIX condition [SCA6: 0.08 ± 0.007, control: 0.061 ± 0.009], and RANDOM-RANDOM condition [SCA6: 0.082 ± 0.009, control: 0.065 ± 0.008] and performed the same analysis. We found that AUCCR of late correct responses did not differ between groups or between conditions. Also, the data did not reveal an interaction effect [main effect of condition: F3,66 = 1.25, P = 0.30; main effect of group: F1,22 = 2.09, P = 0.16; interaction group x condition: F3,66 = 0.17, P = 0.92]. A third analysis including all factors confirmed that the significant group difference was localized to early correct responses only [interaction group x type of correct response (early versus late): F1, 22 = 10.01, P = 0.004; two-way repeated measures ANOVA]. Thus, although there were group differences in early correct responses, randomizing the order and temporal intervals of markers did not influence predictive motor timing performance regardless of the timing of responses. Lower predictive motor timing performance of SCA6 patients is limited to responses during the anticipatory period To further compare performance of groups depending on the timing of response, but regardless of the trial condition, we conducted a principal component analysis (PCA) for dimension reduction. AUCCR scores for early correct responses between trial conditions were highly inter-correlated, at least r = 0.91 [P < 0.001, Pearson correlation]. Kaiser-Meyer-Olkin measure of sampling adequacy was 0.87, Bartlett’s test of sphericity was significant [χ2(6) = 142.67, P < 0.001] and all communalities were above 0.94, confirming that this data was suitable for PCA. We found that the first principal component (PC1) explained 95% of the variance and therefore we used this measure to represent performance of early correct responses. Next, we subjected AUCCR scores for late correct responses to PCA and consistently found a high inter-correlation between performance on different conditions [at least r = 0.85, P < 0.001, Pearson correlation]. Kaiser-Meyer-Olkin measure of sampling adequacy was 0.87, Bartlett’s test of sphericity was significant [χ2(6) = 101.11, P < 0.001] and all communalities were above 0.87. We found that the first principal component (PC1) explained 90.1% of the variance and therefore we used this measure to represent performance of late correct responses. We then used these measures to compare performance between groups, depending on the timing of response, regardless of the trial condition. We found a significant interaction effect between performance (PC1 of early vs late) and group [F1, 22 = 9.76, P = 0.005], although the main effect for group in this analysis did not reach the significance level [F1,22 = 3.65, P = 0.069]. Post-hoc tests confirmed that the group difference was localized to PC1 of early correct responses, but not to PC1 of late correct responses [early: t22 = 3.34, P = 0.006; late: t22 = -1.44, P = 0.328, two-sample t-test with Bonferroni correction]. In summary, although task performance was not influenced by randomization of sequences or temporal intervals, we observed a higher performance of early correct responses in controls only, which is in line with our results from the training session. SCA6 patients and controls perform comparably during reactive timing task training To investigate whether SCA6, in addition to predictive motor timing, also affected reactive motor timing functions, we tested all participants on a related reactive motor timing task where no prediction could be made based on visual cues. We calculated AUCCR of corresponding button presses made after marker flash (0-703ms) as performance measure during the training session, in which eight trials of the FIX-FIX condition were presented (Fig 1D). Since this task was purely reactive, we did not make a distinction between early and late responses. Both groups showed learning of the task, evidenced by an increase in AUCCR over the course of the training (Fig 4A). We did not observe a group difference, although P-value approximation to significance threshold suggested a trend. We also did not observe an interaction effect between groups and trials [main effect of trials: F7,154 = 9.88, P < 0.001; main effect of group: F1,22 = 3.47, P = 0.076; interaction trials x group: F7,154 = 1.41, P = 0.21]. In short, both groups increased their performance during the training, and overall there were no training differences between groups in this respect. 10.1371/journal.pone.0162042.g004Fig 4 Results of reactive timing task. (A) Task performance expressed as AUCCR of correct responses after marker flash for both controls and SCA6 patients, showing that learning occurs over the course of the trials (asterisks horizontal line). (B) AUCCR calculated for different trial conditions during experimental sessions. The data shows group differences, as well as differences between all conditions with the FIX-FIX condition in the control group, but only between the RANDOM-FIX and the FIX-FIX condition in the SCA6 group. (C) Reaction time distributions of correct button presses during training for controls and (D) for SCA6 patients. (F) Distribution of pooled reaction times for all training trials. Variance of the reaction time distribution was smaller in controls (asterisks). (A, B) Error bars represent SEM, * indicates P < 0.05, ** indicates P < 0.01 for main effects of trials (asterisks horizontal line panel A) and main effects of condition and group (asterisks panel B). (D, E) Dotted line indicates 500ms for visual aid, white dots represent the mean and black bars represent the SD. (F) Histogram shows 16.7ms bins. ** indicates P < 0.01. SCA6 patients and controls show differences in reaction times and precision during reactive timing task training We visualized reaction time distributions of the reactive timing task, showing development of reaction times of both groups over the training session (Fig 4C and 4D). We then calculated average reaction times of pooled correct responses on all trials of the training session: trial 1 [SCA6: 518.9 ± 113.8ms, control: 499.8 ± 107ms; mean ± SD], trial 2 [SCA6: 506.8 ± 115.8ms, control 477.4 ± 103.4ms], trial 3 [SCA6: 510 ± 107.6ms, control: 471.5 ± 102ms], trial 4 [SCA6: 521.6 ± 115.9ms, control: 480.4 ± 105.3ms], trial 5 [SCA6: 529.4 ± 111ms, control: 473.1 ± 108.8ms], trial 6 [SCA6: 524.2 ± 104ms, control: 471.7 ± 110.1ms], trial 7 [SCA6: 522.4 ± 115.5ms, control: 475.2 ± 112.2ms] and trial 8 [SCA6: 510.4 ± 112.9ms, control: 474.3 ± 115ms]. Significantly higher reaction times for SCA6 patients were observed for all trials, except for trial 1 [trial 1: mean difference (MD) = 19.1ms, Z = 2.49, P = 0.1; trial 2: MD = 29.4ms, Z = 4.34, P < 0.001; trial 3: MD = 38.5ms, Z = 5.64, P < 0.001; trial 4: MD = 41.1ms, Z = 5.56, P < 0.001; trial 5: MD = 56.3ms, Z = 7.62, P < 0.001; trial 6: MD = 52.5ms, Z = 7.3, P < 0.001; trial 7: MD = 47.2ms, Z = 6.2, P < 0.001; trial 8: MD = 36.1ms, Z = 4.78, P < 0.001; Wilcoxon rank sum test, with Bonferroni correction]. Next, we compared variances of pooled reaction times of correct button presses during training between groups and found a small but significant difference in variance, in which the SCA6 patient group showed a higher variance [SCA6: 1.258x104 ms2, control: 1.172x104 ms2, F1,7107.19 = 249.6, P < 0.001, Brown-Forsythe test of variances] (Fig 4E). We did the same analysis for pooled reaction times during the experimental sessions and found a consistent difference in variance between groups [SCA6: 1.342x104 ms2, control: 1.163x104 ms2, F1,13277.38 = 200.21, P < 0.001]. Together, these results indicate that average reaction times were significantly higher for SCA6 patients compared to controls. Furthermore, SCA6 patients also consistently showed a mildly reduced temporal precision, based on the observation that correct responses were more widely temporally distributed within this group. SCA6 patients show impaired learning of temporal interval order We hypothesized that SCA6 patients would be impaired at learning sequences and temporal intervals of stimuli on the reactive timing task. In addition, controls could potentially use this type of learning as strategy to gain increased performance on conditions using fixed sequence or temporal intervals. To test this hypothesis, we visualized the AUCCR during FIX-FIX condition [SCA6: 0.152 ± 0.02, control: 0.231 ± 0.016], FIX-RANDOM condition [SCA6: 0.153 ± 0.02, control: 0.215 ± 0.017], RANDOM-FIX condition [SCA6: 0.134 ± 0.015, control: 0.181 ± 0.014], and RANDOM-RANDOM condition [SCA6: 0.138 ± 0.014, control: 0.193 ± 0.017] (Fig 4B). Using a repeated measures ANOVA we found a main effect of group, as well as a main effect of condition and an interaction effect [main effect of condition: F1.71, 37.56 = 22.82, P < 0.001; main effect of group: F1,22 = 7.1, P = 0.014; interaction condition x group: F1.71,37.56 = 4.12, P = 0.03, with Greenhouse-Geisser correction]. Closer inspection of only SCA6 performance furthermore showed a main effect of condition [F1.56, 17.15 = 4.29, P < 0.039, with Greenhouse-Geisser correction]. Planned contrasts comparing FIX-FIX condition with the other conditions revealed that the difference was located between the FIX-FIX condition and the RANDOM-FIX condition [F1,11 = 6.23, P = 0.03]. Performance of control participants also showed a significant main effect of condition [F3,33 = 23.45, P < 0.001] and contrasts indicated that AUCCR on the FIX-FIX condition was significantly higher compared to all other conditions [versus FIX-RANDOM: F1,11 = 15.49, P = 0.002; versus RANDOM-FIX: F1,11 = 72.05, P < 0.001; versus RANDOM-RANDOM: F1,11 = 21.02, P = 0.001] (Fig 4B). Together, these results show that although overall performance was reduced in the SCA6 patient group, randomizing the sequential order of stimuli resulted in reduced performance in both groups, suggesting that sequential order learning occurred in both groups. In contrast, randomizing the temporal interval order resulted in reduced performance only in the control group, but not in the SCA6 patient group. This suggests that only controls learned the temporal order of stimuli sufficiently, whereas this type of learning was reduced in SCA6 patients. Complementary to the predictive timing task, we performed a PCA on AUCCR scores of all trial conditions to compare performance on groups independent of trial condition. Scores on all conditions were highly inter-correlated, at least r = 0.92 [P < 0.001, Pearson correlation]. Kaiser-Meyer-Olkin measure of sampling adequacy was 0.81, Bartlett’s test of sphericity was significant [χ2(6) = 173.12, P < 0.001] and all communalities were above 0.93, confirming that this data was suitable for PCA. We found that PC1 explained 96.2% of the variance and therefore we used this measure to represent performance on the experimental sessions of the reactive timing task. Confirming our earlier findings, PC1 differed significantly between groups [t22 = 2.69, P = 0.013, two-sample t-test]. Performance on predictive and reactive motor timing tasks does not correlate with SARA scores We hypothesized that if predictive and reactive motor timing functions are linked to cerebellar function, we could possibly find a relation between SCA6 symptom severity and performance on the tasks. To test this hypothesis, we performed an exploratory correlation analysis between performance on the predictive task and reactive timing task with individual scores on the Scale for the Assessment and Rating of Ataxia (SARA), which we obtained from 11 of the 12 SCA6 participants. Performance of both tasks was represented as the PC1 of scores on each task, as explained earlier. We did not find significant correlations between SARA score and scores on the predictive timing task (PC1 all correct responses) [r = -0.07, P = 0.85, Pearson correlation], PC1 of early correct responses [r = -0.509, P = 0.11, Pearson correlation], PC1 of late correct responses [r = 0.36, P = 0.27, Pearson correlation] and PC1 of reactive timing task [r = -0.5, P = 0.12, Pearson correlation]. Discussion In this study, we subjected cerebellar patients and control subjects to two finger-movement timing paradigms to find out to what extent the cerebellum contributes to spatio-temporal prediction and whether cerebellar dysfunction leads to changes in motor timing and coordination. We found that SCA6 patients were impaired at establishing spatio-temporal prediction of finger movements based on dynamic visual stimuli. Healthy control participants precisely timed their button-presses before a moving stimulus completely overlapped with a target stimulus, thereby anticipating or even ‘over-predicting’ their motor timing with approximately 66ms at the end of the training session, a behavioral effect that has also been observed previously [1]. Instead, cerebellar patients failed to establish spatio-temporal prediction and placed their button presses generally more temporally distributed, indicative of reduced temporal precision in responses. Concomitantly, SCA6 patients had significantly lower performance of early correct responses placed in the anticipatory period, but not of late correct responses placed in the period after the moving stimulus overlapped with the target stimulus. Our results support previous work that has associated the cerebellum with spatio-temporal prediction processes and substantiate recent proposals arguing that the cerebellum is responsible for ‘monitoring’ ongoing timing and adjustment based on temporal predictions [32]. Striking similarities can be found with a study in which cerebellar patients and healthy participants had to precisely time their button press to intercept a moving target with a moving ball [2]. Lower hit-rates were observed in cerebellar patients and errors were equally distributed between (too) early and late responses. Interestingly, both this study and our results show a higher variability in patient response times, although in this study the variability was found to be limited to the late error trials, whereas we did not make such a distinction in our study. Using the same paradigm, cerebellar activations were found to be related to performance and were reduced in SCA patients [4]. Both these and our data indicate a predictive sub-second motor timing deficit in cerebellar patients, which could lead to impaired motor anticipation in daily life situations. For instance, Lang and Bastian (1999) showed that cerebellar patients failed to show anticipatory muscle activity when catching a falling weight, which made them unable to control the impact of the falling weight [33]. In a more general context, a considerable amount of literature has associated the cerebellum with temporal processing in the sub-second range, which are processes that are also engaged during predictive motor timing. A compelling example of this is that cerebellar patients show increased variation in timing, amplitude and velocity of finger opening in overarm throwing of a ball, which consequently results in reduced accuracy of throwing [34]. Moreover, cerebellar lesions in humans, as well as in rodents, affect the acquisition and particularly the timing of conditioned responses in Pavlovian eye blink conditioning [35], highlighting that the cerebellar cortex is specifically involved in the time course and amplitude of this learned motor behavior [36,37]. Other studies employing explicit timing tasks, such as rhythmic finger tapping and temporal interval discrimination and reproduction, have provided convincing evidence for cerebellar involvement in sub-second temporal processing. An increased temporal variability during rhythmic tapping with the finger or foot has been found in cerebellar patients [6,11], particularly associated with lesions in the lateral cerebellar regions [12]. Neuroimaging studies have found reliable activation of the cerebellum among several other areas in the cerebello-diencephalic-parietal network during finger-tapping in synchrony or continuation after tones at constant intervals [17–19]. Consistent with this finding, inactivating the ipsilateral cerebellar regions using repetitive transcranial magnetic stimulation (rTMS) introduced variability in the inter-tap interval [20]. Moreover, impaired temporal discrimination in the sub-second range has been found in both cerebellar patients and animal studies [6,13,14,38] (but see also: [39,40]). Gooch et al. (2010) however did not find impairments in temporal discrimination, although they did find impaired temporal estimation, production and reproduction in cerebellar lesion patients [41]. The reproduction of temporal intervals furthermore recruits cerebellar regions [15,16]. This cerebellar activation commonly coincides with activity in other cortical areas, e.g. basal ganglia, supplementary motor area (SMA), premotor area and inferior parietal cortex. Inactivating the left and right lateral cerebellum with rTMS impairs performance at a temporal reproduction task on short interval (millisecond range), but not at second range intervals [21]. The present view posits that the cerebellum is particularly involved in temporal processing at the sub-second scale, whereas the basal ganglia as part of a wider cortical network processes information at longer timescales [7,15,16,21,42–45]. This view is however still debated, as some researchers claim that the precise role of the cerebellum in timing processes has not been proven so far [46–48]. To assess the relationship between the performance of both tasks and indicators of SCA6 disease progression, we conducted a correlation analysis including the first PC as a measure of task performance in both predictive and reactive timing tasks and scores on the SARA [27]. Against expectations we failed to find a correlation between performance and SARA score. A possible explanation for this is that SARA score is based on multiple sub-scores assessing a wide range of functions, such as walking, stance, sitting and speech, whereas performance on our tasks depends specifically on the ability to precisely coordinate finger movements and to make precisely timed motor executions. A high SARA score does not necessarily translate into reduced fine motor skills, but could be the result of a broad spectrum of symptoms. Not finding a correlation could be explained by the fact that we had a rather small amount of participants for this exploratory correlation analysis, and that the SARA scores of our SCA6 participants were relatively low. Increasing the amount of participants with higher SARA scores could still reveal such an underlying relationship between task performance and SCA6 symptom severity. In our reactive motor timing paradigm, we found that cerebellar patients overall scored worse than controls, which could indicate a general motor coordination deficit. Nevertheless, randomizing the sequence of stimuli resulted in a lower performance in both groups, indicating that sequence learning did occur in both SCA6 and control participants. Randomizing the temporal interval between markers, however, did not affect the performance of SCA6 patients, suggesting that temporal interval learning is impaired or that a deficit exists in the integration of spatial and temporal information in cerebellar patients. Learning of the sequence and/or temporal interval of stimuli could potentially aid in creating an expectation of which button to press and when to make the motor execution, respectively. This spatial and temporal expectation could be used within a strategy used during both the predictive and the reactive timing task, although randomization only resulted in performance differences during the reactive timing task. Learning of new motor sequences has been associated with enhanced activity in several brain areas, including the prefrontal cortex (PFC), putamen, intraparietal sulcus region (IPS), precuneus, premotor cortex (PMC), supplementary motor area (SMA) and cerebellum [49–57] (but see:[58]). It has been shown in multiple studies that during the initial stages of motor sequence learning frontal areas in combination with bilateral cerebellar regions show enhanced activation, after which the representation shifts to activation within the cortical-striatal circuit when the sequence has been learned [50,55–57]. Also, a shift in activity from the cerebellar cortex to deep cerebellar nuclei with sequence learning has been shown, suggesting segregation of neural networks involved at different stages of motor sequence learning [50]. Interestingly, the hippocampus has been shown to be active during learning of sequences [59], but also during retrieval of learned sequences [60]. A relationship between neural activity of hippocampal CA1 neurons and nonspatial sequence coding of odors has been found in rats, thus implying a role for the hippocampus in coding for sequential position of nonspatial objects [61]. These hippocampal related findings are interesting in the light of our recent observations of cerebellar-hippocampal interaction during spatio-temporal prediction [1], although the former sequences may not be related to timing at the millisecond level. Within the temporal domain, sustained activity in the lateral cerebellum and several structures within the neocortical-cerebellar network has been associated with temporal sequence learning in the form of intervals and rhythm [49,62,63]. Based on the cerebellar involvement in these forms of sequence learning, it seems likely that the observed deficits in temporal order learning can be attributed to cerebellar Purkinje cell specific degeneration in SCA6. These findings are in agreement with a previous study showing that cerebellar patients are indeed impaired on temporal, but also spatial sequence learning in a serial reaction time task [64]. Still, it should be noted that we were unable to detect sequence learning in cerebellar patients and healthy controls in our paradigm where spatio-temporal cues were available for motor prediction. There are several limitations to our experimental approach. First, SCA6 covers a wide spectrum of symptoms [25,26], among others poor motor coordination of upper limbs, tremors, physical fatigue and oculomotor deficits. All of these may have affected participants’ ability to execute the tasks, since performance on both tasks was not only dependent on motor timing (i.e. pressing the button on time or as fast as possible), but also depended on precise finger movement coordination for pressing the right button. The possible influence of reduced motor coordination on predictive timing performance in SCA6 patients, as shown by reactive timing group differences, cannot be fully discounted. We attempted to minimize this influence by designing the paradigm in a way that movement of only one finger at a time was required instead of multiple fingers at the same time, and that the movement-speed of markers was acceptable for both SCA6 and control participants. As such, SCA6 patients showed increasing performance during the training on both tasks, which could indicate that reduced motor coordination has played a relatively limited role in our predictive task. We furthermore excluded SCA6 patients exhibiting unilateral or bilateral tremors or pronounced physical fatigue, thereby attempting to reduce these additional influences. Consequently, this reduced the average SARA score in our patient group. Our findings suggest however that cerebellar patients with relatively mild symptoms (i.e. having a low SARA score) already exhibit pronounced deficits in predictive motor timing functions. Attentional deficits may also have played a role in our study. Cerebellar patients have been shown to exhibit deficits in shifting attention between visual stimuli [65,66], although another study argues that response-related preparation processes rather than visuospatial attention shift are affected in cerebellar patients [67]. Both of these are potentially of importance, particularly during motor sequence learning, since differences in brain activation have been found between implicitly and explicitly learned motor sequences, suggesting that a different set of brain structures is activated depending on the degree of attentional resources assigned to the learned motor sequence [52]. Furthermore, cerebellar activity has been shown during a verbal working memory task [68] and a correlation between working memory and performance on temporal estimation, production and reproduction tasks has been found in subjects with cerebellar lesions [41], although it remains unclear to what extent SCA6 affects working memory in our patients. The potential influence of oculomotor deficits can also not be fully discounted. Future research using eye-tracking hardware may clarify to what extent eye movement deficits in SCA6 patients affect predictive motor timing. Recently it has been shown that in the mouse inducing the expression of C terminus polypeptides in Purkinje cells, as seen in SCA6 patients, yields both physiological and behavioral consequences, and may provide us with a new mouse model to study disease-related mechanisms of SCA6 in greater depth [23]. To conclude, we hereby provided evidence that SCA6 patients are impaired at establishing spatio-temporal prediction in timing of responses based on dynamic visual stimuli, reflecting a deficit in integration of spatial and temporal information and subsequently motor anticipation to upcoming events, using a task that demanded a prediction of when a stimulus would be at a certain target location. We propose that SCA6 patients are impaired in updating their forward internal model using spatial and temporal cues in the sub-second range, a process that has been postulated to require the cerebellum [69,70]. We thank Drs. E. Brusse and R.S. van der Giessen for their clinical contribution to the study. ==== Refs References 1 Onuki Y , Van Someren EJW , De Zeeuw CI , Van Der Werf YD . Hippocampal-cerebellar interaction during spatio-temporal prediction . Cereb. Cortex . 2015 ;25 : 313 –321 . 10.1093/cercor/bht221 23968839 2 Bareš M , Lungu OV , Liu T , Waechter T , Gomez CM , Ashe J . Impaired predictive motor timing in patients with cerebellar disorders . Exp. Brain Res . 2007 ;180 : 355 –365 . 17256160 3 Bareš M , Lungu OV , Husárová I , Gescheidt T . Predictive Motor Timing Performance Dissociates Between Early Diseases of the Cerebellum and Parkinson’s Disease . The Cerebellum . 2010 ;9 : 124 –135 . 10.1007/s12311-009-0133-5 19851820 4 Bareš M , Lungu OV , Liu T , Waechter T , Gomez CM , Ashe J . The neural substrate of predictive motor timing in spinocerebellar ataxia . The Cerebellum . 2011 ;10 : 233 –44 . 10.1007/s12311-010-0237-y 21110147 5 O’Reilly JX , Mesulam MM , Nobre AC . The Cerebellum Predicts the Timing of Perceptual Events . J Neurosci . 2008 ;28 : 2252 –2260 . 10.1523/JNEUROSCI.2742-07.2008 18305258 6 Ivry RB , Keele SW . Timing functions of the cerebellum . J Cogn Neurosci . 1989 ;1 : 136 –152 . 10.1162/jocn.1989.1.2.136 23968462 7 Ivry RB . The representation of temporal information in perception and motor control . Curr. Opin. Neurobiol . 1996 ;6 : 851 –857 . 9000026 8 Ivry RB , Spencer RMC , Zelaznik HN , Diedrichsen J . The cerebellum and event timing . Ann N Y Acad Sci . 2002 ;978 : 302 –317 . 12582062 9 Spencer RMC , Ivry RB . Cerebellum and Timing In: Manto M , Schmahmann JD , Rossi F , Gruol DL , Koibuchi N , editors. Handbook of the Cerebellum and Cerebellar Disorders . Dordrecht : Springer Netherlands ; 2013 pp. 1201 –1219 . 10 Coull JT , Nobre AC . Dissociating explicit timing from temporal expectation with fMRI . Curr Opin Neurobiol . 2008 ;18 : 137 –144 . 10.1016/j.conb.2008.07.011 18692573 11 Matsuda S , Matsumoto H , Furubayashi T , Hanajima R , Tsuji S , Ugawa Y , et al The 3-Second Rule in Hereditary Pure Cerebellar Ataxia: A Synchronized Tapping Study . PLOS ONE . 2015 ;10 :e0118592 10.1371/journal.pone.0118592 25706752 12 Ivry RB , Keele SW , Diener HC . Dissociation of the lateral and medial cerebellum in movement timing and movement execution . Exp Brain Res . 1988 ;73 : 167 –180 . 3208855 13 Grube M , Cooper FE , Chinnery PF , Griffiths TD . Dissociation of duration-based and beat-based auditory timing in cerebellar degeneration . Proc Natl Acad Sci . 2010 ;107 : 11597 –11601 . 10.1073/pnas.0910473107 20534501 14 Casini L , Ivry RB . Effects of divided attention on temporal processing in patients with lesions of the cerebellum or frontal lobe . Neuropsychology . 1999 ;13 : 10 –21 . 10067771 15 Bueti D , Walsh V , Frith C , Rees G . Different brain circuits underlie motor and perceptual representations of temporal intervals . J Cogn Neurosci . 2008 ;20 : 204 –214 . 10.1162/jocn.2008.20017 18275329 16 Jahanshahi M , Jones CRG , Dirnberger G , Frith CD . The Substantia Nigra Pars Compacta and Temporal Processing . J Neurosci . 2006 ;26 : 12266 –12273 . 17122052 17 Rao SM , Harrington DL , Haaland KY , Bobholz JA , Cox RW , Binder JR . Distributed neural systems underlying the timing of movements . J Neurosci . 1997 ;17 : 5528 –5535 . 9204934 18 Jänke L , Loose R , Lutz K , Specht K , Shah NJ . Cortical activations during paced finger-tapping applying visual and auditory pacing stimuli . Cogn Brain Res . 2000 ;10 : 51 –66 . 19 Pollok B , Gross J , Kamp D , Schnitzler A . Evidence for anticipatory motor control within a cerebello-diencephalic-parietal network . J Cogn Neurosci . 2008 ;20 : 828 –840 . 10.1162/jocn.2008.20506 18201129 20 Del Olmo MF , Cheeran B , Koch G , Rothwell JC . Role of the cerebellum in externally paced rhythmic finger movements . J Neurophysiol. Am Physiological Soc ; 2007 ;98 : 145 –152 . 21 Koch G , Oliveri M , Torriero S , Salerno S , Gerfo EL , Caltagirone C . Repetitive TMS of cerebellum interferes with millisecond time processing . Exp Brain Res . 2007 ;179 : 291 –299 . 17146647 22 Zhuchenko O , Bailey J , Bonnen P , Ashizawa T , Stockton DW , Amos C , et al Autosomal dominant cerebellar ataxia (SCA6) associated with small polyglutamine expansions in the alpha1A-voltage-dependent calcium channel . Nat Genet . 1997 ;15 : 62 –69 . 8988170 23 Mark MD , Krause M , Boele H-J , Kruse W , Pollok S , Kuner T , et al Spinocerebellar Ataxia Type 6 Protein Aggregates Cause Deficits in Motor Learning and Cerebellar Plasticity . J Neurosci . 2015 ;35 : 8882 –8895 . 10.1523/JNEUROSCI.0891-15.2015 26063920 24 Koeppen A . The pathogenesis of spinocerebellar ataxia . The Cerebellum . 2005 ;4 : 62 –73 . 15895563 25 Solodkin A , Gomez CM . Spinocerebellar ataxia type 6 . Ataxic Disord . 2011 ;103 : 461 –473 . 26 Gomez CM , Thompson RM , Gammack JT , Perlman SL , Dobyns WB , Truwit CL , et al Spinocerebellar ataxia type 6: gaze-evoked and vertical nystagmus, Purkinje cell degeneration, and variable age of onset . Ann Neurol . 1997 ;42 : 933 –950 . 9403487 27 Schmitz-Hübsch T , Du Montcel ST , Baliko L , Berciano J , Boesch S , Depondt C , et al Scale for the assessment and rating of ataxia Development of a new clinical scale . Neurology . 2006 ;66 : 1717 –1720 . 16769946 28 Gobel EW , Parrish TB , Reber PJ . Neural correlates of skill acquisition: decreased cortical activity during a serial interception sequence learning task . NeuroImage . 2011 ;58 : 1150 –1157 . 10.1016/j.neuroimage.2011.06.090 21771663 29 Brainard DH . The Psychophysics Toolbox . Spat Vis . 1997 ;10 : 433 –436 . 9176952 30 Pelli DG . The VideoToolbox software for visual psychophysics: transforming numbers into movies . Spat Vis. Brill ; 1997 ;10 : 437 –442 . 31 Brown MB , Forsythe AB . Robust Tests for the Equality of Variances . J Am Stat Assoc . 1974 ;69 : 364 –367 . 32 Petter EA , Lusk NA , Hull C , Hesslow G , Meck WH . Interactive roles of the cerebellum and striatum in sub-second and supra-second timing: Support for an initiation, continuation, adjustment, and termination (ICAT) model of temporal integration . Neurosci Biobehav Rev . In press. 33 Lang CE , Bastian AJ . Cerebellar subjects show impaired adaptation of anticipatory EMG during catching . J Neurophysiol . 1999 ;82 : 2108 –2119 . 10561391 34 Timmann D , Watts S , Hore J . Failure of cerebellar patients to time finger opening precisely causes ball high-low inaccuracy in overarm throws . J Neurophysiol . 1999 ;82 : 103 –114 . 10400939 35 Topka H , Valls-Solé J , Massaquoi SG , Hallett M . Deficit in classical conditioning in patients with cerebellar degeneration . Brain . 1993 ;116 : 961 –969 . 8353718 36 McCormick DA , Thompson RF . Neuronal responses of the rabbit cerebellum during acquisition and performance of a classically conditioned nictitating membrane-eyelid response . J Neurosci . 1984 ;4 : 2811 –2822 . 6502205 37 Perrett SP , Ruiz BP , Mauk MD . Cerebellar cortex lesions disrupt learning-dependent timing of conditioned eyelid responses . J Neurosci . 1993 ;13 : 1708 –1718 . 8463846 38 Breukelaar JWC , Dalrymple-Alford JC . Effects of lesions to the cerebellar vermis and hemispheres on timing and counting in rats . Behav Neurosci . 1999 ;113 : 78 –90 . 10197908 39 Nichelli P , Alway D , Grafman J . Perceptual timing in cerebellar degeneration . Neuropsychologia . 1996 ;34 : 863 –871 . 8822733 40 Mangels JA , Ivry RB , Shimizu N . Dissociable contributions of the prefrontal and neocerebellar cortex to time perception . Cogn Brain Res . 1998 ;7 : 15 –39 . 41 Gooch CM , Wiener M , Wencil EB , Coslett HB . Interval timing disruptions in subjects with cerebellar lesions . Neuropsychologia . 2010 ;48 : 1022 –1031 . 10.1016/j.neuropsychologia.2009.11.028 19962999 42 Meck WH , Penney TB , Pouthas V . Cortico-striatal representation of time in animals and humans . Curr Opin Neurobiol . 2008 ;18 : 145 –152 . 10.1016/j.conb.2008.08.002 18708142 43 Lee K-H , Egleston PN , Brown WH , Gregory AN , Barker AT , Woodruff PW . The role of the cerebellum in subsecond time perception: evidence from repetitive transcranial magnetic stimulation . J Cogn Neurosci . 2007 ;19 : 147 –157 . 17214571 44 Aso K , Hanakawa T , Aso T , Fukuyama H . Cerebro-cerebellar interactions underlying temporal information processing . J Cogn Neurosci . 2010 ;22 : 2913 –2925 . 10.1162/jocn.2010.21429 20044898 45 Fierro B , Palermo A , Puma A , Francolini M , Panetta ML , Daniele O , et al Role of the cerebellum in time perception: A TMS study in normal subjects . J Neurol Sci . 2007 ;263 : 107 –112 . 17655867 46 Harrington DL . Does the representation of time depend on the cerebellum?: Effect of cerebellar stroke . Brain . 2004 ;127 : 561 –574 . 14711883 47 Harrington DL . Reply to: Evaluating the role of the cerebellum in temporal processing: beware of the null hypothesis . Brain . 2004 ;127 :E14 . 48 Ivry RB . Evaluating the role of the cerebellum in temporal processing: beware of the null hypothesis . Brain . 2004 ;127 :E13 15277307 49 Sakai K , Ramnani N , Passingham RE . Learning of sequences of finger movements and timing: frontal lobe and action-oriented representation . J Neurophysiol . 2002 ;88 : 2035 –2046 . 12364526 50 Doyon J , Song AW , Karni A , Lalonde F , Adams MM , Ungerleider LG . Experience-dependent changes in cerebellar contributions to motor sequence learning . Proc Natl Acad Sci . 2002 ;99 : 1017 –1022 . 11805340 51 Grafton ST , Hazeltine E , Ivry R . Functional mapping of sequence learning in normal humans . J Cogn Neurosci . 1995 ;7 : 497 –510 . 10.1162/jocn.1995.7.4.497 23961907 52 Hazeltine E , Grafton ST , Ivry R . Attention and stimulus characteristics determine the locus of motor-sequence encoding. A PET study . Brain . 1997 ;120 : 123 –140 . 9055803 53 Honda M , Deiber MP , Pascual-Leone A , Zhuang P , Hallett M . Dynamic cortical involvement in implicit and explicit motor sequence learning . Brain . 1998 ;121 : 2159 –2173 . 9827775 54 Jenkins IH , Brooks DJ , Nixon PD , Frackowiak RS , Passingham RE . Motor sequence learning: a study with positron emission tomography . J Neurosci . 1994 ;14 : 3775 –3790 . 8207487 55 Jueptner M , Stephan KM , Frith CD , Brooks DJ , Frackowiak RSJ , Passingham RE . Anatomy of motor learning. I. Frontal cortex and attention to action . J Neurophysiol . 1997 ;77 : 1313 –1324 . 9084599 56 Sakai K , Hikosaka O , Miyauchi S , Takino R , Sasaki Y , Pütz B . Transition of brain activation from frontal to parietal areas in visuomotor sequence learning . J Neurosci . 1998 ;18 : 1827 –1840 . 9465007 57 Toni I , Krams M , Turner R , Passingham RE . The time course of changes during motor sequence learning: a whole-brain fMRI study . Neuroimage . 1998 ;8 : 50 –61 . 9698575 58 Seidler RD , Purushotham A , Kim S-G , Uğurbil K , Willingham D , Ashe J . Cerebellum activation associated with performance change but not motor learning . Science . 2002 ;296 : 2043 –2046 . 12065841 59 Schendan HE , Searl MM , Melrose RJ , Stern CE . An FMRI study of the role of the medial temporal lobe in implicit and explicit sequence learning . Neuron . 2003 ;37 : 1013 –1025 . 12670429 60 Ross RS , Brown TI , Stern CE . The retrieval of learned sequences engages the hippocampus: Evidence from fMRI . Hippocampus . 2009 ;19 : 790 –799 . 10.1002/hipo.20558 19219919 61 Allen TA , Salz DM , McKenzie S , Fortin NJ . Nonspatial Sequence Coding in CA1 Neurons . J Neurosci . 2016 ;36 : 1547 –1563 . 10.1523/JNEUROSCI.2874-15.2016 26843637 62 Schubotz RI , van Cramon Y . Interval and ordinal properties of sequences are associated with distinct premotor areas . Cereb Cortex . 2001 ;11 : 210 –222 . 11230093 63 Ramnani N , Passingham RE . Changes in the human brain during rhythm learning . J Cogn Neurosci . 2001 ;13 : 952 –966 . 11595098 64 Shin J , Ivry RB . Spatial and temporal sequence learning in patients with Parkinson’s disease or cerebellar lesions . J. Cogn. Neurosci. 2003 ;15 : 1232 –1243 . 14709239 65 Akshoomoff NA , Courchesne E . ERP evidence for a shifting attention deficit in patients with damage to the cerebellum . J. Cogn. Neurosci. 1994 ;6 : 388 –399 . 10.1162/jocn.1994.6.4.388 23961733 66 Courchesne E , Townsend J , Akshoomoff NA , Saitoh S , Yeung-Couchesne R , Lincoln AJ , et al Impairment in shifting attention in autistic and cerebellar patients . Behav Neurosci . 1994 ;108 : 848 –865 . 7826509 67 Yamaguchi S , Tsuchiya H , Kobayashi S . Visuospatial attention shift and motor responses in cerebellar disorders . J Cogn Neurosci . 1998 ;10 : 95 –107 . 9526085 68 Desmond JE , Gabrieli JD , Wagner AD , Ginier BL , Glover GH . Lobular patterns of cerebellar activation in verbal working-memory and finger-tapping tasks as revealed by functional MRI . J Neurosci . 1997 ;17 : 9675 –9685 . 9391022 69 Wolpert DM , Miall RC , Kawato M . Internal models in the cerebellum . Trends Cogn Sci . 1998 ;2 : 338 –347 . 21227230 70 Miall RC , Wolpert DM . Forward models for physiological motor control . Neural Netw . 1996 ;9 : 1265 –1279 . 12662535

PMC005xxxxxx/PMC5003365.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 10.1371/journal.pone.0160429PONE-D-16-14976Research ArticleResearch and Analysis MethodsResearch FacilitiesResearch LaboratoriesGovernment LaboratoriesMedicine and Health SciencesInfectious DiseasesMedicine and Health SciencesEpidemiologyDisease SurveillanceInfectious Disease SurveillanceMedicine and Health SciencesInfectious DiseasesInfectious Disease ControlInfectious Disease SurveillanceMedicine and Health SciencesInfectious DiseasesSexually Transmitted DiseasesBiology and Life SciencesMicrobiologyPeople and PlacesGeographical LocationsEuropeBelgiumMedicine and Health SciencesPulmonologyRespiratory InfectionsBiology and Life SciencesImmunologyVaccination and ImmunizationVaccinesMedicine and Health SciencesImmunologyVaccination and ImmunizationVaccinesMedicine and Health SciencesPublic and Occupational HealthPreventive MedicineVaccination and ImmunizationVaccinesSurveillance of Infectious Diseases by the Sentinel Laboratory Network in Belgium: 30 Years of Continuous Improvement 30 Years of Surveillance of Infectious Diseases by the Sentinel Laboratory NetworkMuyldermans Gaëtan *Ducoffre Geneviève Leroy Mathias Dupont Yves Quolin Sophie participating sentinel laboratories ¶WIV-ISP, OD Public Health and Surveillance, Unit ‘Epidemiology of infectious diseases’, Brussels, BelgiumChuang Jen-Hsiang EditorCenters for Disease Control, TAIWANCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: GM GD. Data curation: ML YD. Formal analysis: GM GD. Funding acquisition: SQ. Investigation: GM GD. Methodology: GM GD. Project administration: GM GD. Resources: SQ. Software: ML. Supervision: GM GD. Validation: GM GD. Visualization: GM GD. Writing – original draft: GM GD. Writing – review & editing: GM GD ML SQ. ¶ Membership of the participating sentinel laboratories is provided in the Acknowledgments. * E-mail: gaetan.muyldermans@wiv-isp.be29 8 2016 2016 11 8 e016042914 4 2016 19 7 2016 © 2016 Muyldermans et al2016Muyldermans et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.In 1983 the sentinel laboratory network was established because of the need to describe the epidemiological evolution of infectious diseases. During the study period of 30 years (1983–2013), microbiology laboratories reported on weekly basis the laboratory diagnosed cases for a selection of infectious diseases. This resulted in a large longitudinal laboratory based database allowing to provide trends over time and distribution by person and place. During this period, adaptations to data collection were made due to changes in diagnostic methods and public health priorities, introduction and application of digital revolution, and multiple reorganizations of the laboratories. Since the surveillance network is dynamic, it necessitates a continuous evaluation to ensure that, over time, it continues to be representative of the general epidemiological trends in the country. Secondly the aim is to examine the robustness and stability of this surveillance system. Here we demonstrated that the flexibility of the data collection methodology by the sentinel laboratory network is unique and that adaptations do not affect the capacity of the system to follow trends. Therefore, the surveillance by this network is representative of the current epidemiological situation in Belgium. To our knowledge, no such surveillance network with such a long-term follow-up and demonstrated stability for multiple infectious diseases in the general population was earlier described. Furthermore, expected trends due to the implementation of vaccination or other events were accurately detected. The collected data obtained from this network allows interesting comparisons with other national and international information sources. The sentinel laboratory network was funded by the Flemish region and the Walloon-Brussels federation. Data AvailabilityAs the data contains patient’s demographic information, we are not allowed to publicly provide these data for ethical reasons. However, we accept to share data in case interested readers or researchers should request underlying data. In this case, data can be requested to the corresponding author. Aggregated data are available and graphically represented by the Epistat platform (https://epistat.wiv-isp.be/).Data Availability As the data contains patient’s demographic information, we are not allowed to publicly provide these data for ethical reasons. However, we accept to share data in case interested readers or researchers should request underlying data. In this case, data can be requested to the corresponding author. Aggregated data are available and graphically represented by the Epistat platform (https://epistat.wiv-isp.be/). ==== Body Introduction In Belgium, the sentinel laboratory network was established in 1983 in order to obtain information on the epidemiology of infectious diseases [1]. The laboratories participating to this network are further called the sentinel laboratories. This sentinel laboratory network is coordinated by the Scientific Institute of Public Health (WIV-ISP) organizing the data collection and data storage, and facilitating the data processing and dissemination of this information. The main objective of this network is to monitor the evolution of different infectious diseases over time, both within a calendar year and over several years. The collected information allows as well to fulfil national and international (i.e. ECDC, WHO) requests. Next to this sentinel laboratory network, other surveillance networks for human infectious diseases, complementing each other, are available in Belgium, i.e. the notification of infectious diseases organised by the Flemish Community [2], Brussels Capital and the French-speaking Community [3], the network of paediatrics collecting mainly data on vaccine preventable infectious diseases in children since 2002 [4], the network for surveillance of sexually transmitted diseases since 2000 [5], the network of national reference laboratories and the national reference centers collecting public health microbiology data since 2011 (see also materials and methods) [6] and the registration network of general practitioners since 1979 [7]. Each surveillance system has its strengths and weaknesses. Although the mandatory notification system and the sentinel laboratory network are both fed by microbiology laboratories, additional clinical information is provided to the mandatory notification system by the treating physicians which renders this system exhaustive. On the other hand, the sentinel laboratory network is based on a fraction of the microbiology laboratories raising concerns about the representativeness nationwide and regional [8]. By monitoring 12 pathogens, it was previously demonstrated [9,10] that the coverage of the sentinel network was stable over time and close to, or greater than 50%. Test coverage was in this study calculated by the ratio of reimbursed tests performed by participating laboratories to the total number of tests performed. These results indicate that the network is sensitive and representative for the surveillance of the selected pathogens. Furthermore, these results hold for the 3 regions of Belgium but at the provincial level, a lower test coverage was shown for some pathogens [10]. Moreover, the information provided by the sentinel network is usually considered to be timelier and more complete due to a better compliance of voluntary reporting laboratories [8]. This paper describes the historic changes and developments of the Belgian sentinel laboratory network such as the flexibility of adaptation of the network towards pathogen changes, diagnostic method changes, data transfer methodology and multiple reorganization of the participating sentinel laboratories. The lack of impact of these changes on the robustness of the longitudinal surveillance of this network was investigated by a detailed description of the trend changes for a series of infectious diseases. Materials and Methods Organisation of the network The sentinel laboratory network was implemented from 1983 onwards as described previously by Walckiers et al. (1991) [1]. Briefly, microbiology laboratories transferred on a weekly basis their laboratory diagnosed cases to the WIV-ISP on paper form by regular mail. The information for a limited number of variables was collected for a selection of infectious diseases. The encoded variables included the diagnosed infectious disease, some patients demographic data allowing the identification of duplicates i.e. date of birth (or previously age), gender and postal code. In addition the specimen and its sample identification number, the diagnostic method and the date of diagnosis are recorded as well. If applicable, the registration of the country of infection is foreseen. For confidence reasons, all data transfers are kept anonymous towards the patient. All data were collected in a central database and analysed for a quarterly and annual report. The participation by the sentinel laboratories was and is still today voluntary and without remuneration. If requested by a dedicated reference laboratory, the sentinel laboratories are encouraged to send strains to the reference laboratories/centers for further characterisation such as geno- and/or phenotyping or antimicrobial follow-up. The project on sentinel laboratory network is scientifically supervised by a steering committee composed by representatives of the sentinel laboratories, reference laboratories/centers and authorities from the Flemish Community, Brussels Capital and the French-speaking Community having infection control and prevention into their competencies. The list of infectious diseases is yearly reviewed by this steering committee selecting pathogens based on the current need of public health. In 2013, the trend analysis included 35 pathogens and covered respiratory infections, gastrointestinal infections, sexually transmitted infections, imported infections as well as zoonosis and vaccine preventable diseases [11]. The list of pathogens is chosen such that it does not overburden the administrative work for the sentinel laboratories. Study period The impact of historical adaptations during the study period of 30 years, starting from the implementation of the project (1983) until 2013, were recorded and described. Historical adaptations to improve the functioning of the network During the study period, the network underwent several adaptations according to the evolution at the level of microbial diagnosis and due to the digital revolution. A summary of all adaptations is briefly described underneath. Since the implementation of the network, the diagnostic methods were yearly revised. As of 1983 all data covered culture positive cases for the 26 infectious diseases. Meanwhile cases diagnosed by serological methods were included gradually since 1987, those by molecular diagnosis (PCR) gradually since 2004 and currently the molecular diagnosis is included for 31 out of the 36 infectious diseases. The weekly transfer of data from the laboratories to the WIV-ISP was initially on paper format. Since the beginning of 2000, some of the laboratories were able to extract the needed information directly from their databases. These cases were recorded in batch to the database. During the same period, other laboratories reported their data via a web application, developed and made available by the WIV-ISP. An online submission tutorial was provided to the participating laboratories extracting their data from their laboratory information management system (LIMS) or those reporting by the webtool. As further described in the results section, the fractions of data obtained by the different data transfer methods by the participating sentinel laboratories changed during the last few years. All historical datasets obtained by the different data transfer methods were collected on a SQL server at the public health institute WIV-ISP for further data cleaning. Since the microbiology laboratory test results populating the database throughout all these years are neither uniformly coded nor documented in a standardized manner, some quality assurance measurements were taken. The application of a systematic approach was introduced since 2009 by the definition of harmonized variables and their formats including well defined own coded values [11]. This was felt necessary for an ongoing oversight and a management of the database to remain valid and useful. All procedures were described in standard operating procedures (SOP) and are available in the internal document management system (not shared in public but available on request). A number of quality checks were gradually implemented to improve the completeness and correctness of the reported cases in the database. Firstly, the completeness of the recorded cases reported by paper format was double checked by a second person. Secondly, a monthly feedback was sent since 2012 to all sentinel laboratories to provide an overview of all transferred data since the beginning of the current year. This feedback allows the participating laboratories to determine the completeness of their data in the database and moreover to compare the rate of diagnosed infectious diseases of their laboratories with those from all participating laboratories. This feedback system allows to receive the missing data from the participating laboratories and improving thereby the completeness of the data. Lastly, quality checks toward variable completeness, consistency, content, alignment with specifications (case definitions), expected trends (seasonal and yearly) are performed on weekly basis for a few cases, selected by cherry picking from all received data. Recently case definitions for each of the infectious diseases were defined [11] based on available international information [12,13] to harmonise the inclusion of cases and to ensure their fulfilment of the case criteria. The availability of reference laboratories collaborating with this network on a voluntary basis was since 2011 replaced partially by a network of National Reference Centers (NRC’s) [6]. These NRC’s were selected based on defined criteria, are reimbursed for their activities and need to fulfil to the proper quality assurance level (ISO15189). The tasks of these laboratories and NRC’s are to diagnose or to confirm rare diseases or diseases difficult to diagnose, to perform some further typing of strains, to determine the resistance towards antimicrobials or their resistance mechanisms and finally provide these epidemiological and microbial data for reporting at national and international levels. Data cleaning The definition of the variables and their formats allowed assembling all historical data into one database on which longitudinal trend analysis can be performed. Since 2012, a data cleaning program was implemented with SAS software (SAS Institute Inc.®, Cary, NC, USA). The program mainly executes the following steps: append all historical tables, harmonize variable names and their format, compute new variables (i.e. age, agegroup), and remove duplicates on basis of sample identification or patient demographic data i.e. date of birth (or previously age), gender and postal code. The encoded variables included the diagnosed infectious disease, some patient’s demographic data and specimen and analysis method information. This clean table is optimised by statistical programs and is automatically updated twice a day. Data analysis A web platform called “Epistat” (https://epistat.wiv-isp.be/) was built since 2012 for secured feedback and real time analysis of the clean database. This platform allows the construction of different types of graphs for epidemiological monitoring and for further investigations according to the needs of the sentinel laboratories and other stakeholders: time distributions, geographical maps, age pyramids, pie charts and histograms. Created graphs may be used in surveillance reports and scientific publications by both the WIV-ISP and the sentinel laboratories. The tool is composed of three different parts: a webform, written in HTML5 and Javascript; a SAS program (SAS Institute Inc. ®, Cary, NC, USA) generating the different options appearing in the form; and a SAS stored process receiving the selected options in the form and creating the final output in the browser window. Since the software runs on the web, the application is accessible only with an internet connection and a browser. Access to this application is restricted to the participating laboratories, the WIV-ISP staff members and the officers from the health authorities with logins and passwords. Results Laboratory participation rate During the study period, the number of participating laboratories has decreased over time (Fig 1A). After a first introduction period of the network, 159 laboratories participated to the surveillance in 1985 and currently we observed 97 sentinel laboratories in 2013. However due to structural reorganizations and fusions of some clinical laboratories, the ratio of participating sentinel laboratories from all microbiology laboratories has increased from 40% in 1985 to 58% in 2004 and remained constant since then (Fig 1B). In 2013, the fraction of participating laboratories on the total number of registered microbiology laboratories in Belgium was 59% (97 participating laboratories out of 163 microbiology laboratories). We could not determine whether the academic and non-academic hospital related laboratories, and non-hospital related laboratories participating to the network are proportionally representative from the laboratories performing microbiological diagnoses in Belgium. 10.1371/journal.pone.0160429.g001Fig 1 Number of sentinel laboratories participating to the sentinel laboratory network (A) and percentage of microbiology laboratories participating to the network as compared to the total number of registered microbiology laboratories (B). However, the coverage, a measure of the proportion of target population included in the surveillance system was previously studied [9, 10] demonstrating that the participating laboratories were performing more than 50% of all tests, and that the coverage was constant in Belgium between 1999 and 2002 [9] and between 2007 and 2012 [10]. In 2013, the distribution of sentinel laboratories by region was 54% in the Flemish Community, 34% in the French-speaking Community and 12% in Brussels Capital. This distribution is comparable to that of all the registered microbiology laboratories in Belgium (data not shown) demonstrating its regional representativity. Also the distribution of the Belgian population (n = 11.099.554 in 2013) is similar: 57% in the Flemish Community, 32% in the French-speaking Community and 10% in Brussels Capital. Evolution in Number of Pathogens and Data Transfers The number of infectious diseases for which data were collected, increased from 26 in 1983 to the currently 35. A gradual increase of the number of reported cases was observed during the study period of 30 years (data not shown). Especially during the last decade when the digital data transfer became available, a tremendous increase of the number of reported cases was observed (Fig 2). The availability of a digital data transfer system reduced the workload substantially and improved the speed of data transmission. However, this necessitates the availability of an automatically process to clean the data especially removing the duplicates. As demonstrated further, the implementation of digital data transfer improved the completeness of the data without impact on the number of cases after removal of duplicates. 10.1371/journal.pone.0160429.g002Fig 2 Number of reported cases before removal of the duplicates from 2007 when the digital data transfer became available until 2013. The total number of reported cases (Total) is transferred by sending the information on paper format by regular mail (Paper), by importing the cases via a web application developed by the WIV-ISP (WebForm), or by sending an extraction of the cases from the laboratory information management system (Export). In 2013 only 12 laboratories (12%) used the paper form to report their data. Long standing follow-up of infectious diseases For 10 infectious diseases (Campylobacter, Chlamydia trachomatis, Entamoeba histolitica, Haemophilus influenzae, Legionella pneumophila, Neisseria gonorrhoeae, N. meningitidis, Plasmodium, Streptococcus pneumoniae, and Yersinia enterocolitica), the surveillance covered the entire study period of 30 years. Fig 3 demonstrates the yearly trend analysis for some of these including the seasonal activities. Although for Yersinia enterocolitica a continuous decrease was observed since 1983, the Campylobacter surveillance demonstrated a continuous number of reported cases. Within this waving incidence over the years (slight increase till 2000 followed by a decrease and relapse again from 2010 onwards), we observed a seasonal decrease in summer 1999. During the same period, the contamination of feedstock with polychlorinated biphenyls was demonstrated leading to the destruction of massive amounts of animal food products, mainly eggs and chicken [14]. This so called dioxin affair led to the reduced consumption of these animal food products and thereby a significant decline (40%) in the number of infections [15]. 10.1371/journal.pone.0160429.g003Fig 3 Trend analyses for some representative infectious diseases as measured by the sentinel laboratory network from 1993 (if available) until 2013. The dashed line represents the monthly number of cases while the orange line represents a smoothed curve obtained by the Loess statistical method (SAS Institute Inc. ®, Cary, NC, USA), a weighted scatterplot through these data points. Campylobacter, Y. enterocolitica, Chlamydia trachomatis, Neisseria gonorrhoeae, Haemophilus influenzae, Neisseria meningitidis, Borrelia burgdorferi, Rotavirus, and RSV. For the follow-up of the sexual transmitted infections (STI), a continuous increase is observed as demonstrated in Fig 3 for C. trachomatis and N. gonorrhoeae without notification of a similar increase of reimbursed tests [10]. For C. trachomatis (in 2001: N = 775, in 2013: N = 5232) the number of cases in the age range of young women from 15 to 29 years explains largely the increase, while for N. gonorrheae (in 2001: N = 230, in 2013: N = 1063) the increase is situated in particular in men from 20 to 34 years and for T. pallidum (in 2002: N = 114, in 2013: N = 1293), the increase is situated in particular in men from 35 to 49 years old. These increases are mainly observed in the Flemish Community (district of Antwerp) and in Brussels Capital (data not shown). These evolutions are also observed in other European countries [16]. It is also necessary to remain vigilant watchful as for the evolution of the used techniques of diagnosis to explain our observations. For example, the development since 2002 of more sensitive techniques of diagnosis could partially explain the increased diagnosis of sexual transmitted infections. However, the implementation of a clinical STI network confirmed the increase of STI [17]. For Haemophilus influenzae and Neisseria meningitidis, a sudden decrease was observed from 2002 onwards. This could be brought in relation with the introduction and/or the reimbursement policy of the vaccines against these infectious diseases. For Neisseria meningitidis, a drop from 280 cases in 2001 to 108 cases in 2013 was observed. The reference laboratory/center for Neisseria meningitidis receiving the strains for further subtyping, demonstrated the decrease in incidence of serogroup C from 49.4% in 2001 to 10.4% in 2013 [18] showing that the N. meningitidis serogroup C was brought under control. For 11 other pathogens (Borrelia burgdorferi, Bordetella, Chlamydia psittaci, Cryptosporidium, Giardia, Hantavirus, Hepatitis A virus, Influenza, Mycoplasma pneumoniae, Streptococcus pyogenes, and VTEC), a surveillance is available spanning a period of more than 20 years. The trend analysis and other surveillance systems are described in details in the annual reports and are available on the website of the network (https://epidemio.wiv-isp.be). The serological confirmation of Borrelia burgdorferi, to diagnose Lyme borreliose, is also included in the surveillance program from the beginning of the years '90. Between 1998 and 2012 the incidence raised reaching yearly between 1000–1500 cases. Within the period 2007–2012, respectively 120.000 to 280.000 tests were performed per year [10]. Since then, due to multiple discussion forums concerning the so called ‘chronic Lyme disease’ or post-treatment Lyme disease syndrome and an increased interest by the media of the Lyme disease in Belgium, the reported cases increased to 2090 in 2013. Also an increased test frequency was demonstrated during this period [10]. For Rotavirus, cases were recorded from 1999 onwards but due to the workload to transfer the huge number of records, it was interrupted a few years later (2002–2004). It was again initiated in 2005 to obtain a background measurement of the number of diagnosed cases before the introduction of 2 vaccines (Rotarix® [GlaxoSmithKline Biologicals Rixensart, Belgium] in June 2006 and RotaTeq® [Merck&Co., Inc.Whitehouse Station, New Jersey] in June 2007) [19]. The number of diagnosed Rotavirus cases decreased from 9414 cases in 2005 to 2359 cases in 2013, demonstrating the possibility of the network to follow-up the effect of the vaccine policy. For RSV a continuous seasonal increase of cases was reported, starting every year from week 40 i.e. at the beginning of October. The maximum number of cases per week is generally situated at mid-December. The majority of the cases are diagnosed on young children of less than 5 years. In 2013, 8294 cases were diagnosed by the network. Representativity of the data Although the geographical representativeness was previously investigated [9,10] we further investigated whether all sentinel laboratories participated equally well for each pathogen. Table 1 demonstrates the number of participating laboratories per pathogen in decreasing order. Campylobacter was reported at median level over the 5 last years by 91 laboratories. Further, Rotavirus and RSV, 2 pathogens introduced some years after the implementation of the network were reported by respectively 82 and 74 participants. 10.1371/journal.pone.0160429.t001Table 1 Overview of the number of sentinel laboratories reporting a particular infectious disease. The indicated number of reporting laboratories is calculated from the median number of sentinel laboratories reporting cases during the last 5 years of the study period. Pathogen Reporting laboratories (last 5 y) Median Range Campylobacter 91 85–94 Rotavirus 82 79–84 RSV 74 70–78 Giardia 73 69–75 S. pneumoniae 74 67–77 Y. enterocolitica 66 58–70 N. gonorrhoeae 64 61–67 S. pyogenes 53 51–59 M. pneumoniae 53 48–54 C. trachomatis 53 51–55 Considering the respiratory infectious diseases RSV, M. pneumoniae, Adenovirus, and Parainfluenza virus, the number of reporting laboratories varied respectively with 74, 53, 41 and 17 (data not shown). This implicates that the cross pathogen comparison for symptomatic diseases is hampered by the reporting participation of sentinel laboratories. Discussion There is an increasing need, both at national and international levels, to obtain epidemiological information on many infectious diseases. We demonstrated here that the sentinel laboratory network is an important tool to provide the necessary data and thereby to accomplish these tasks. It provides a robust surveillance for multiple infectious diseases in Belgium and this despite some adaptations throughout these 30 years of the surveillance period. The most important adaptations are: changes in diagnostic methodology by the advent of the PCR technology and the implementation of this sensitive technique to replace mainly the culture, structural reorganisations and fusions between clinical laboratories and the digital revolution, all described in materials and methods. It has been implemented in Belgium since 1983 and was since then one of the most important surveillance network in infectious diseases with a good coverage demonstrating a good sensitivity and geographical representativeness [9, 10]. The collection of a limited number of variables encourages approximately 58% of the available laboratories to voluntarily participate to this network. Whether this high participation rate of laboratories reflects a similar coverage of the Belgian population was not assessed in this study but was previously investigated [10]. During these 30 years of study, the surveillance system demonstrated also a great flexibility by having the capacity to monitor pathogens for the entire period (n = 10) while others (n = 11) were added or removed depending of their public health needs. The longitudinal surveillance of these pathogens was exemplified in the results section for some gastrointestinal infectious diseases (Y. enterocolitica, Campylobacter and Rotavirus), sexually transmitted diseases (Chlamydia trachomatis and N. gonorrhoeae), vaccine preventable diseases (N. meningitidis, H. influenzae and Rotavirus) and the respiratory diseases (RSV). For all of these pathogens we could demonstrate a surveillance which is in line with current international findings [20]. For the vaccine preventable diseases, we could determine the impact of the introduction and reimbursement of the vaccine policy (H. influenzae, N. meningitidis and Rotavirus). For Rotavirus, it was previously demonstrated using the sentinel laboratory records that the infectious season was delayed compared to pre-vaccination seasons [19]. We could demonstrate the effect of some environmental factors such as the dioxin affair when during the summer period of 1999 the contamination of the feedstock resulted in the destruction of massive amounts of animal food products [14]. Also during the same period less animal food products such as eggs and poultry were consumed resulting in a decreased incidence of Campylobacter infections [15]. On the other hand the yearly increase in intensity of the seasonal peaks for RSV suggests for an increased incidence during the winter period. No indications of shift of the seasonal peaks neither a broadening of the curves were determined. The huge increase in incidence of sexually transmitted infections (STI) during the last decade is probably a reflection of the waning of prevention campaign. Why the incidence of the different STI is different according to the gender and age group could not be unravelled by the sentinel network due to the lack of clinical information. This observation is nevertheless confirmed by data from other national networks [17]. We demonstrated as well that this network has the characteristic to be dynamic. This feature is supported by the fact that it is possible to add a germ to the content of the recording at any time in the course of the year. Therefore it requires a regular follow-up which is assured by the scientific advice given by the yearly steering committee of the network bringing together the sentinel laboratories, the reference centers, the epidemiologists and the sponsors of the network. The addition of a pathogen as proposed by the steering committee can very easily be added to the list and after informing the participating laboratories, data can be collected without further investments in partners, infrastructure, or methodology of working. It is also necessary to remain vigilant in continuing the surveillances even if a low incidence is monitored. For instance the low burden of sexually transmitted infections during the 90’s [21] was tempting to remove them from the list. Those years the decrease in incidence was a profit from the massive prevention campaigns for HIV/AIDS. The continuous monitoring of three main sexually transmitted infectious diseases demonstrates that monitoring even at low burden remains important for the long term surveillance. The roles of the reference centers are to confirm the diagnosis of the received samples and to supply these with other invaluable microbial information, such as the type of circulating strains at the human and/or food level or their sensibility towards antimicrobials. By collecting and providing these microbial informations, the surveillance of particular pathogens is further accomplished. As an example, the reference centers have shown the effect of the vaccination on the decrease of N. meningitidis, in particular the serogroup C present in the vaccine [18]. A high participation rate (more than 50%) was measured for 10 pathogens. However we found different levels of participation for different infectious diseases belonging to the same symptomatic disease. This is true for the respiratory infectious diseases, but is also true for the gastrointestinal diseases and the sexually transmitted diseases (data not shown). This difference in participation rate between pathogens hampers cross pathogen comparison for infectious diseases belonging to the same symptomatic disease. By the implementation in the near future of a new electronic data transfer system based on harmonized national coding standards and a common data transfer route for all laboratories [22], the partial registration by some laboratories will be overcome. The advantage of the limited number of variables asked to the participating laboratories makes it easier to stimulate them to participate. It allowed the monitoring of the epidemiology in time and place. The drawback of this limited information is that no clinical information is requested and thus no clinical surveillance can be provided. In conclusion, the data supplied by this network represent a unique source of information from the point of view of the public health. It allowed us to accurately detect and describe trends, contribute to the estimation of the burden of disease and unravel the effect of the implementation of a vaccine policy or other events. These observations are currently presented in pathogen specific reports and allow furthermore interesting comparisons with other national and international information sources. We would like to thank all people in charge of the sentinel laboratories for their invaluable collaboration without which this network would not exist. We are grateful for their remarkable dedication and their regular transfer of data. The reference centers are grateful for their voluntary investigations in typing, antimicrobial resistance testing and exchanging their clinical expertise with colleagues. The Flemish Community, Brussels Capital and the French-speaking Community for their financial support. Our team members for initiating the project (André Stroobant), their help in the development of the network (Denise Walckiers, Viviane Van Casteren, Frank Van Loock, Germaine Hanquet, Sophie Quoilin and Geneviève Ducoffre), the data management and development of the electronic tools for data transfer (Roger Cornelis, Yves Dupont, Guy Jeanfils and Mathias Leroy), the investigators of the representativity studies (Hans Vandenberghe and Nicolas Berger) and all the colleagues to import the data (Dominique Meunier, Marleen Meganck, Ellen De Blanc, Luce De Gendt and Nathalie Verhocht). Members of the participating sentinel laboratories Vael C., A.Z. klina laboratorium, Brasschaat; Goossens H., Universitair ziekenhuis Antwerpen, Edegem; Ceyssens C., Algemeen ziekenhuis St.-Jozef, Malle; Stalpaert M., A.M.L., Antwerpen; Van Esbroeck M., Instituut voor Tropische Geneeskunde, Antwerpen; Bruynseels P., ZNA klinisch laboratorium, Antwerpen; Vermaelen K., A.Z. St.-Maarten, Duffel; Laffut W., H. Hartziekenhuis, Lier; Crabbe G., Somedi, Heist-Op-Den-Berg; Frans J, Imeldaziekenhuis, Bonheiden; Verbeeck P., Heilig Hartziekenhuis, Mol; De Muelenaere G., Algemeen ziekenhuis St.-Dimpna, Geel; Spiritus T., A.Z. St.-Elisabethziekenhuis, Herentals; Van Kerkhoven D., St.-Jozefkliniek, Turnhout; Reybrouck R., Regionaal Ziekenhuis Heilig Hart, Tienen; Van Hentenrijk C., Regionaal Ziekenhuis Heilig Hart, Leuven; De Vuyst D., Vereniging Diestse Ziekenhuizen, Diest; Van Meensel B., Medisch Centrum Huisartsen, Leuven; Patteet S., UZLeuven, Leuven; Verbelen V., Clinique St-Pierre, Ottignies-Louvain-la-Neuve; Mascart G., C.H.U. Brugmann, Bruxelles; Vandenberg O., Laboratoire de la porte de Hal, Bruxelles; Mulongo B., Cliniques Roi Baudouin, Bruxelles; Kabamba B., Cliniques Universitaires Saint-Luc, Bruxelles; Hing.M., Hôpital Militaire Reine Astrid, Neder-Over-Heembeek; Claude B., ULB Institut de biologie clinique, Bruxelles; Allemeersch D., Clinique de l'Europe, Bruxelles; Bertrand S., WIV-ISP, Bruxelles; Van Gucht S., WIV-ISP, Bruxelles; Denis O., Hôpital Erasme, Bruxelles; de Moreau de Gerbehaye A.I., Hôpitaux Iris Sud, Bruxelles; Glorieux T., Clinilabo, Brussel; Pierard D., Universitair Ziekenhuis Brussel, Brussel; De Craemer, St.-Rembert Kliniek, Torhout; Reynders M., A.Z. St.-Jan, Brugge; Vandewal W., A.Z. St.-Lucas, Brugge; Vandecandelaere P., Regionaal Ziekenhuis—Jan Yperman, Ieper; Maenhout P., Klinisch Labo Maenhout, Waregem; Boudewijns M., AZ Groeninge, Kortrijk; Segers H., O.L.V. Van Lourdes Ziekenhuis, Waregem; Bruyland, Medisch Labo Bruyland, Kortrijk; Alliet G., A.Z. Damiaan, Oostende; Declercq Ph., St.-Jozefkliniek, Izegem; De Laere E., AZ Delta, Roeselare; Desmedt R., Stedelijk Ziekenhuis, Roeselare; Van Hoecke F., St.-Andriesziekenhuis, Tielt; Van Liedekerke A., Algemeen Ziekenhuis St.-Elisabeth, Zottegem; Van Vaerenbergh K., O.L.V. Ziekenhuis, Aalst; Watelle M., Medisch Labo Medina, Dendermonde; Beckers J., Algemeen Ziekenhuis St.-Blasius, Dendermonde; Piette A., A.Z. Alma, Eeklo; Vanrompay D., Universiteit Gent, Gent; Verdonck A., Laboratorium Nuytinck, Brugge; Ide L., A.Z. Jan Palfijn Gent, Gent; Dierick J., A.Z. Maria Middelares, Gent; Vandenabeele A.M., A.Z. St.-Lucas, Gent; Claeys G., Universitair Ziekenhuis Gent, Gent; Coppens A., C.R.I., Zwijnaarde; Verbruggen A.M., A.Z. Nikolaas, St.-Niklaas; Govaerts D., C.H.U. André Vésale, Montigny-Le-Tilleul; Schatt P., Clinique Notre-Dame de Grâce, Gosselies; Lissoir B., Grand Hopital de Charleroi, Gilly; Fameree D., C.H.U. de Charleroi, Charleroi; Moonens F., RHMS Louis Caty Laboratoire, Baudour; Mansoor I., RHMS Louis Caty Laboratoire, Baudour; Vatlet M., C.H.U. Ambroise Paré, Mons; Van Bosterhaut B., Centre Hospitalier de Mouscron, Mouscron; Woestyn J., Laboratoire J. Woestyn, Mouscron; Potvliege C., Centre Hospitalier de Tivoli, La Louviere; Godet S., Labassos, Braine-Le-Comte; Loosen I., laboratoire de biologie médicale—Bauduin, Enghien; Mzougui T., Centre Hospitalier de Jolimont, Lobbes; Parmentier M-Fr., Centre de Santé des Fagnes, Chimay; Marchal J.F., Chwapi Tournai; Sion C., Centre Hospitalier Régional de Huy, Huy; Renard, Laboratoire d'analyses médicales—Ralet, Fléron; Boeras A., Clinique St-Joseph, Liège; Melin P., Hôpital universitaire du Sart Tilman, Liège; Carpentier M., Centre Hospitalier Régional de la Citadelle, Liège; Collard D., CHRV Centre Hospitalier de Verviers et de l’Est de la Belgique, Verviers; Schleck, Clinique Reine Astrid—labo de biologie clinique, Malmedy; Gobbels P., Hôpital Saint-Nicolas, Eupen; Rousseau D., Centre de diagnostic, Verviers; Van Goethem G., A.Z. St-Franciskus, Heusden; Waumans L., Jessaziekenhuis, Hasselt; Gabriels P., St.-Trudo, St.-Truiden; Oris E., Ziekenhuis Oost-Limburg, Genk; Bafort K., Maria ziekenhuis Noord-Limburg, Overpelt; Goffinet P., les Cliniques du Sud Luxembourg, Arlon; Andre M., Centre Hospitalier de l'Ardenne, Libramont; Glupczynski Y., Cliniques Universitaires (UCL)—de Mont-Godinne, Godinne; Jacquemin J.P., Centre Hospitalier de Dinant, Dinant; Garrino M.G., Centre Hospitalier Régional de Namur, Namur; Tamigniau A., C.H.R. du Val de Sambre, Auvelais; Dubois E., Labo Médic, Belgrade; Van Heule D., Sint-Jozef-Kliniek, Willebroek; Van Rompay, UGent, Gent. ==== Refs References 1 Walckiers D , Stroobant A , Yourassowsky E , Lion J , Cornelis R . A sentinel network of microbiology laboratories as a tool for surveillance of infectious diseases in Belgium . Epidemiol Infect . 1991 4 ;106 (2 ):297 –303 . 2019299 2 Agentschap zorg en gezondheid. Meldingsplichtige infectieziekten. Available from: https://www.zorg-en-gezondheid.be/overzicht-richtlijnen-infectieziekten 3 Matra. Déclaration des maladies transmissibles. Available from: https://www.wiv-isp.be/matra/CF/liste_matra.aspx 4 Institute of Public Health (Belgium). PediSurv. Available from: https://epidemio.wiv-isp.be/ID/Surveillance/Pages/pedisurv.aspx 5 Institute of Public Health (Belgium). Surveillance of sexually transmitted infections. Available from: https://epidemio.wiv-isp.be/ID/Surveillance/Pages/STI.aspx 6 Muyldermans G , Litzroth A , Ducoffre G , Quoilin S , Medical-Technical Advisory Board . Correction: Establishment and reinforcement of the National Reference Centers for Human Microbiology in Belgium . Arch Public Health . 2013 4 22 ;71 (1 ):8 doi: 10.1186/0778-7367-71-8 23607794 7 Stroobant A , Lamotte JM , Van Casteren V , Cornelis R , Walckiers D , Colyn Y . Epidemiological surveillance of measles through a network of sentinel general practitioners in Belgium . Int J Epidemiol . 1986 9 ;15 (3 ):386 –91 . 3771076 8 Dufour B , La Vieille S . Epidemiological surveillance of infectious diseases in France . Vet Res . 2000 Mar-Apr ;31 (2 ):169 –85 . 10779198 9 Vandenbergh H. Estimation de la représentativité nationale du réseau des laboratoires vigies. Brochure du séminaire ‘Diagnostic et surveillance des maladies infectieuses ‘WIV-ISP, 2004. Available from: www.wiv-isp.be/epidemio/epifr/plabfr/semfr/sem20.pdf 10 Berger N , Muyldermans G , Dupont Y , Quoilin S . Assessing the sensitivity and representativeness of the Belgian Sentinel Network of Laboratories using reimbursement data . 2016 Arch of Public Health , 74 : 29 .27504181 11 Institute of Public Health (Belgium). Laboratory case definitions. Available from: https://epidemio.wiv-isp.be/ID/Surveillance/Pages/sentinelLabs.aspx 12 European Centre for Disease Prevention and Control (ECDC). Laboratory case definitions (ECDC) Available from: http://ec.europa.eu/health/ph_threats/com/docs/1589_2008_en.pdf 13 Centers for Disease Control and Prevention (CDC). Case Definitions for Infectious Conditions Under Public Health. Available from: http://www.cdc.gov/mmwr/PDF/rr/rr4610.pdf 14 Bernard A , Fierens S . The Belgian PCB/dioxin incident: a critical review of health risks evaluations . Int J Toxicol . 2002 Sep-Oct ;21 (5 ):333 –40 . 12396678 15 Vellinga A , Van Loock F . The dioxin crisis as experiment to determine poultry-related campylobacter enteritis . Emerg Infect Dis . 2002 1 ;8 (1 ):19 –22 . 11749743 16 European Centre for Disease Prevention and Control (ECDC). Annual epidemiological report: Sexually transmitted infections including HIV and blood-borne viruses, 2014. Available from: http://ecdc.europa.eu/en/publications/Publications/sexually-transmited-infections-HIV-AIDS-blood-borne-annual-epi-report-2014.pdf 17 Surveillance des infections sexuellement transmissibles dans la population générale–données de 2013 pour la Belgique et les 3 régions. Available from: https://epidemio.wiv-isp.be/ID/diseases/Documents/STI%20uploads/Rapport_IST_2013.pdf 18 Institute of Public Health (Belgium). Annual report national reference center Neisseria meningitidis. Available from: http://bacterio.wiv-isp.be/reporting/reports 19 Hanquet G , Ducoffre G , Vergison A , Neels P , Sabbe M , Van Damme P , Van Herck K . Impact of rotavirus vaccination on laboratory confirmed cases in Belgium . Vaccine . 2011 6 24 ;29 (29–30 ):4698 –703 . doi: 10.1016/j.vaccine.2011.04.098 21571023 20 European Centre for Disease Prevention and Control (ECDC). Annual epidemiological report: Reporting on 2011 surveillance data and 2012 epidemic intelligence data. 2013. Available from: http://ecdc.europa.eu/en/publications/Publications/Annual-Epidemiological-Report-2013.pdf 21 Walckiers D , Piot P , Stroobant A , Van der Veken J , Declercq E . Declining trends in some sexually transmitted diseases in Belgium between 1983 and 1989 . Genitourin Med . 1991 10 ;67 (5 ):374 –7 . 1743708 22 Voldstedlund M , Haarh M , Mølbak K ; MiBa Board of Representatives . The Danish Microbiology Database (MiBa) 2010 to 2013 . Euro Surveill . 2014 1 9 ;19 (1 ).

PMC005xxxxxx/PMC5003366.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757120010.1371/journal.pone.0161684PONE-D-16-00266Research ArticleBiology and Life SciencesBiotechnologyMedical Devices and EquipmentCathetersMedicine and Health SciencesMedical Devices and EquipmentCathetersBiology and Life SciencesOrganismsBacteriaStaphylococcusBiology and Life SciencesMicrobiologyMedical MicrobiologyMicrobial PathogensBacterial PathogensStaphylococcusMedicine and Health SciencesPathology and Laboratory MedicinePathogensMicrobial PathogensBacterial PathogensStaphylococcusSocial SciencesEconomicsEconomic AnalysisCost-Effectiveness AnalysisBiology and Life SciencesOrganismsBacteriaStaphylococcusStaphylococcus AureusBiology and Life SciencesMicrobiologyMedical MicrobiologyMicrobial PathogensBacterial PathogensStaphylococcusStaphylococcus AureusMedicine and Health SciencesPathology and Laboratory MedicinePathogensMicrobial PathogensBacterial PathogensStaphylococcusStaphylococcus AureusBiology and Life SciencesAnatomyBody FluidsBloodMedicine and Health SciencesAnatomyBody FluidsBloodBiology and Life SciencesPhysiologyBody FluidsBloodMedicine and Health SciencesPhysiologyBody FluidsBloodMedicine and Health SciencesHematologyBloodMedicine and Health SciencesInfectious DiseasesBacterial DiseasesBacteremiaBiology and life sciencesOrganismsBacteriaStaphylococcusStaphylococcus aureusMethicillin-resistant Staphylococcus aureusBiology and life sciencesMicrobiologyMedical microbiologyMicrobial pathogensBacterial pathogensStaphylococcusStaphylococcus aureusMethicillin-resistant Staphylococcus aureusMedicine and health sciencesPathology and laboratory medicinePathogensMicrobial pathogensBacterial pathogensStaphylococcusStaphylococcus aureusMethicillin-resistant Staphylococcus aureusMedicine and Health SciencesPharmacologyDrugsAntimicrobialsAntibioticsBiology and Life SciencesMicrobiologyMicrobial ControlAntimicrobialsAntibioticsRapid Diagnosis of Staphylococcal Catheter-Related Bacteraemia in Direct Blood Samples by Real-Time PCR Rapid Diagnosis of Staphylococcal Catheter-Related Bacteraemia by RT-PCRZboromyrska Yuliya 1De la Calle Cristina 2Soto Marcelo 3Sampietro-Colom Laura 3Soriano Alex 2Alvarez-Martínez Míriam José 1Almela Manel 1Marco Francesc 14Arjona Ruth 1Cobos-Trigueros Nazaret 2Morata Laura 2Mensa José 2Martínez José Antonio 2Mira Aurea 5http://orcid.org/0000-0002-8025-3926Vila Jordi 14*1 Department of Clinical Microbiology, Biomedical Diagnostic Centre (CDB), Hospital Clínic, School of Medicine, University of Barcelona, Barcelona, Spain2 Department of Infectious Diseases, Hospital Clínic, University of Barcelona, Barcelona, Spain3 Health Technology Assessment Unit, Hospital Clínic, University of Barcelona, Barcelona, Spain4 ISGlobal, Barcelona Centre for International Health Research (CRESIB), Hospital Clínic, University of Barcelona, Barcelona, Spain5 CDB, Hospital Clínic, School of Medicine, University of Barcelona, Barcelona, SpainFriedrich Alex EditorUniversity Medical Center Groningen, NETHERLANDSCompeting Interests: We received funding from Cepheid, however, this does not alter our adherence to PLOS ONE policies on sharing data and materials. Conceptualization: JM JV. Data curation: MS LSC. Formal analysis: JAM MS AS JV. Funding acquisition: AM. Investigation: YZ CDC MA NCT LM RA. Methodology: JV AS MS LS. Project administration: MJAM. Resources: JV FM. Software: JAM MS LSC. Supervision: JV. Writing – original draft: YZ MS. Writing – review & editing: JV AS. * E-mail: jvila@clinic.ub.es29 8 2016 2016 11 8 e01616845 1 2016 10 8 2016 © 2016 Zboromyrska et al2016Zboromyrska et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Catheter-related bacteremia (CRB) is an important cause of morbidity and mortality among hospitalized patients, being staphylococci the main etiologic agents. The objective of this study was to assess the use of a PCR-based assay for detection of staphylococci directly from blood obtained through the catheter to diagnose CRB caused by these microorganisms and to perform a cost-effectiveness analysis. A total of 92 patients with suspected CRB were included in the study. Samples were obtained through the catheter. Paired blood cultures were processed by standard culture methods and 4 ml blood samples were processed by GeneXpert-MRSA assay for the detection of methicillin-susceptible (MSSA) or methicillin-resistant (MRSA) Staphylococcus aureus, and methicillin-resistant coagulase-negative staphylococci (MR-CoNS). Sixteen CRB caused by staphylococci were diagnosed among 92 suspected patients. GeneXpert detected 14 out of 16 cases (87.5%), including 4 MSSA and 10 MR-CoNS in approximately 1 hour after specimen receipt. The sensitivity and specificity of GeneXpert were 87.5% (CI 95%: 60.4–97.8) and 92.1% (CI 95%: 83–96.7), respectively, compared with standard culture methods. The sensitivity of GeneXpert for S. aureus was 100%. Regarding a cost-effectiveness analysis, the incremental cost of using GeneXpert was of 31.1€ per patient while the incremental cost-effectiveness ratio of GeneXpert compared with blood culture alones was about 180€ per life year gained. In conclusion, GeneXpert can be used directly with blood samples obtained through infected catheters to detect S. aureus and MR-CoNS in approximately 1h after sampling. In addition, it is cost-effective especially in areas with high prevalence of staphylococcal CRB. http://dx.doi.org/10.13039/501100002943Departament d'Innovació, Universitats i Empresa, Generalitat de Catalunya2014SGR0653http://orcid.org/0000-0002-8025-3926Vila Jordi Ministerio de Economía y Competitividad, Instituto de Salud Carlos IIIREIPI RD12/0015http://orcid.org/0000-0002-8025-3926Vila Jordi Cepheidhttp://orcid.org/0000-0002-8025-3926Vila Jordi This study was funded by Cepheid (Sunnyvale, CA; USA). This study was supported by grant 2014SGR0653 from the Departament de Universitats, Recerca i Societat de la Informació de la Generalitat de Catalunya, by the Ministerio de Economía y Competitividad, Instituto de Salud Carlos III, co-financed by European Regional Development Fund (ERDF) “A Way to Achieve Europe,” the Spanish Network for Research in Infectious Diseases (REIPI RD12/0015). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. We received fundings from Cepheid, however, this does not alter our adherence to PLOS ONE policies on sharing data and materials. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Catheter-related bloodstream infection (CRBSI) is an important cause of morbidity and mortality among hospitalized patients. Staphylococci are the main aetiological agents of CRBSI [1]. Bacteremia due to Staphylococcus aureus is a severe disease with a high risk of complications and a high mortality rate [2]. In addition, coagulase-negative staphylococci (CoNS) are also important aetiological agents of CRBSI [3]. Early adequate antibiotic treatment of staphylococcal bacteremia is associated with a better prognosis [4] and early removal of the catheter reduces the risk of developing haematogenous complications [5]. However, the conventional blood culture (BC) method usually requires more than 16 h for bacterial growth detection in the case of staphylococci [6]. To address this problem, several molecular assays have been developed in recent years for direct detection of pathogens in whole blood samples [7]. Although these assays provide results more rapidly, the low bacterial density in blood during bacteremia (1–10 CFU/mL) often results in low sensitivity [8, 9]. However, the intraluminal density of bacteria in CRBSI cases is high (>1,000 CFU/ml) [10–12], which explains the shorter time-to-positivity (TTP) of BC obtained from a catheter compared to those from a peripheral vein [13]. We therefore hypothesized that a rapid Real-Time Polymerase Chain Reaction (rt-PCR)-based assay performed in blood obtained through the infected catheter would have a high sensitivity for detecting the presence of microorganisms causing CRBSI. The main objective of this study was to assess the use of the GeneXpert MRSA/SA BC assay (Cepheid, Sunnyvale, USA) to detect methicillin-susceptible (MSSA), methicillin-resistant (MRSA) S. aureus, and methicillin-resistant CoNS (MR-CoNS) directly from blood obtained through the catheter and to perform a cost-effectiveness analysis comparing this new test to the traditional BC method. Materials and Methods Sample Collection From October 2012 to October 2014 we collected 92 whole blood samples from 92 patients in whom CRBSI was suspected during hospitalization in our centre, a 700-bed university hospital. All patients were prospectively followed until a definitive source of bacteremia was established. Criteria for suspecting CRBSI included: sudden onset of fever with local inflammatory signs at the catheter insertion site, fever without local signs or any other evident source of infection. In all cases, 4 ml of whole blood were obtained through the catheter and submitted to the Microbiology laboratory in a sterile EDTA-containing tube for the GeneXpert assay and blood samples for paired BC (about 10 ml for a single BC vial) were drawn sequentially from the catheter and a peripheral vein without waiting between draws. Patients were included in the study during morning working time, and GeneXpert was performed within 1 hour before sample collection for BC in all cases. The criteria of the definitive diagnosis of CRBSI were: isolation of the same microorganism from BCs obtained from the catheter and peripheral vein with a differential TTP of ≥2 hours or the isolation of the same microorganism from both a catheter tip and at least one percutaneous BC [1]. Microorganisms isolated from positive BC were considered as contaminants if commensal bacteria were detected in only one set of paired BC or different CoNS were isolated from each set of BC and another source of fever was identified. The study was approved by the Hospital Clinic of Barcelona Ethics Committee (study no. 2012/7798). Patient records were anonymized and de-identified prior analysis. No informed consent was considered necessary for this study. Routine microbiological techniques BCs were collected and transported to the laboratory in less than 2 hours. BCs vials were incubated in BACTEC FX (Becton Dickinson, MD, USA) for a maximum of 120 h. In case of positivity, Gram staining and subculture on appropriate solid media were performed. The TTP of each BC bottle was recorded. If the catheter was removed, the Maki roll-on semiquantitative method for catheter tip culture was used. The final identification of bacterial species was achieved using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) (Bruker Daltonics, Bremen, Germany). Routine antimicrobial susceptibility included the Phoenix system (Becton Dickinson, MD, USA) for S. aureus and the disk diffusion method for CoNS. Results of susceptibility testing were interpreted according to EUCAST guidelines (http://www.eucast.org). The GeneXpert assay To recover bacteria from whole blood, the EDTA collection tube was centrifuged at 430 × g for 5 min and then the supernatant was centrifuged at 15,600 × g for 2 min. The pellet obtained was resuspended in 100 μl of sterile saline solution (0.9% NaCl) and used for the GeneXpert test according to the manufacturer's instructions. The GeneXpert used was Xpert MRSA-SA BC G3 Version 24. GeneXpert was considered positive for MR-CoNS if only the methicillin resistance gene (mecA) was detected; positive for MSSA if only the staphylococcal protein A (spa) target was detected; and positive for MRSA if the spa, mecA genes and the staphylococcal chromosome genomic island (SCCmec) were identified. The current cycle threshold (Ct) cut-off values of the GeneXpert assay were established for positive BCs containing high bacterial concentrations, with a valid maximum cycle for all three targets of 36. Since the expected bacterial inoculum in whole blood is much lower, two bacterial strains of S. aureus (MSSA and MRSA) were used to study the analytical detection limits of GeneXpert assay. Four millilitres of whole blood were spiked with bacteria to achieve three different concentrations (10, 100 and 1000 CFU/ml). The GeneXpert assay was performed in duplicate for each S. aureus strain and each bacterial concentration. Taking into account the results of the detection limits study, prolonged Ct values (>36) were also considered as positive. Cost-effectiveness analysis The information was not provided to the physician in charge of the patient. Therefore, we performed an analysis according to the sensitivity and specificity of the test and using previous literature about complications related with S. aureus catheter-related bacteremia [5]. A cost-effectiveness analysis (CEA) of using GeneXpert to detect CRBSI caused by S. aureus and MR-CoNS was performed. The comparator was the conventional BC method. The main clinical outcome considered was the expected number of life years gained. Costs were measured from the perspective of our hospital and were obtained from hospital sources. Differences in costs and life years between testing strategies were used to compute the incremental cost-effectiveness ratio (ICER). A decision tree based on the standard clinical approach to CRBSI was developed. The tree was used to model outcomes following different strategies (Fig 1). Individuals enter the model as patients with suspected CRBSI. All patients are tested with conventional BC. Under the standard protocol (lower branch of the tree) all patients receive empirical wide-spectrum antibiotics and a fraction of patients have the catheter removed before BC results are available. The removal decision is initially based solely on clinical criteria. In patients tested with GeneXpert (upper branch) the catheter is removed depending on the test result, which is known before BC results. The model assumes that a positive test leads to immediate catheter removal and patients are treated with specific antibiotics. 10.1371/journal.pone.0161684.g001Fig 1 Decision tree for cost-effectiveness analysis. CRBSI, catheter-related bloodstream infection; BC, blood culture; -, negative; +, positive. Variables determining the probability of each health outcome are: the prevalence rate of S. aureus and MR-CoNS among suspected patients, the sensitivity and specificity of the GeneXpert test, the probability of haematogenous complications when the catheter is removed on suspicion of CRBSI and when catheter removal is delayed until BC results are available, and the probability of death related to CRBSI. The S1 File provides a detailed description of the economic model and data sources. Resources considered for the cost analysis included BC costs, the GeneXpert kit, technical staff, catheters, wide-spectrum and specific antibiotics; imaging (abdominal ultrasound, transthoracic echocardiogram, and PET scan) and the use of an intensive care unit in case of haematogenous complications. Treatment of co-morbidities or CRBSI caused by bacteria other than staphylococci was not considered in the cost analysis. One-way deterministic and multivariate probabilistic sensitivity analyses were performed to evaluate the robustness of CEA. Statistical analysis Statistical analysis was performed using SPSS software (SPSS, Chicago, IL). Differences were considered significant with a P< 0.05. Results The results of the analytical detection limits study were as follows: samples with 10 CFU/ml bacterial concentration were negative by GeneXpert for the three targets (spa, mecA, SCCmec); samples with 100 CFU/ml were positive but with prolonged Ct values (>36); samples with 1000 CFU/ml were positive with recommended Ct cut-off values (≤36) for the three targets. We evaluated 92 blood samples obtained from 92 patients and processed with paired BCs and GeneXpert. The source of bacteremia was presumed to be a short peripheral intravenous catheter in 3 patients (3.2%), a peripherally inserted central catheter in 18 (19.6%), a central subclavian or jugular venous catheter in 63 (68.5%), and a tunnelled catheter in 8 patients (8.7%). The median (IQR) length of stay of patients was 13 (5–22) days and the median catheter dwell time was 12 (7–23) days. BCs were positive in 34 patients. Eighteen cases did not fulfil the CRBSI definition. In 15 cases, the microorganisms isolated (14 CoNS and 1 viridans group Streptococcus) were considered contaminants. None of the 14 contaminant CoNS was detected by GeneXpert. No specific treatment was prescribed in these cases. Two patients had bacteremia by Gram-negative bacilli, with the suspected source being a urinary tract infection in both; and the other patient had bacteremia due to Enterococcus faecium from an unknown source. A total of 16 CRBSI were diagnosed among 92 suspected patients, therefore, the prevalence of staphylococcal CRBSI in our study population was 17.4%. Among these 16 CRBSI no septic metastasis and no death related was observed. GeneXpert detected 14 out of 16 cases (87.5%), including 4 MSSA and 10 MR-CoNS. Table 1 shows the concordance between definitive CRBSI cases and GeneXpert results. There were 2 false negative GeneXpert results, both caused by MR-CoNS. 10.1371/journal.pone.0161684.t001Table 1 Comparison of results obtained by GeneXpert and convencional BC. No. of patients Positive BC (%) Negative BC (%) Total (%) GeneXpert positive 14 (15.2)a 6 (6.5)b 20 (21.7) GeneXpert negative 2 (2.2) 70 (76.1)c 72 (78.3) Total (%) 16 (17.4) 76 (82.6) 92 (100) BC, blood culture a Including 4 Staphylococcus aureus detected. b Including 3 S. aureus detected. c Including 14 contaminant CoNS. The median TTP in the BC of the 10 MR-CoNS detected by GeneXpert was 16.6h (IQR: 12.1–17.8), whereas the TTP of two MR-CoNS not detected by GeneXpert was 20.8 h and 24.1 h, respectively. Interestingly, the median TTP of not detected CoNS, including 14 contaminants and 2 false negatives, was 25.5 h (IQR: 20.0–41.8). The difference between the median TTP of MR-CoNS detected by GeneXpert (16.6 h) and those not detected (25.5 h) was statistically significant (P = 0.005). On the other hand, GeneXpert identified 6 additional staphylococci (presumably false positives), including 3 MR-CoNS, 2 MSSA, and one sample, which offered two possibilities since the genes mecA and spa but not SCCmec were detected. These two possibilities are: 1) that both MR-CoNS and MSSA were present or 2) a MRSA strain with a SCCmec variant that was not detected. Indeed, the last option seems more plausible since the Ct values for both mecA and spa genes were similar with 37.7 and 37.5, respectively. Interestingly, 4 out of these 6 (66.7%) false positives were receiving active antibiotic treatment initiated at least 24 h before sampling. Briefly, one patient with MSSA was treated with ceftriaxone, the other patient with MSSA with meropenem and vancomycin; one patient with mixed MSSA and MR-CoNS was receiving meropenem and linezolid, and the other with MR-CoNS, received vancomycin. In contrast, among the 14 patients with GeneXpert results concordant to CRBSI diagnosis, only 2 (2/14, 14.3%) patients were receiving correct antimicrobial treatment for at least 24 h before sampling (P = 0.037). The sensitivity, specificity, positive predictive value, and negative predictive value of GeneXpert in the present study were 87.5% (CI 95%: 60.4–97.8), 92.1% (CI 95%: 83–96.7), 70% (CI 95%: 45.7–87.2) and 97.2% (CI 95%: 89.4–99.5), respectively. The sensitivity, specificity, positive predictive value, and negative predictive value of GeneXpert for S. aureus were 100% (CI 95%: 39.6–100), 96.6% (CI 95%: 89.7–99.1), 57.1% (CI 95%: 20.2–88.2) and 100% (CI 95%: 94.6–100), respectively. The sensitivity of GeneXpert for MR-CoNS was 83.3%. Cost-effectiveness results Clinical outcomes Using prevalence figures in the hospital and mortality rates published in the literature[14], the probability of death for patients with suspected CRBSI was estimated at 2.59% for patients tested with conventional BC alone and 1.59% for patients tested with GeneXpert (Table 2). Taking life expectancy of the general population in Spain as reference, the average discounted life expectancy (with a 3% discount rate) for patients tested with BC and with GeneXpert were 17.01 and 17.18 years, respectively. The use of GeneXpert allowed earlier detection of 14 cases of CRBSI among 92 suspected patients (15.2%). 10.1371/journal.pone.0161684.t002Table 2 Cost-effectiveness results (base case). GeneXpert and BC Only BC Difference Clinical outcomes Death probability 1.59% 2.59% -1.00% Expected Life years (discounted) 17.183 17.010 0.173 Early detections 15.2% - 15.2% Costs Test price (€) 67+60 60 67 Cost per patient (€) 411.5 380.4 31.1 Cost-effectiveness (ICER) Incremental cost (€) per life year gained (discounted) 179.5 Incremental cost (€) per early detected case 204.2 BC, blood culture; ICER, incremental cost-effectiveness ratio. Costs and cost-effectiveness Costs per test were 60€ for one BC (on average, 15€ for one BC vial) and 67€ for the GeneXpert test. In the baseline analysis the cost per patient were 380.4€ and 411.5€, respectively. As mentioned in the methods section, these numbers only take into account testing and treatment of CRBSI caused by S. aureus and MR-CoNS. Because the use of GeneXpert is associated with a lower probability of haematogenous complications due to early detection and, therefore, lower treatment costs, the incremental cost of using GeneXpert is only of 31.1€ per patient. The incremental cost-effectiveness ratio (ICER) of GeneXpert compared with BC alone was about 180€ per life year gained. Similarly, the ICER per early detection was estimated at 204€. The sensitivity analysis showed that the probability of haematogenous complication when the catheter removal is delayed, the joint prevalence of S. aureus and MR-CoNS and the sensitivity of the GeneXpert test were the parameters that introduced the highest variability in the ICER (Fig 2). For complication probability in case of delayed catheter removal, the ICER for the GeneXpert test ranged between 9€ per life year gained (when the probability = 42%) and 2,405€ (when probability = 13%). As regards the influence of prevalence, the ICER for the GeneXpert test ranged between minus 46€ per life year gained (i.e., dominance of GeneXpert) when the prevalence rate was 50% and 435€ per life year gained when the prevalence was lower (i.e., 10%). Changes in sensitivity also affected ICER, which varied from 124€ to 470€ per life year gained for high (98%) and low (62%) sensitivity rates, respectively. The remaining parameters of the model (complication probability when the removal of the catheter is not delayed, the unconditional death probability for true positives, life expectancy, and specificity of GeneXpert) introduced less variability in the ICER. 10.1371/journal.pone.0161684.g002Fig 2 Univariate deterministic sensitivity analysis. This analysis addresses changes in incremental cost-effectiveness ratio (€/life year gained) when individual parameter values are varied. The vertical line corresponds to the base case incremental cost-effectiveness ratio of the model. CRBSI, catheter-related bloodstream infection Results from the probabilistic sensitivity analysis are summarized in Fig 3. At a willingness-to-pay of 0€ per life year gained GeneXpert had a 27% probability of being cost-effective compared with BC. As the willingness-to pay increased, the probability that Genexpert was cost-effective quickly raised to 98%. 10.1371/journal.pone.0161684.g003Fig 3 The probabilistic sensitivity analysis. Discussion In this study, we demonstrated that GeneXpert is a useful and rapid test for the detection of S. aureus and MR-CoNS directly in blood samples obtained through an infected catheter. As GeneXpert results are available in approximately 1h, this test may help clinicians to early select the appropriate antibiotic treatment and to decide whether immediate catheter removal is necessary or not. A delay in the initiation of appropriate antibiotic therapy and prompt catheter removal are critical variables associated with the outcome of patients with CRBSI and the risk of developing septic metastases [5, 15]. In addition, as vancomycin treatment compared with a beta-lactam has been associated with a higher mortality rate in case of MSSA bacteremia, a rapid discrimination between MSSA and MRSA should improve the outcome of these patients.[16] Previously reported sensitivity of the GeneXpert from positive BC for MSSA and MRSA was 100% and 98.3%, respectively [17]. In the present study, the global sensitivity performed using blood obtained from an infected catheter was 87.5%, including MR-CoNS, and it increases up to 100% for MSSA. According to a recent study from our institution, CoNS are the aetiology of 45% of CRBSI cases with a rate of methicillin resistance higher than 70% [18]. Interestingly, GeneXpert detects MR-CoNS in case of negative spa gene and positive mecA. To our knowledge, this is the first time that GeneXpert is used for this proposal and our results suggest that this is a promising diagnostic tool particularly because contaminants were not amplified. This could be attributed to the low bacterial density in contaminant cases as it was suggested by a significantly longer median TTP in concomitant BC of negative GeneXpert samples than of positive ones (25.5 h vs 16.6 h, P = 0.005). However, there were 2 true CRBSI due to MR-CoNS not detected by GeneXpert. In these cases the TTP was also ≥20 h and this is a limitation but it probably indicates less severe infections [19]. An interesting finding was the identification of six presumably false positive results. The most reasonable explanation is the previous exposure to active antibiotics, which inhibited their growth in culture media. However, we cannot rule out false positive results due to unspecific amplification. Previously, the GeneXpert assay had been shown to have an important impact on the mean length of hospital stay and the mean hospital costs for bacteremic patients [20]. In another study, using GeneXpert results to guide anti-infective therapy proved to be less costly than empiric treatment for MRSA and also showed the potential to reduce mortality rates [21]. Regarding our results, the estimated ICER was only about 180€ per life year gained. Although there are no well-defined cost-effectiveness thresholds for life years gained, there are thresholds for quality adjusted life years (i.e., QALYs; range US $20,000/16,800€ and US $100,000/84,000€) [22]. The ICER estimated here is well below these figures; if any adjustment for quality were made the main results will not be affected. There are several limitations in our study. First, the number of patients included was small and the study was performed in a single hospital. Second, as GeneXpert allows detection of only S. aureus and MR-CoNS, CRBSI caused by other etiological agents, including methicillin-susceptible CoNS, cannot be identified by this system. Third, results of the GeneXpert test were not taken into account for patients’ management; this fact and the low prevalence of S.aureus CRB in our sample led to take data from the literature regarding potential complications of bacteremia. Therefore, CEA represents a partially literature-derived model. Fourth, the numbers of discounted life years gained were likely to overestimate real gains in life years because a large share of patients with suspected bacteremia have a clinical condition different from staphylococcal CRBSI that can potentially lead to death before reaching the general population’s expected age of death. Similarly, no attempt was done to estimate gains in quality-adjusted life years. The heterogeneity of patients’ original health condition would have made such calculation pointless. However, given that the GeneXpert test was found to be highly cost-effective in terms of unadjusted gains of life years, adjustments for quality are unlikely to change the qualitative results. Finally, the life expectancy figures could be overestimated, because large share of patients with CRBSI have a clinical condition (different from staphylococcal CRBSI) that can potentially lead to death before reaching the general population’s expected age of death. However, given that the bias applies to the estimated life expectancy obtained with both tests, the change in life years of using GeneXpert compared with BC is not likely to suffer a significant bias. In conclusion, GeneXpert can be used directly with whole blood sample obtained through an infected catheter for the detection of S. aureus and MR-CoNS in approximately 1 h after sampling. In addition, GeneXpert is cost-effective, especially in settings with a high prevalence of staphylococcal CRBSI. Finally, this test has a potential to identify cases not detected by BC, probably due to ongoing antibiotic treatment. However, further studies are necessary to confirm this finding. Supporting Information S1 File Economic evaluation of the use of GeneXpert to detect staphylococcal catheter-related bloodstream infection (CRBSI). (DOCX) Click here for additional data file. This study was funded by Cepheid (Sunnyvale, CA; USA). This study was supported by grant 2014SGR0653 from the Departament de Universitats, Recerca i Societat de la Informació de la Generalitat de Catalunya, by the Ministerio de Economía y Competitividad, Instituto de Salud Carlos III, co-financed by European Regional Development Fund (ERDF) “A Way to Achieve Europe,” the Spanish Network for Research in Infectious Diseases (REIPI RD12/0015). Study results were presented at the 54th Interscience Conference on Antimicrobial Agents and Chemotherapy (ICAAC), Washington, DC, USA, September 5–9, 2014. ==== Refs References 1 Mermel LA , Allon M , Bouza E , Craven DE , Flynn P , O'Grady NP , et al Clinical practice guidelines for the diagnosis and management of intravascular catheter-related infection: 2009 Update by the Infectious Diseases Society of America . Clinical infectious diseases: an official publication of the Infectious Diseases Society of America . 2009 ;49 (1 ):1 –45 . Epub 2009/06/06. 10.1086/599376 ; PubMed Central PMCID: PMCPmc4039170.19489710 2 Laupland KB , Ross T , Gregson DB . Staphylococcus aureus bloodstream infections: risk factors, outcomes, and the influence of methicillin resistance in Calgary, Canada, 2000–2006 . The Journal of infectious diseases . 2008 ;198 (3 ):336 –43 . Epub 2008/06/05. 10.1086/589717 .18522502 3 Lorente L , Henry C , Martin MM , Jimenez A , Mora ML . Central venous catheter-related infection in a prospective and observational study of 2,595 catheters . Crit Care . 2005 ;9 (6 ):R631 –5 . Epub 2005/11/11. 10.1186/cc3824 ; PubMed Central PMCID: PMCPmc1414031.16280064 4 Soriano A , Martinez JA , Mensa J , Marco F , Almela M , Moreno-Martinez A , et al Pathogenic significance of methicillin resistance for patients with Staphylococcus aureus bacteremia . Clinical infectious diseases: an official publication of the Infectious Diseases Society of America . 2000 ;30 (2 ):368 –73 . Epub 2000/02/15. 10.1086/313650 .10671343 5 Fowler VG Jr., Justice A , Moore C , Benjamin DK Jr., Woods CW , Campbell S , et al Risk factors for hematogenous complications of intravascular catheter-associated Staphylococcus aureus bacteremia . Clinical infectious diseases: an official publication of the Infectious Diseases Society of America . 2005 ;40 (5 ):695 –703 . Epub 2005/02/17. 10.1086/427806 .15714415 6 Martinez JA , Pozo L , Almela M , Marco F , Soriano A , Lopez F , et al Microbial and clinical determinants of time-to-positivity in patients with bacteraemia . Clinical microbiology and infection . 2007 ;13 (7 ):709 –16 . Epub 2007/05/09. 10.1111/j.1469-0691.2007.01736.x .17484763 7 Liesenfeld O , Lehman L , Hunfeld KP , Kost G . Molecular diagnosis of sepsis: New aspects and recent developments . European journal of microbiology & immunology . 2014 ;4 (1 ):1 –25 . Epub 2014/03/29. 10.1556/EuJMI.4.2014.1.1 ; PubMed Central PMCID: PMCPmc3955828.24678402 8 Dark P , Blackwood B , Gates S , McAuley D , Perkins GD , McMullan R , et al Accuracy of LightCycler® SeptiFast for the detection and identification of pathogens in the blood of patients with suspected sepsis: a systematic review and meta-analysis . Intensive care medicine . 2015 ;41 (1 ):21 –33 . Epub 2014/11/25. 10.1007/s00134-014-3553-8 .25416643 9 Yagupsky P , Nolte FS . Quantitative aspects of septicemia . Clinical microbiology reviews . 1990 ;3 (3 ):269 –79 . Epub 1990/07/01. ; PubMed Central PMCID: PMCPmc358159.2200606 10 Wing EJ , Norden CW , Shadduck RK , Winkelstein A . Use of quantitative bacteriologic techniques to diagnose catheter-related sepsis . Archives of internal medicine . 1979 ;139 (4 ):482 –3 . Epub 1979/04/01. .107870 11 Andremont A , Paulet R , Nitenberg G , Hill C . Value of semiquantitative cultures of blood drawn through catheter hubs for estimating the risk of catheter tip colonization in cancer patients . Journal of clinical microbiology . 1988 ;26 (11 ):2297 –9 . Epub 1988/11/01. ; PubMed Central PMCID: PMCPmc266880.3069861 12 Quilici N , Audibert G , Conroy MC , Bollaert PE , Guillemin F , Welfringer P , et al Differential quantitative blood cultures in the diagnosis of catheter-related sepsis in intensive care units . Clinical infectious diseases: an official publication of the Infectious Diseases Society of America . 1997 ;25 (5 ):1066 –70 . Epub 1997/12/24. .9402359 13 Blot F , Nitenberg G , Chachaty E , Raynard B , Germann N , Antoun S , et al Diagnosis of catheter-related bacteraemia: a prospective comparison of the time to positivity of hub-blood versus peripheral-blood cultures . Lancet . 1999 ;354 (9184 ):1071 –7 . Epub 1999/10/06. .10509498 14 Cosgrove SE , Qi Y , Kaye KS , Harbarth S , Karchmer AW , Carmeli Y . The impact of methicillin resistance in Staphylococcus aureus bacteremia on patient outcomes: mortality, length of stay, and hospital charges . Infection control and hospital epidemiology: the official journal of the Society of Hospital Epidemiologists of America . 2005 ;26 (2 ):166 –74 . Epub 2005/03/11. 10.1086/502522 .15756888 15 Molina J , Penuela I , Lepe JA , Gutierrez-Pizarraya A , Gomez MJ , Garcia-Cabrera E , et al Mortality and hospital stay related to coagulase-negative Staphylococci bacteremia in non-critical patients . The Journal of infection . 2013 ;66 (2 ):155 –62 . Epub 2012/10/30. 10.1016/j.jinf.2012.10.021 .23103291 16 Kim SH , Kim KH , Kim HB , Kim NJ , Kim EC , Oh MD , et al Outcome of vancomycin treatment in patients with methicillin-susceptible Staphylococcus aureus bacteremia . Antimicrobial agents and chemotherapy . 2008 ;52 (1 ):192 –7 . Epub 2007/11/07. 10.1128/aac.00700-07 ; PubMed Central PMCID: PMCPmc2223910.17984229 17 Wolk DM , Struelens MJ , Pancholi P , Davis T , Della-Latta P , Fuller D , et al Rapid detection of Staphylococcus aureus and methicillin-resistant S. aureus (MRSA) in wound specimens and blood cultures: multicenter preclinical evaluation of the Cepheid Xpert MRSA/SA skin and soft tissue and blood culture assays . Journal of clinical microbiology . 2009 ;47 (3 ):823 –6 . Epub 2009/01/16. 10.1128/jcm.01884-08 ; PubMed Central PMCID: PMCPmc2650929.19144803 18 Marcos M , Soriano A , Inurrieta A , Martinez JA , Romero A , Cobos N , et al Changing epidemiology of central venous catheter-related bloodstream infections: increasing prevalence of Gram-negative pathogens . The Journal of antimicrobial chemotherapy . 2011 ;66 (9 ):2119 –25 . Epub 2011/06/15. 10.1093/jac/dkr231 .21665905 19 Kassis C , Rangaraj G , Jiang Y , Hachem RY , Raad I . Differentiating culture samples representing coagulase-negative staphylococcal bacteremia from those representing contamination by use of time-to-positivity and quantitative blood culture methods . Journal of clinical microbiology . 2009 ;47 (10 ):3255 –60 . Epub 2009/08/21. 10.1128/jcm.01045-09 ; PubMed Central PMCID: PMCPmc2756952.19692564 20 Bauer KA , West JE , Balada-Llasat JM , Pancholi P , Stevenson KB , Goff DA . An antimicrobial stewardship program's impact with rapid polymerase chain reaction methicillin-resistant Staphylococcus aureus/S. aureus blood culture test in patients with S. aureus bacteremia . Clinical infectious diseases: an official publication of the Infectious Diseases Society of America . 2010 ;51 (9 ):1074 –80 . Epub 2010/10/01. 10.1086/656623 .20879856 21 Brown J , Paladino JA . Impact of rapid methicillin-resistant Staphylococcus aureus polymerase chain reaction testing on mortality and cost effectiveness in hospitalized patients with bacteraemia: a decision model . PharmacoEconomics . 2010 ;28 (7 ):567 –75 . Epub 2010/06/17. 10.2165/11533020-000000000-00000 .20550222 22 Bell CM , Urbach DR , Ray JG , Bayoumi A , Rosen AB , Greenberg D , et al Bias in published cost effectiveness studies: systematic review . BMJ (Clinical research ed) . 2006 ;332 (7543 ):699 –703 . Epub 2006/02/24. 10.1136/bmj.38737.607558.80 ; PubMed Central PMCID: PMCPmc1410902.16495332

PMC005xxxxxx/PMC5003367.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 10.1371/journal.pone.0162227PONE-D-16-33410CorrectionCorrection: Trends in the Prevalence of Depression in Hospitalized Patients with Type 2 Diabetes in Spain: Analysis of Hospital Discharge Data from 2001 to 2011 Lopez-de-Andrés Ana Jiménez-Trujillo Mª Isabel Hernández-Barrera Valentín de Miguel-Yanes José Mª Méndez-Bailón Manuel Perez-Farinos Napoleón de Burgos Lunar Carmen Cárdenas-Valladolid Juan Salinero-Fort Miguel Ángel Jiménez-García Rodrigo Carrasco-Garrido Pilar 29 8 2016 2016 11 8 e0162227© 2016 Lopez-de-Andrés et al2016Lopez-de-Andrés et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Trends in the prevalence of depression in hospitalized patients with type 2 diabetes in Spain: analysis of hospital discharge data from 2001 to 2011 ==== Body There are errors in the Funding section. The correct funding information is as follows: This study was funded by the FIS (Fondo de Investigaciones Sanitarias—Health Research Fund, grant no. PI13/00118, Instituto de Salud Carlos III) co-financed by the European Union through the Fondo Europeo de Desarrollo Regional (FEDER, “Una manera de hacer Europa”). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. ==== Refs Reference 1 Lopez-de-Andrés A , Jiménez-Trujillo MI , Hernández-Barrera V , de Miguel-Yanes JM , Méndez-Bailón M , Perez-Farinos N , et al (2015 ) Trends in the Prevalence of Depression in Hospitalized Patients with Type 2 Diabetes in Spain: Analysis of Hospital Discharge Data from 2001 to 2011 . PLoS ONE 10 (2 ): e0117346 doi: 10.1371/journal.pone.0117346 25706646

PMC005xxxxxx/PMC5003368.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 27570966PONE-D-16-0756410.1371/journal.pone.0161671Research ArticleComputer and Information SciencesComputer ArchitectureComputer HardwareComputer and Information SciencesComputer SoftwareResearch and Analysis MethodsMicroscopyLight MicroscopyComputer and Information SciencesData AcquisitionComputer and Information SciencesComputer ArchitectureUser InterfacesEngineering and TechnologyHuman Factors EngineeringMan-Computer InterfaceGraphical User InterfaceComputer and Information SciencesComputer ArchitectureUser InterfacesGraphical User InterfaceComputer and Information SciencesSoftware EngineeringSoftware DesignEngineering and TechnologySoftware EngineeringSoftware DesignResearch and Analysis MethodsMicroscopyLight MicroscopyFluorescence MicroscopySoftware Framework for Controlling Unsupervised Scientific Instruments Software Framework for Controlling Unsupervised Scientific InstrumentsSchmid Benjamin 12*Jahr Wiebke 1Weber Michael 13Huisken Jan 14* 1 Max Planck Institute of Molecular Cell Biology and Genetics, 01307 Dresden, Germany 2 Optical Imaging Centre Erlangen, Friedrich-Alexander-University of Erlangen-Nuremberg, 91054 Erlangen, Germany 3 Harvard Medical School, Boston, Massachusetts 02115, United States of America 4 Morgridge Institute for Research, Madison, Wisconsin 53715, United States of America Gilestro Giorgio F Editor Imperial College London, UNITED KINGDOM Competing Interests: The authors have declared that no competing interests exist. Conceived and designed the experiments: JH BS MW. Performed the experiments: BS WJ. Analyzed the data: BS. Contributed reagents/materials/analysis tools: BS WJ. Wrote the paper: BS WJ JH. * E-mail: benjamin.schmid@fau.de (BS); jhuisken@morgridge.org (JH)2016 29 8 2016 11 8 e01616712 3 2016 11 7 2016 © 2016 Schmid et al2016Schmid et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Science outreach and communication are gaining more and more importance for conveying the meaning of today’s research to the general public. Public exhibitions of scientific instruments can provide hands-on experience with technical advances and their applications in the life sciences. The software of such devices, however, is oftentimes not appropriate for this purpose. In this study, we describe a software framework and the necessary computer configuration that is well suited for exposing a complex self-built and software-controlled instrument such as a microscope to laymen under limited supervision, e.g. in museums or schools. We identify several aspects that must be met by such software, and we describe a design that can simultaneously be used to control either (i) a fully functional instrument in a robust and fail-safe manner, (ii) an instrument that has low-cost or only partially working hardware attached for illustration purposes or (iii) a completely virtual instrument without hardware attached. We describe how to assess the educational success of such a device, how to monitor its operation and how to facilitate its maintenance. The introduced concepts are illustrated using our software to control eduSPIM, a fluorescent light sheet microscope that we are currently exhibiting in a technical museum. Deutsche Forschungsgemeinschaft (DE)GSC97/3Jahr Wiebke WJ is supported by a DIGS-BB fellowship provided by the DFG in the context of the excellence Initiative. Deutsche Forschungsgesellschaft GSC97/3 (http://www.dfg.de/). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files. The software is publicly available at https://github.com/bene51/eduSPIM.git.Data Availability All relevant data are within the paper and its Supporting Information files. The software is publicly available at https://github.com/bene51/eduSPIM.git. ==== Body Introduction Science outreach can be performed in different ways, e.g. through web publishing, public discussions, school visits or dedicated outreach events. A more exploratory and interactive possibility that is particularly well suited for conveying technical advances in scientific instrument development is the exhibition of new instrument prototypes to the public, e.g. in museums. Oftentimes it is possible to use the hardware of a laboratory prototype more or less unchanged for a public exhibition. The software, however, comprises the interface to the user and must account for the altered role of the instrument. Modern scientific devices are highly complex, and scientists typically receive special training to operate the software. In contrast, visitors in a museum with diverse educational backgrounds explore the instrument by trial. Unlike in a laboratory environment, it is the software’s responsibility to provide an attractive user interface and guarantee robust, non-stop operation with only little or no supervision. In this paper we elaborate the concepts of a modular software framework that is well suited for controlling an autonomously running scientific instrument. Our software focuses on user-friendliness and elaborate error handling. It recovers automatically from errors, originating, e.g. from the communication with the hardware. If a severe outage prohibits correct functioning, our software keeps providing a seemingly functioning instrument by simulating normal operation. The introduced framework allows one to operate a fully functioning device, but can at the same time be used to simulate normal function if mock hardware is attached for illustration purposes only, or in the complete absence of hardware. The concepts introduced here can generally be used to control unsupervised scientific devices, perhaps operated at locations that are difficult to access, e.g. at contaminated areas, in deep ocean or for speleology. As an example, we provide a specific implementation that is used to control eduSPIM [1], an educational light sheet fluorescence microscope (or Selective Plane Illumination Microscope, SPIM [2]) that we have developed recently for an interactive exhibition in a technical museum, to support the UNESCO International Year of Light 2015 [3]. Framework for operating real hardware, mock hardware or a purely virtual instrument Thorough error handling is critical for the unsupervised operation of a fully functional instrument with working hardware attached to it. Otherwise, malfunctioning hardware or errors during the communication with peripherals would immediately lead to software crashes and leave the system unusable. To recover from this, the administrator would be forced to check frequently to ensure continuous functioning. Here, we introduce an elaborate error handling design that allows the software to recover from unforeseen situations. We propose a modular framework that defines an interface for each attached peripheral. For example, a camera interface defines functions for setting the exposure time, starting the acquisition and reading the acquired image data. These interfaces are implemented by individual software modules that handle the communication to the corresponding peripheral, typically using the manufacturer’s application programming interface (API). At the same time we implement a second module for each interface that simulates its operation. In the camera example, the simulating implementation loads and uses pre-acquired data from the hard-drive. The software uses the real implementation for each device during normal operation and without any error occurring. In case of a failure, communication to all devices is closed and re-initialized. If the error persists, the software exits completely to make sure that all hardware resources are released properly. To provide ongoing operation in such a case, the software is started from a batch script repeatedly from inside a loop. In case it exits with an error code indicating a hardware problem, it is restarted automatically (Fig 1). If any error occurs during startup, the software switches to the simulating versions of all modules, thereby providing a seemingly functional instrument. In case an error occurs that renders the instrument completely unusable (e.g. user input cannot be gathered any more), even the simulating versions won’t allow a usable device, and an emergency screen notifies users about the problem and asks for their patience until the device has been repaired. 10.1371/journal.pone.0161671.g001Fig 1 Error handling during initialization and normal operation. The main program was started from a Windows batch script in a loop. If any error occurred during hardware initialization, the software switched to fall-back mode, which simulated normal operation but used a pre-acquired data set. If switching to fall-back mode also failed, the software entered fatal mode, and an error message was displayed. Otherwise, the program waited for the visitor to press one of the buttons, upon which the corresponding functionality was executed. If any error occurred, all hardware devices were closed and re-initialized. If the error persisted, the application exited with an error code. Back in the batch script, the application was re-started as long as the exit code indicated an error. If the software’s purpose is the control of an educational instruments, e.g. an exhibition in a museum or school, there may only be limited resources for the necessary maintenance service. Our software can be be started explicitly in simulating mode, so that the exhibition can continue without disruption until maintenance is possible. Pursuing this idea further, it is even possible to exhibit an instrument with only mock hardware or without any hardware attached to it at all. Only the software needs to be installed, which keeps costs and efforts minimal. Our framework checks autonomously for attached hardware. If none is available, it automatically downloads a pre-acquired example data set. It is therefore straightforward to replicate the exhibit numerous times and use it practically anywhere, e.g. in school lessons or at special events. Computer setup A number of considerations regarding computer setup and configuration are important for guaranteeing continuous availability of an unsupervised instrument. We set up the PC’s BIOS to boot the computer at a fixed time early every morning. We put a shortcut to our software into the Startup folder under the Windows start menu, so that our software starts automatically after booting. It switches immediately to full-screen mode to hide the operating system from users and to focus their attention to the instrument’s software. Our software closes itself at a fixed time every evening, well after any expected usage. Additionally we configured the Windows Task Scheduler to shut down the computer during the night, to save power and freshly initialize the connections to the peripherals every day. To guarantee that no other application window (such as an update notification of the operating system) obscures our software during normal usage times, we implemented a tool for putting the main window of our application into foreground, using the Windows API. This tool was called at a regular interval of 5 seconds from within our software. We decided for a mini PC (Intel NUC), because the space available in the museum was limited, and we wanted to hide the computer and all hardware controllers inside the table to not distract the visitors from the optical setup. As an operating system we are using Microsoft Windows 7 64-bit professional. Monitoring and maintenance For monitoring and remote maintenance of an unsupervised instrument, the computer needs to be connected to the Internet, possibly via WLAN. Our proposed software solution uses a cloud service (Dropbox, https://www.dropbox.com) to synchronize folders with our computers in the lab. Information about any exception that occurred is not only sent to an administrator via email, but also saved and archived to a synchronized folder. Additional emails on successful starts and shut downs inform the administrator about the integrity of the system. For trouble-shooting and occasional maintenance such as installing security updates for the operating system, we used a screen-sharing software (TeamViewer, https://www.teamviewer.com) to log in remotely. In contrast to related software, TeamViewer does not require the computer to have a public IP address, so that the computer can reside behind a router outside of our institute’s local network. Usage statistics and website To assess the success of a remotely operated instrument, it is important to collect and visualize information about its usage and the data acquired by it. We used the logging framework SLF4j (http://www.slf4j.org) to log not only exceptions, but also all actions triggered by the user to a synchronized folder. In another synchronized folder, we regularly saved and archived renderings of the instrument’s output. We analyze, visualize and publish these data on a dedicated website, which is dynamically updated in regular intervals using JavaScript, AJAX and jQuery (https://jquery.com). For eduSPIM, we save stack renderings in a publicly shared folder. To limit the amount of data we only keep one rendering per day. Furthermore, we save the most recently acquired image and a table with the number of times each button was pressed per day. The accompanying website (http://www.eduspim.org, Fig 2) shows the latest snapshot, a diagram with the number of stacks acquired since the opening of the exhibition and a histogram with the distribution of button presses on the current day. Our website was implemented such that it updated itself automatically and could be displayed on public screens, e.g. in the lobby of our institute. To limit data transfer, only the small table file was fetched at a fixed interval of 5 seconds, using JavaScript, AJAX and jQuery (https://jquery.com), and checked for modification. The diagrams were updated and the latest snapshot image was downloaded and displayed on the page only if necessary. 10.1371/journal.pone.0161671.g002Fig 2 Screenshot of the eduSPIM website, http://www.eduspim.org. The page was divided into three panels: the main panel containing a snapshot of the latest rendering displayed on the microscope computer (left), a diagram showing the number of stacks acquired since the exhibition was opened (upper right) and a histogram indicating how often each button was pressed on the current day (lower right). All panels were dynamically updated using JavaScript. The data shown here were acquired in stack mode. The fluorescence signal was rendered using a combination of volume ray casting and depth colour coding. An intuitive interface for users The user interface consists of both the controls for the user to interact with the system, as well as a visual representation of the instrument’s results. Generally, the functionality offered to the user and the input controls should be kept minimal and concentrate on the key aspects. If the instrument is located at a museum, this is essential to attract the visitors’ attention and to convey its educational message. For a scientific instrument operated under extreme conditions, it is important to be intuitively usable. In both cases, controls such as buttons must be robust, under certain conditions they may also need to be water- or fireproof, or they may need to be operable with gloves. Although input controls can be implemented in software, hardware solutions are preferable to avoid the need of a keyboard or mouse and thereby prevent users from interacting with the operating system. To connect the hardware to the PC and read its state from within the control software, some kind of input/output (I/O) interface is required, e.g. a data acquisition device. Sometimes the signal sent by the hardware needs additional filtering, e.g. to avoid flickering when pressing a button, in which case we recommend to use a microcontroller board. For eduSPIM, we designed a control panel with seven buttons (Fig 3): two for positioning the sample along the vertical axis, two for adjusting the imaging plane from dorsal to ventral, and one button each for recording a stack, for switching the laser manually on and for displaying an information screen. We connected the buttons to the computer via an Arduino board (https://www.arduino.cc, S1 Fig). Custom software on the Arduino took care of unwanted flickering. 10.1371/journal.pone.0161671.g003Fig 3 User interface. During normal operation, the user interface is clean and simple. The computer screen displays only a rendering of the acquired microscope data (centre), together with a small 3D model of the zebrafish sample indicating the current imaging plane (upper right). The data shown here were acquired during preview mode. The transmission image was rendered as an opaque surface in grey, overlaid with the fluorescence signal from the vasculature in green. A rendering of the data acquired in stack mode is shown in Fig 2. A plate with seven buttons (lower right), connected to the control computer via an Arduino board, lets the visitors move the sample forward and backward, record a stack, manually switch the laser on or display a screen with additional information. For maintenance, e.g. when the sample is exchanged, we implemented an additional panel (left) to adjust numerous settings conveniently. This panel is hidden during normal operation and only displayed via a keyboard shortcut by an administrator. Visual representation of the instrument’s results depend very much on the kind of data that is obtained. In general, the representation needs to be as simple as possible to be easily comprehensible. In the laboratory, data acquired by scientific instruments is usually processed post-acquisition. Processing and analysis steps are tailored to an experiment and aim to confirm or disprove a scientific hypothesis, typically in a quantitative manner. In contrast, a standalone instrument, either in an exhibition or in the field, has a fixed purpose with a given processing and visualization pipeline. If the instrument has an educational purpose, an illustrative representation of the data that best demonstrates its functionality is essential. If it is an imaging device such as a microscope, the acquired images need to be rendered in a way that the imaged structure can easily be recognized. This is particularly important in case of three-dimensional imaging. At the same time, processing and rendering must be efficient to be applicable in real-time for immediate display. For eduSPIM we implemented a preview mode that was active while the user moves the sample manually through the light sheet, and a stack mode that captures a three-dimensional z-stack at the current position. In both modes, the display was continuously updated while images were obtained from the cameras. To support the visitor in navigating through the imaged sample, a small 3D model of it was displayed next to the rendering, indicating the light sheet position and the current imaging plane (Fig 3). In preview mode, the transmission image was rendered in grey, overlaid with the fluorescence signal using a green lookup table. Scaling the images corresponding to the current axial imaging position and placing them into a coordinate system provided a 3D impression. In stack mode, we combined basic volume ray casting with depth colour coding for real-time 3D rendering. We pre-calculated lookup tables with a different colour for each plane. As in preview mode, the display was continuously updated as images were acquired. Each pixel in an image was given an RGB value according to the corresponding lookup table and a transparency value proportional to the square of its intensity. Images were then scaled depending on their z position and overlaid from back to front, taking into account their transparency values (Fig 2). This was accomplished in real-time without the need for sophisticated graphics hardware using Java2D drawing functions. An augmented interface for maintainers For a standalone instrument to be intuitively usable, the functionality exposed to users is typically limited. Normally, it is however necessary to adjust several settings from time to time to keep the instrument working optimally. The software needs to provide an augmented interface to maintainers to make these changes. This interface should not be visible to normal users. We implemented a graphical maintenance panel (Fig 3) that was hidden during normal usage but could be made visible by connecting a keyboard and pressing a shortcut. Furthermore, we integrated a panel for entering and executing BeanShell (http://www.beanshell.org) commands. BeanShell is a scripting language that can be embedded into Java applications to execute Java code dynamically at run-time. It provides a flexible way for adding new functionality without a GUI. The BeanShell interface was also hidden during normal operation. For eduSPIM, we needed to replace the sample regularly, despite an optimized protocol for sample fixation and staining. Upon sample exchange, we adjusted the bounding box of the imaged volume, calibrated the optical elements and fixed acquisition parameters like exposure time and laser power, using the graphical maintenance panel. We used the BeanShell interface to perform non-repetitive tasks, e.g. for time-lapse recording, for recording raw 3D data sets and for automatic acquisition and stitching of multi-tile data sets [4, 5]. Programming environment We have implemented our software in Java and C/C++. Java was used for the graphical user interface and the logic flow. C/C++ was used to implement hardware communication to the cameras and motors. The C/C++ routines were called from Java using the Java Native Interface (JNI). Maven (https://maven.apache.org/) was used to build the Java part of the project, while Microsoft NMAKE was used to build the C/C++ modules, using the Microsoft compiler and the Windows 7 SDK. The source code is kept in a Git repository (https://git-scm.com/) for version control and can be publicly accessed on GitHub (https://github.com/bene51/eduSPIM.git). To run the software, no special hardware is needed, since it automatically searches for attached peripherals and, in case of their absence, falls back to the simulating mode. Discussion The concepts presented in this paper apply whenever a scientific instrument needs to be operated reliably without supervision. Our framework focuses on intuitive usage, robustness against hardware failures and ease of maintenance. Its modular design allows to exchange hardware parts easily. To test and validate the concepts we have implemented them in a software framework to control, monitor and efficiently maintain eduSPIM, a light sheet microscope exhibited in a technical museum in Dresden. At the time of writing, it has been continuously running for over nine months, and almost 20,000 stacks have been acquired since the opening of the exhibition. Although our implementation specifically targets the fully-functional hardware of eduSPIM, it can also run without hardware and imitate normal functionality of a light sheet microscope. It can therefore be used to provide a virtual software-based light sheet microscope on any computer. This can be used to easily demonstrate not only the working principles of light sheet microscopy, but also its application, e.g. to image the vasculature in zebrafish, without the need for any hardware. Such a virtual light sheet microscope on a laptop can be taken anywhere, e.g. to class rooms, conferences, public discussions or it can be used for exhibitions in public locations such as universities or municipal offices, to convey the importance of science for today’s society to the general public and raise the interests of young people in research. The ideas introduced in this paper are not specific to light sheet microscopes, not even to microscopes in general, and can be used for exhibitions or remote operation of similar devices. Beyond the educational mission, our design can as well serve as a blueprint for unsupervised instruments that are operated at remote locations that might be hard to reach or contaminated. Supporting Information S1 Fig Arduino wiring diagram. For eduSPIM, we connected seven buttons to an Arduino board. Pull-down resistors kept the signal low if a button’s state was open (not pressed). When a button was pressed, the input signal at the corresponding pin switched to high. A change in the buttons’ states was recognized by the software running on the Arduino, which notified the microscope control software on the PC through a serial connection. After a button press, consecutive switches were only allowed after 50 ms delay, to avoid bouncing. Additionally, we used the Arduino for controlling the LED that was used for transmission imaging on the microscope. We automatically switched the LED off if no button was pressed for 5 minutes. We adjusted the LED’s brightness using the LED controller and connected it to the Arduino using a transistor (TIP 120) for switching. We provide the software running on the Arduino on github (https://github.com/bene51/EduSPIM). (EPS) Click here for additional data file. We thank the Technische Sammlungen Dresden, particularly Lidia Westermann and Holger Seifert, for providing a great platform for eduSPIM; Florian Frisch from the MPI-CBG media-outreach group for establishing the contact to the museum and all of the members of the Huisken lab for discussions. ==== Refs References 1 Jahr W. , Schmid B. , Weber M. & Huisken J. eduSPIM: Light Sheet Microscopy in the Museum . PLOS One (2016 ). 2 Huisken J. , Swoger J. , Del Bene F. , Wittbrodt J. & Stelzer E. H. K . Optical sectioning deep inside live embryos by selective plane illumination microscopy . Science (New York, N.Y.) 305 , 1007 –9 (2004 ). 10.1126/science.1100035 3 UNESCO. International Year of Light and Light-Based Technologies (2013). Available: http://www.light2015.org. 4 Preibisch S. , Saalfeld S. & Tomancak P. Globally optimal stitching of tiled 3D microscopic image acquisitions . Bioinformatics (Oxford, England) 25 , 1463 –5 (2009 ). 10.1093/bioinformatics/btp184 5 Schindelin J. et al Fiji: an open-source platform for biological-image analysis . Nature Methods 9 , 676 –82 (2012 ). 10.1038/nmeth.2019 22743772

PMC005xxxxxx/PMC5003369.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 10.1371/journal.pone.0161466PONE-D-16-20443Research ArticleBiology and Life SciencesDevelopmental BiologyCell DifferentiationBiology and Life SciencesCell BiologyCell PhysiologyCell MetabolismBiology and Life SciencesDevelopmental BiologyCell DifferentiationOsteoblast DifferentiationResearch and Analysis MethodsBiological CulturesCell CulturesBiology and Life SciencesCell BiologyCellular TypesAnimal CellsConnective Tissue CellsOsteoblastsBiology and Life SciencesAnatomyBiological TissueConnective TissueConnective Tissue CellsOsteoblastsMedicine and Health SciencesAnatomyBiological TissueConnective TissueConnective Tissue CellsOsteoblastsResearch and Analysis MethodsMicroscopyLight MicroscopyFluorescence MicroscopyBiology and Life SciencesCell BiologyCell ProcessesCell ProliferationBiology and Life SciencesGeneticsGene ExpressionControlling Growth and Osteogenic Differentiation of Osteoblasts on Microgrooved Polystyrene Surfaces Osteoblasts on Microgrooves SurfacesSun Lanying 123Pereira Daniel 34Wang Qibao 2Barata David Baião 34Truckenmüller Roman 34Li Zhaoyuan 2Xu Xin 1*Habibovic Pamela 34*1 Shandong Provincial Key Laboratory of Oral Tissue Regeneration, School of Stomatology, Shandong University, Jinan, Shandong Province, China2 Oral Implantology Center, Stomatology Hospital of Jinan, Jinan, Shandong Province, China3 Department of Tissue Regeneration, MIRA Institute for Biomedical Technology and Technical Medicine, University of Twente, Enschede, Overijssel, The Netherlands4 MERLN Institute for Technology-Inspired Regenerative Medicine, Maastricht University, Maastricht, Limburg, The NetherlandsEngler Adam J. EditorUniversity of California, San Diego, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceived and designed the experiments: PH XX RT LS DP. Performed the experiments: LS DP. Analyzed the data: LS PH DP QW DBB ZL XX. Contributed reagents/materials/analysis tools: PH RT QW DBB. Wrote the paper: LS PH RT DP ZL XX. * E-mail: p.habibovic@maastrichtuniversity.nl (PH); Xinxu@sdu.edu.cn (XX)29 8 2016 2016 11 8 e016146620 5 2016 5 8 2016 © 2016 Sun et al2016Sun et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Surface topography is increasingly being recognized as an important factor to control the response of cells and tissues to biomaterials. In the current study, the aim was to obtain deeper understanding of the effect of microgrooves on shape and orientation of osteoblast-like cells and to relate this effect to their proliferation and osteogenic differentiation. To this end, two microgrooved polystyrene (PS) substrates, differing in the width of the grooves (about 2 μm and 4 μm) and distance between individual grooves (about 6 μm and 11 μm, respectively) were fabricated using a combination of photolithography and hot embossing. MG-63 human osteosarcoma cells were cultured on these microgrooved surfaces, with unpatterned hot-embossed PS substrate as a control. Scanning electron- and fluorescence microscopy analyses showed that on patterned surfaces, the cells aligned along the microgrooves. The cells cultured on 4 μm-grooves / 11 μm-ridges surface showed a more pronounced alignment and a somewhat smaller cell area and cell perimeter as compared to cells cultured on surface with 2 μm-grooves / 6 μm-ridges or unpatterned PS. PrestoBlue analysis and quantification of DNA amounts suggested that microgrooves used in this experiment did not have a strong effect on cell metabolic activity or proliferation. However, cell differentiation towards the osteogenic lineage was significantly enhanced when MG-63 cells were cultured on the 2/6 substrate, as compared to the 4/11 substrate or unpatterned PS. This effect on osteogenic differentiation may be related to differences in cell spreading between the substrates. the Open Foundation of Shandong Provincial Key Laboratory of Oral Tissue RegenerationSDKQ201504Sun Lanying the Sci-Tech Development Program of Jinan City201221055Sun Lanying the Medical Sci-Tech Development Program of Jinan City2008-30Sun Lanying Dutch Province of Limburg LINKHabibovic Pamela NIRM: Model systems and novel technology for skeletal regenerationBarata David Baião This work was supported by the Open Foundation of Shandong Provincial Key Laboratory of Oral Tissue Regeneration (SDKQ201504, LS), the Sci-Tech Development Program of Jinan City (201221055, LS) and the Medical Sci-Tech Development Program of Jinan City (No. 2008-30, LS). DBB gratefully acknowledges the financial support of the NIRM (Netherlands Institute of Regenerative Medicine, PH). This research has been in part made possible with the support of the Dutch Province of Limburg (PH). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper.Data Availability All relevant data are within the paper. ==== Body Introduction Establishing successful integration of a biomedical implant into the host bone tissue is of prime importance in orthopedics and dental surgery [1–4]. Efforts invested in optimizing the interface between an implant and its biological environment are growing, as a result of a widespread use of, for example, dental implants. Surface-structural features of biomaterials in the form of roughness and topography, are, in addition to surface-chemical properties, increasingly being recognized as crucial factor to control the response of cells and tissues to biomaterials [5–10]. Surface topography has been shown important for the early events of attachment and formation of focal adhesions, activating mechanotransduction events, which eventually may be determinant for cell fate and consequent tissue formation. Among various types of designed topographies, microsized grooved surfaces have been extensively studied for their effects on cell alignment because they can be relatively easily produced using a variety of microfabrication techniques [4, 8, 11–16]. Regarding the behavior of osteogenic cells on grooved surfaces, it has been demonstrated that in vitro, they strongly orient in the direction of grooves, unlike on flat surfaces, where a random orientation is generally observed [5, 16–19]. It has also been demonstrated that microgrooves with widths comparable to the cell size induce remarkable cell guidance, while the effect of the grooves with widths appreciably larger than the cells is weak [20]. In a study of SaOs-2 osteoblastic cells on microgrooves with widths ranging from 4 to 38 μm, it was shown that the narrower grooves (4 to 16 μm), in the range of 0.5–2 fold of cell size, are more effective in guiding the cell orientation [17]. A recent report showed that microgrooves with a width ranging from 2 to 12 μm exhibits great contact guidance effects on the shape and orientation of rat bone marrow cells and fibroblasts [14, 21]. Another study showed that microgrooves with a width between 1 and 10 μm can change rat dermal fibroblasts cell morphology and induce cell guidance [11, 22]. In again another study, it was revealed that the degree of cell guidance and alignment is greatest on narrow grooves (2 and 4 μm), a width that is below the MG-63 cell size [16]. However, it was also demonstrated that microgrooves with the width considerably smaller than the cell size, have a less pronounced effects on cell shape. For example, contact guidance was not observed when fibroblast cells were cultured on grooves smaller than 100 nm [23]. While contact guidance is a generally accepted effect of microgrooved surfaces on cell orientation and morphology, there is less consistency in results regarding the effect of such micropatterns on osteoblast proliferation and osteogenic differentiation [12, 13, 24–27]. For example, in the study by Matsuzaka et al. [12], no significant difference in the amount of mineralized extracellular matrix, or alkaline phosphatase (ALP) expression of rat bone marrow cells cultured on polystyrene (PS) were observed as a result of groove depth or width, whereas when the cells were cultured on poly(lactic acid) substrates with a depth of 1 μm and a width of 1 μm or 2 μm, more mineralized extracellular matrix formation was observed than on grooves with larger sizes. Further study by the same group [13] showed no effect on proliferation of rat bone marrow cells as a result of presence of microgrooves on either PS or poly(lactic acid). A higher calcium content was observed on microgrooved poly(lactic acid) as compared to PS, without a significant effect of the microgroove dimensions. Another study, by Kenar et al. [24] showed a positive effect of a substrate made of a blend of poly(3-hydroxybutyrate-co-3-hydroxyvalerate) (PHBV) and poly(l/d,l-lactic acid) (P(l/dl)LA) with 27 μm wide grooves, on ALP expression and calcium deposition by rat bone marrow derived osteoblasts, as compared to unpatterned controls. A higher ALP expression by SaOs-2 cells was observed on microgrooved calcium phosphate substrate as compared to flat silicon substrate or tissue culture plastic, although no direct comparison with unpatterned calcium phosphate was made. There was no effect on cell proliferation of micropatterns [25]. Yang et al. [26] showed a significantly higher proliferation of human fetal osteoblasts on microgrooved calcium phosphate ceramic as compared to the unpatterned one, but no consistent results regarding the effect of the groove dimensions were observed. Finally, in the study by Jiang et al. [27], a negative effect of microgrooved titania was observed on both proliferation and ALP expression of MC3T3-E1 cells, as compared to the unpatterned substrate, whereas no differences were observed between 12- or 40 μm wide grooves. In the current study, the focus was on the properties of microgrooved surfaces, i.e. their size and periodicity. To investigate these properties, hot embossing based on photolithographically patterned micromoulds was used as a means to provide surfaces of PS, with distinct microscale features. As one of the first microfabrication techniques applied to the field of biology and biomedicine, photolithography has been widely used to generate microstructures such as grooves and wells in inorganic materials such as silicon and silicon oxide [28, 29]. This enabled the creation of a more controlled microenvironment and further study of the influence of surface topography on cell behaviour [30, 31]. PS was selected for this study as it is a widely used cell culture substrate, that itself does not promote or inhibit (pre)-osteoblast differentiation towards the osteogenic lineage. Furthermore, this thermoplastic polymer is amenable for precise patterning on the micron- or submicron scale [15], allowing detailed studies into the effect of surface topographies on cell behaviour. Here, two types of microgrooved PS substrates, differing in the width of the grooves (2 or 4 μm) and the distance between adjacent grooves (6 or 11 μm, respectively) were used, but both having subcellular dimensions to profit from the previously demonstrated contact guidance effect. MG-63 human osteosarcoma cells were cultured on these surfaces, and their attachment, metabolic activity, proliferation and osteogenic differentiation were assessed. Materials and Methods Micropatterning of PS surfaces Standard photolithography followed by reactive ion etching was used to produce a silicon wafer carrying two types of periodical patterns of parallel, straight microgrooves. Upon production, the dimensions of topographical features were measured using a white light interferometer (ContourGT-I, Bruker). The patterns were then embossed into PS sheets with a thickness of 50 μm (Goodfellow Ltd.) by applying a pressure of 50 bar at 120°C for 2 min in a nano imprint lithography system (EITRE® 6, Obducat). A PS substrate that was hot-embossed with a flat silicon wafer served as a control. The accuracy of pattern transfer was evaluated by an environmental scanning electron microscope (SEM; XL30, ESEM-FEG, Philips) in the secondary electron mode and quantified again using white light interferometry. For cell culture, substrates with a diameter of 10 mm were punched from the patterned and flat surfaces, followed by activation by air plasma in a plasma cleaner (PDC-002, Harrick Scientific) for 15 seconds. Prior to cell seeding, patterned and flat substrates were placed in ultra-low attachment 48-well-plates, fixed with o-rings and sterilized with 70% ethanol for 15 min with refreshments every 5 min. After complete ethanol evaporation at RT, the samples were washed twice with sterile phosphate buffered saline (PBS) and incubated in cell culture medium overnight. Cell culture on patterned PS Osteoblast-like MG-63 cells (ATCC® CRL-1427TM), originally derived from osteosarcoma-affected bone, were maintained in proliferation medium comprising α-minimal essential medium (α-MEM; Gibco), 10% fetal bovine serum (Lonza), 100 U ml-1 penicillin (Gibco), 100 μg ml-1 streptomycin (Gibco), 0.2 mM ascorbic acid (Sigma-Aldrich), 2 mM L-glutamine (Gibco), and 1 ng ml-1 basic fibroblast growth factor (bFGF; Instruchemie). Cells were expanded at 37°C in a humidified atmosphere with 5% CO2. Medium was refreshed twice a week. Upon reaching 80% confluence, cells were trypsinized and seeded on the PS substrates, and incubated in 1 mL medium at 37°C in a humidified atmosphere with 5% CO2. For cell attachment and morphology analysis, 2500 cells were seeded on each sample in basic cell culture medium (BM; proliferation medium without bFGF). For the assessment of metabolic activity, proliferation and osteogenic differentiation, 4000 MG-63 cells were seeded on each sample and incubated in either BM or osteogenic medium (OM; BM supplemented with 10 nM dexamethasone). Both media were refreshed every 2 to 3 days. SEM, fluorescence microscopy and CellProfiler analysis To assess cell shape on different substrates by SEM and fluorescence microscopy, the cells were cultured for 24 hours in BM. This time point, at which a sufficient number of cells was attached on the surface to allow analysis, but the cells were not confluent yet, was selected based on the previous work, which showed that the effect of surface microfeatures on cell shape occurred early, and the maintenance of this effect was dependent on the properties of the substrate [32, 33]. After 24-hour culture in BM, the cells were washed with PBS, and fixed with 10% formalin (Sigma) for 30 min. For fluorescence microscopy analysis, the samples were permeabilized with 0.1% Triton-X 100 for 5 min and blocked with 1% bovine serum albumin (BSA) in PBS. To stain cell cytoskeleton, Alexa Fluor 488 (1:60 dilution in 1% BSA in PBS; Invitrogen) was added to the samples and incubated for 45 min. Then, DAPI (1:100 dilution in 1% BSA in PBS; Sigma-Aldrich/Fluka) was added for 20 min to stain cell nuclei. Cell images were acquired using an automated fluorescence microscope (BD Pathway™ 435; BD Biosciences) and then analyzed using the CellProfiler software with built-in modules (Measure Object Area Shape) [34]. 6 different areas of each sample and at least 228 cells were used to determine cell area and perimeter as output parameters to compare the effect of different microgrooved topographies. Samples intended for SEM analysis were rinsed with PBS, dehydrated with a graded series of isopropanol (70, 80, 90, 96 and 100%) for 20–30 min each, and finally completed in hexamethyldisilazane (HMDS; Sigma-Aldrich) 2 times for 15 min. The samples were then air-dried, mounted onto SEM stubs, and sputter-coated with gold. The orientation angle (OA), which is defined as the angle between the long axis (maximum length) of the cell and the direction of the groove [35], was used to evaluate the contact guidance generated by the microgrooves. A cell that perfectly aligns with the groove should has an OA of 0°, representing the strongest contact guidance introduced by a microgrooved substrate, whereas the OA of a randomly spread cell is close to 45°. OA was calculated for about 40 cells for each topography, based on the SEM images taken from 6 to 8 randomly selected areas. Metabolic activity, cell proliferation and ALP activity assay PrestoBlue®, a non-destructive cell viability assay (Life Technologies) containing a growth indicator which is reduced by metabolically active cells to a fluorescent agent, was used to quantitatively analyse cell viability (n = 3) after 7, 14 and 21 days of culture, according to the manufacturer’s protocol. In brief, 1 mL of cell culture medium containing 100 μL of the PrestoBlue reagent was added into each well after washing with PBS. The plate was then incubated at 37°C in a humidified atmosphere with 5% CO2 for 40 min. Fluorescence was measured at 590 nm in a spectrophotometer (Victor 3, Perkin Elmer), after which the culture was continued. To assess proliferation, cells (n = 3) were harvested at 7, 14 and 21 days. Total DNA amount was quantified with CyQuant Cell Proliferation Assay kit (Sigma) as a measure of total cell number, according to the manufacturer’s protocol, and a fluorescence measurement (excitation at 480 nm and emission at 520 nm) using a spectrophotometer (Victor 3, Perkin Elmer). ALP activity was then determined with a CDP-star assay kit (Roche Applied Science), according to the manufacturer’s instructions. The ALP activity was evaluated by measurement of luminescence using a spectrophotometer (Victor 3, Perkin Elmer) and normalized for the DNA content. Osteogenic gene expression by quantitative real-time PCR To evaluate the effect of microgrooves on the expression of a set of osteogenic genes, MG-63 cells (n = 3) were cultured on samples for 7, 14 and 21 days. Total RNA was extracted by using a combination of TRIzol® (Invitrogen) and a Nucleospin® RNA isolation and purification kit (Macherey-Nagel). In brief, 1 mL of TRIzol reagent was added to each well, followed by one freeze/thaw cycle. After mixing with 200 μL chloroform, the samples were centrifuged and the aqueous phase was collected. 350 μL 70% ethanol was added to each sample before loading onto the RNA binding column of the NucleoSpin RNA II isolation kit. Subsequent steps were in accordance with the manufacturer’s instruction. After quantification using a NanoDrop spectrophotometer (Nanodrop Technologies), RNA samples were reverse-transcribed into cDNA with an iScriptcDNA Synthesis kit (BioRad) according to the manufacturer’s instructions. Quantitative real-time PCR was performed for analyzing expression of bone morphogenetic protein-2 (BMP-2), runt-related transcription factor 2 (Runx2), ALP, collagen Type 1 (Col-1) and osteocalcin (OC). The CT values were normalized to the glyceraldehyde 3-phosphate dehydrogenase (GAPDH) housekeeping gene and fold induction was calculated using the comparative ΔCT method. Primer sequences of the selected markers are listed in Table 1. 10.1371/journal.pone.0161466.t001Table 1 Primer sequences of the osteogenic genes, the expression of which was investigated using qPCR analysis. Gene Forward Primer Reverse Primer GAPDH CGCTCTCTGCTCCTCCTGTT CCATGGTGTCTGAGCGATGT BMP-2 CCAAGTAAGTCCAACGAAAG GGTGATGTCCTCGTCTGTA Runx2 ATGGCGGGTAACGATGAAAAT ACGGCGGGGAAGACTGTGC COL-I AGGGCCAAGACGAAGACATC AGATCACGTCATCGCACAACA ALP ACAAGCACTCCCACTTCATC TTCAGCTCGTACTGCATGTC OCN TGAGAGCCTCACACTCCTC CGCCTGGGTCTCTTCACTAC Statistical analysis For cell area, perimeter and OA analysis, one-way ANOVA with Bonferroni post-hoc test was used to evaluate the differences between samples. To analyze the topography effect on cell proliferation, metabolic activity, ALP activity and gene expression, two-way ANOVA with Bonferroni post-hoc tests was used. The significance level was set at p < 0.05. Results Characterization of micropatterns Light interferometry measurements showed that the two patterns of the silicon wafer, used to hot-emboss PS, were different in the width of the grooves and the ridge width, i.e. distance between the grooves (Fig 1A). Pattern A had a groove width of 5.1±0.1 μm and a ridge width of 2.9±0.1, whereas the groove and the ridge width of pattern B were 10.0±0.1 μm and 5.0±0.1 μm, respectively. In both cases, the grooves had the same depth of 4.5 μm. Microgrooved surfaces were successfully hot-embossed on PS substrates, resulting in substrates with groove/ridge width of 2.0±0.1/6.2±0.1 μm (substrate 2/6) and 4.0±0.1/11.2±0.2 μm, (substrate 4/11), respectively (Fig 1). 10.1371/journal.pone.0161466.g001Fig 1 Dimensions of grooves and ridges of silicon wafers and of respective hot-embossed polystyrene films of the narrow (A, 2/6) and wide (B, 4/11) designs measured using white light interferometry (n = 10) (a) and SEM images of 2/6 and 4/11 (scale bar = 10 μm) (b). PS films were successfully hot-embossed using the Si wafer. The width of the grooves (ridges on PS substrate) consistently increased with about 1 μm upon hot embossing. Cell attachment, morphology and orientation on micropatterned PS To investigate the effect of microgrooved topographies on cell attachment and morphology, fluorescence microscopy (Fig 2A–2C) and SEM (Fig 2D–2F) analyses were performed after 24-hour attachment, showing that all surfaces allowed cell attachment and that the cell morphology was dependent on the surface-topographical features. While on the flat, unpatterned PS surface, MG-63 cells were randomly orientated and displayed a spread phenotype with distinct cytoplasmic processes, on the microgrooved surfaces, the cells were aligned in the direction parallel to the grooves with clear elongation of the cytoskeleton. On 2/6, the substrate with narrower grooves and ridges, the cells were predominantly observed on the ridges. A “bridging” effect was occasionally observed, whereby a cell spread over grooves connecting two or more ridges. The cells grown on 4/11, with broader grooves and ridges, appeared more confined to the topographical features. They were predominantly found inside the grooves and on the edges of the ridges, but rarely on top of the ridges. The “groove-bridging” effect was less frequently observed on 4/11 as compared to 2/6. 10.1371/journal.pone.0161466.g002Fig 2 Fluorescent images of DAPI/phalloidin-stained MG-63 cells (a-c) and SEM images (d–f) after 24-hour attachment on 2/6 (a, d), 4/11 (b, e) and flat control (c, f). Both microgrooved surfaces induced alignment of the cells in the direction of the grooves. While cells on 2/6 were predominantly found on the ridges, bridging over two or more grooves, on 4/11, the cells were predominantly found inside the grooves. Cells on the flat control appeared randomly oriented and spread. The CellProfiler analysis of the parameters cell area and cell perimeter (Fig 3A and 3B) confirmed these qualitative observations. Values for both cell area and cell perimeter were higher when cells were cultured on 2/6 than on 4/11 or the flat control. Cells cultured on the unpatterned substrate showed a larger cell area and cell perimeter as compared to the cells cultured on 4/11. These data confirmed the effect of surface topography on cell morphology. 10.1371/journal.pone.0161466.g003Fig 3 Quantification of cell area (a), cell perimeter (b) and orientation angle (c) of MG-63 cells cultured for 24 hours on 2/6, 4/11 and flat control. Cells cultured on 2/6 showed a significantly larger cell area and cell perimeter as compared to cells cultured on 4/11 or the flat control. Furthermore, cells cultured on 4/11 showed a smaller cell area and cell perimeter as compared to the flat control. Both microgroove topographies strongly enhanced cell orientation as compared to cells cultured on the unpatterned PS. Statistically significant differences are marked with * for p < 0.05, ** for p < 0.01 and ***for p < 0.001. The OA analysis (Fig 3C) was performed after the initial cell attachment to investigate the contact guidance effect of the microgrooves. The cells cultured on the flat substrate had an average OA of 43.29°±25.61°, confirming a random orientation. In contrast, the average OA of the cells seeded on 2/6 and 4/11 was 3.91°±1.03° and 1.91°±0.59°, respectively, suggesting a strong effect of the microgrooves on cell orientation. Cell viability and proliferation on micropatterned PS PrestoBlue analysis of metabolic activity, performed after 7, 14 and 21 days of culture (Fig 4A) showed that all surfaces supported the growth of MG-63 cells. Over the 21-day culture period, the metabolic activity of the cells increased steadily in both BM and OM, for both microgrooved topographies and the flat control. No significant effect of the medium or the topography was observed on cell viability. 10.1371/journal.pone.0161466.g004Fig 4 Metabolic activity (a) and DNA amount (b) of MG-63 cells cultured on 2/6, 4/11 and flat surface in basic or osteogenic medium for up to 21 days. A gradual increase in both metabolic activity and DNA amount was observed for all surfaces and in both media, without significant effects of surface topography. Quantification of DNA amounts (Fig 4B) was in accordance with metabolic activity data. A slow increase in DNA amount was observed on all substrates over the culture period of 21 days in both BM and OM. At 7 days, a slightly lower DNA amount was measured for cells cultured in OM on 2/6 and 4/11, however, no significant differences were observed between patterned and unpatterned PS. The cells were semi-confluent after 7 days and reached full confluence at the later time points. While it is known that confluence of cells may have a negative effect on osteogenic differentiation, the fact that no significant differences in cell proliferation were observed on different materials makes the comparison of osteogenic differentiation among them still possible. These data suggested that microgrooves used in this experiment did not have a strong effect on cell metabolic activity or proliferation. Osteogenic differentiation of MG-63 cells on microgrooved PS Based on the assumption that a potential effect of the initial cell morphology change caused by the surface topography on osteogenic differentiation would be observed later, 7, 14 and 21 days were selected to measure ALP activity and mRNA transcript expression. ALP enzymatic activity (Fig 5) remained at comparable levels throughout the 21-day cell culture period for all samples, with the exception of cells cultured on the unpatterned substrate in OM, the activity of which increased between day 7 and 14. In BM, no significant effect of the surface topography was observed at any time point. In OM, however, a higher ALP activity was observed on the flat PS as compared to 4/11 at day 14 and day 21. Furthermore, cells cultured on 2/6 also showed a higher ALP activity as compared to those cultured on 4/11 at 21 days. 10.1371/journal.pone.0161466.g005Fig 5 ALP activity, normalized to DNA amount of MG-63 cells cultured on 2/6, 4/11 and flat surface in basic or osteogenic medium for up to 21 days. The effect of topography on the ALP activity was mild, with cells cultured in OM on 4/11 showing a significantly lower activity as compared to the flat control at 14 days, and to both the flat control and 2/6 at 21 days. Statistically significant differences are marked with * for p < 0.05 and **for p < 0.01. To further investigate the differentiation of MG-63 cells at mRNA level, the expression of a panel of osteogenic markers, BMP-2, Runx2, ALP, Col-1 and OC was investigated upon 7, 14 and 21 days of culture in BM or OM (Fig 6). 10.1371/journal.pone.0161466.g006Fig 6 Normalized expression of a panel of osteogenic markers at mRNA level of MG-63 cells cultured on 2/6, 4/11 and flat surface in basic or osteogenic medium for up to 21 days. Cells cultured on 2/6 in BM for 21 days showed a higher expression of BMP-2 and Runx2 than the cells cultured on 4/11 or flat substrate. A similar effect was observed for ALP expression after 21 days in OM. Col-1 expression in BM by cells cultured on 2/6 was higher in comparison to the flat control after 14 days and in comparison to 4/11 after 21 days. The only topography effect on OC expression was seen after 7 days, where the cells cultured on 2/6 showed a higher expression as compared to the flat control. Statistically significant differences are marked with * for p < 0.05, **for p < 0.01 and ***for p < 0.001. A temporal increase in the expression of BMP-2, a highly expressed marker in cells involved in bone morphogenesis [36, 37], was observed on all substrates and independent of the medium used. In general, the BMP-2 expression was higher when cells were cultured in BM as compared to OM, with significant differences for 2/6 and 4/11 at 14 days, and for all substrates at 21 days. While in OM, no significant effect of the PS surface topography was observed at any of the time points, in BM at 21 days, the BMP-2 mRNA expression of cells cultured on 2/6 was significantly higher as compared to the other two substrates. The expression of Runx2, an osteoblast-specific transcription factor indicative of the initiation of osteogenesis [38], also increased in time, comparable to the BMP-2 expression. However, in contrast to BMP-2, Runx2 expression was in general higher when cells were cultured in OM as compared to BM, with significant differences for all substrates at 14 days, and for 4/11, and flat control at 21 days. Regarding the effect of surface topography, in BM, the cells cultured on 2/6 for 21 days showed a significantly higher Runx2 expression than those cultured on the flat substrate or 4/11, which was in accordance with the data observed for BMP-2 mRNA expression. No significant topography effects were observed in OM. The expression of ALP, a membrane-associated protein that is expressed during the post-proliferative period of extracellular matrix maturation [39, 40], remained relatively constant in time, which was in accordance with data for ALP enzymatic activity. In general, the ALP mRNA expression was higher when cells were cultured in OM as compared to BM, with the significant difference reached for 2/6 after 21 days. While no topography effect was observed in BM at any of the time points, at day 21 in OM, the cells cultured on 2/6 showed a higher ALP mRNA expression than cells cultured on 4/11, which was in line with the data on enzymatic activity. The expression of Col-1, a collagenous protein that is expressed during the initial period of proliferation and matrix synthesis [39], was in general low, and no temporal changes were observed. A somewhat higher expression was observed when cells were cultured in BM as compared to OM, independent of the time point or topography. At 14 days, cells cultured on 2/6 and 4/11 in OM showed a downregulation of Col-1 expression as compared to cells cultured in BM. The same effect was observed for flat PS and 2/6 after 21 days of culture. Concerning the topography effect, in BM, cells cultured on 2/6 for 14 days showed a higher expression as compared to cells cultured on the flat substrate, whereas at 21 days, both 2/6 and the flat substrate showed a higher expression than the culture on 4/11. No topography effects were observed when cells were cultured in OM. Finally, the expression of OC, a vitamin K- and D-dependent protein that is expressed during the mineralization stage [39, 40], remained low throughout the culture period, decreasing with time, in particular in cells cultured in BM. While at the earlier time points, the effect of the medium was only observed at 7 days for 2/6, after 21 days of culture, a significantly higher OC mRNA expression was observed in OM as compared to BM, on all substrates. The only topography effect was observed in BM after 7 days, with a higher OC mRNA expression on 2/6 as compared to the flat substrate. The analysis of the osteogenic differentiation showed that cells cultured on 2/6, the substrate with narrow grooves and ridges, in general showed the highest expression of the osteogenic markers. This effect was seen in both BM and OM, dependent on the marker analyzed. Discussion The aim of the current study was to obtain deeper understanding of the effects of microgrooves on shape and orientation of osteoblast-like cells and to relate this effect to metabolic activity, proliferation and osteogenic differentiation of the cells. In the past 10 years, a number of studies have investigated the effect of microgrooved surfaces of polymers and metals on cell behavior, with emphasis on the effect on cell shape and orientation [4, 8, 11–16]. The size of the grooves and their periodicity have been suggested to play an important role in the effect of the microtopographies on cell shape [11, 14, 16, 17, 20, 41], which is plausibly related to the relationship between the groove/ridge size and the cell size. Based on the earlier studies [16, 17, 20], which showed remarkable cell guidance on surfaces having microgrooves with dimensions comparable to the cell size, here we have selected microgrooved patterns with dimensions in the range 2 to 11 μm. Indeed, qualitative analysis showed that MG-63 cells exhibited an elongated morphology aligned in the direction parallel to the grooves, which was in contrast to the unpatterned flat substrate where the cells showed a more spread, circular morphology without clear orientation. Differences were also observed between microgrooved surfaces of different size. While on 2/6, the surface with narrower grooves and ridges, cells were mainly observed on the ridges, with a bridging effect over the grooves, on the 4 /11 surfaces, the cells were predominantly found inside the grooves. The size of the wider grooves appeared sufficient to “host” the cells, and the fact that they were constrained inside the grooves resulted in a more pronounced orientation, as compared to the topography where cells were found on the ridges. While an obvious effect of the microgrooved topography was observed on shape and orientation of MG-63 cells, there was no influence on cell proliferation or metabolic activity. Regarding the osteogenic differentiation, ALP enzymatic activity was mildly affected by the surface topography, with 2/6, the substrate with narrow grooves and ridges, showing higher values after 21 days of culture in OM that contained dexamethasone as a biological stimulator of osteogenic differentiation. A similar mild effect was observed on the expression of ALP at mRNA level. Interestingly, the strongest effect of topography was observed when cells were cultured in BM, i.e. medium without biological stimulators of osteogenesis. Cells cultured on 2/6, with narrow grooves/ridges, showed a higher expression of BMP-2 and Runx2, and to a lesser extent of Col-1 and OC. No such effect was observed in cells cultured on 4/11, the substrate with wider surface microfeatures. Taken together, these data show that there was an effect of surface topography on osteogenic differentiation, although this effect was not strong for all markers tested, which suggests that further optimization of topographical features is needed. Earlier studies on the microgroove effects on osteoblasts proliferation, metabolic activity and osteogenic differentiation have shown varying results. For example, it has been reported that microgrooved poly(lactic acid) surfaces do not affect osteoblast proliferation while the grooves with a width of 1 and 2 μm had a positive effect on osteogenic differentiation and bone formation [12, 13]. These results are in accordance with what was observed in the current study. The 2/6 substrate favoured osteogenic differentiation, as compared to 4/11, the substrate with wider grooves and ridges, and the flat substrate. This observation indeed suggests that the effect of microgrooves on osteogenic differentiation was dependent on the size of grooves/ridges and their periodicity. On the other hand, it has been shown that microgrooved titania limits the osteoblast proliferation and is even detrimental to their differentiation towards the osteogenic lineage at short culture periods [27]. It should be mentioned that the grooves in this study were wider than what we used here, namely 12 or 40 μm, partly explaining differences observed between the two studies. It is nevertheless also plausible that, besides, the surface topography, surface chemistry plays an important role too. This is also in accordance with studies in which the effect of microgrooves was compared between PS and poly(lactic acid) [12, 13]. A significant effect was observed for the type of medium on the osteogenic differentiation. Although no consensus exists regarding the best stimulator of osteogenic differentiation of MG-63 cells, some work has been published on the use of 1,25(OH)2D3 [42]. Here, we have selected dexamethasone, based on its earlier proven stimulatory effect on ALP activity in MG-63 cells [43, 44] although there were also studies in which no effect of ALP or OC secretion was observed [45]. Furthermore, we intended to compare this data to our previous work with hMSCs, where dexamethasone is a know stimulator of osteogenic differentiation. Our results indeed showed a positive effect of the osteogenic medium on the ALP activity, on flat controls only, whereas the effect on ALP mRNA expression was not as obvious. Furthermore, the expression of Runx2 and OC was enhanced both on microgrooved substrates and the control. The fact that BMP-2 was downregulated in presence of dexamethasone is possibly related to the fact that dexamethasone may decrease intracellular calcium levels [45], resulting in lower expression of BMP-2, which is a calcium-responsive gene. This is also in accordance with our previous work showing lower BMP-2 mRNA expression in osteogenic medium, when cultured on calcium phosphates (e.g. [46]). Regarding the correlation between the effect of the microgrooved topography on cell shape and cell behavior in terms of osteogenic differentiation, the results of this study suggested that the cells with larger area and perimeter exhibited a more pronounced osteogenic differentiation. This corroborates with the results from previous studies on the relationship between cell spreading on fibronectin islands and osteogenic differentiation [47, 48]. It has been shown that cells cultured on larger islands of fibronectin commit to osteogenic lineage, whereas those cultured on smaller islands tend to differentiation towards the adipogenic lineage [48]. This effect may be related to enhanced contractility with increased cell spreading, that in turn promotes osteogenic differentiation [49]. Regarding cell orientation, the wider microgrooves that accommodated cell spreading and had a strongest effect on OA did not appear to positively affect the osteogenic differentiation of MG-63 cells, which is in agreement with previous results [27]. Previous studies mainly focused on the effect of microgrooved surfaces either on cell shape, orientation or on proliferation and osteogenic differentiation by detecting ALP activity or mineralized extracellular matrix production. Here, we elucidated on the effect of microgrooves/microridges on cell behaviour comprehensively from cell shape and orientation, through metabolic activity and growth, to ALP activity and expression of a panel of osteogenic markers at gene level. These data could be useful as input for developing PS-based cell culture platforms that are microstructured in such a way that they can directly affect cell behavior, in absence of biological stimulators. Finally, it should be emphasized that this study, like many others focusing on surface topography, was performed in a 2D environment, which is not a natural microenvironment for the cell. Predictive value of such a study is therefore limited for the cell behavior in 3D environment [50–53]. The challenge therefore lies is developing methods that would allow micropatterning of functional, 3D biomaterials that can be applied clinically. Conclusions The results of this study have demonstrated the effect of microgrooved PS surfaces on the morphology, metabolic activity, proliferation and osteogenic differentiation of MG-63 osteoblast-like cells. The cells grown on PS surfaces with 4 μm-grooves/11 μm-ridges showed a more pronounced cell alignment and a somewhat lower cell areas and cell perimeters as compared to cells cultured on 2 μm-groove/6 μm-ridge surfaces or unpatterned PS. While no effects were observed on metabolic activity or proliferation of cells, their differentiation towards the osteogenic lineage was enhanced on the substrate with the narrower micropattern. This positive effect may be related to a more pronounced cell spreading. This study will contribute to the existing knowledge on employing surface microtopography to control cell and tissue response to biomaterials. The authors appreciated Dr. Nick Beijer of Twente University, the Netherlands, for his help with the CellProfiler analysis. ==== Refs References 1 Advincula MC , Rahemtulla FG , Advincula RC , Ada ET , Lemons JE , Bellis SL . Osteoblast adhesion and matrix mineralization on sol-gel-derived titanium oxide . Biomaterials 2006 4 ; 27 (10 ). 2 Feighan JE , Goldberg VM , Davy D , Parr JA , Stevenson S . The influence of surface-blasting on the incorporation of titanium-alloy implants in a rabbit intramedullary model . The Journal of Bone and Joint Surgery American Volume 1995 9 ; 77 (9 ). 3 Suzuki K , Aoki K , Ohya K . Effects of surface roughness of titanium implants on bone remodeling activity of femur in rabbits . Bone 1997 12 ; 21 (6 ). 4 Anselme K , Bigerelle M , Noel B , Iost A , Hardouin P . Effect of grooved titanium substratum on human osteoblastic cell growth . Journal of Biomedical Materials Research 2002 6 15 ; 60 (4 ). 5 Anselme K , Bigerelle M . Statistical demonstration of the relative effect of surface chemistry and roughness on human osteoblast short-term adhesion . Journal of Materials Science Materials in Medicine 2006 5 ; 17 (5 ). 6 Huang HH , Ho CT , Lee TH , Lee TL , Liao KK , Chen FL . Effect of surface roughness of ground titanium on initial cell adhesion . Biomolecular Engineering 2004 11 ; 21 (3–5 ). 7 Liu X , Lim JY , Donahue HJ , Dhurjati R , Mastro AM , Vogler EA . Influence of substratum surface chemistry/energy and topography on the human fetal osteoblastic cell line hFOB 1.19: Phenotypic and genotypic responses observed in vitro . Biomaterials 2007 11 ; 28 (31 ). 8 Li J , Wu M , Chu J , Sochol R , Patel S . Engineering micropatterned surfaces to modulate the function of vascular stem cells . Biochemical and Biophysical Research Communications 2014 2 21 ; 444 (4 ). 9 Hayman MW , Smith KH , Cameron NR , Przyborski SA . Enhanced neurite outgrowth by human neurons grown on solid three-dimensional scaffolds . Biochemical and Biophysical Research Communications 2004 2 6 ; 314 (2 ). 10 Thakar RG , Ho F , Huang NF , Liepmann D , Li S . Regulation of vascular smooth muscle cells by micropatterning . Biochemical and Biophysical Research Communications 2003 8 8 ; 307 (4 ). 11 den Braber ET , de Ruijter JE , Smits HT , Ginsel LA , von Recum AF , Jansen JA . Quantitative analysis of cell proliferation and orientation on substrata with uniform parallel surface micro-grooves . Biomaterials 1996 6 ; 17 (11 ). 12 Matsuzaka K , Walboomers XF , de Ruijter JE , Jansen JA . The effect of poly-L-lactic acid with parallel surface micro groove on osteoblast-like cells in vitro . Biomaterials 1999 7 ; 20 (14 ). 13 Matsuzaka K , Walboomers F , de Ruijter A , Jansen JA . Effect of microgrooved poly-l-lactic (PLA) surfaces on proliferation, cytoskeletal organization, and mineralized matrix formation of rat bone marrow cells . Clinical Oral Implants Research 2000 8 ; 11 (4 ). 14 Grew JC , Ricci JL , Alexander H . Connective-tissue responses to defined biomaterial surfaces. II. Behavior of rat and mouse fibroblasts cultured on microgrooved substrates . Journal of Biomedical Materials Research Part A 2008 5 ; 85 (2 ). 15 Walboomers XF , Croes HJ , Ginsel LA , Jansen JA . Growth behavior of fibroblasts on microgrooved polystyrene . Biomaterials 1998 10 ; 19 (20 ). 16 Ismail FS , Rohanizadeh R , Atwa S , Mason RS , Ruys AJ , Martin PJ , et al The influence of surface chemistry and topography on the contact guidance of MG63 osteoblast cells . Journal of Materials Science Materials in Medicine 2007 5 ; 18 (5 ). 17 Lu X , Leng Y . Quantitative analysis of osteoblast behavior on microgrooved hydroxyapatite and titanium substrata . Journal of Biomedical Materials Research Part A 2003 9 1 ; 66 (3 ). 18 Holthaus MG , Stolle J , Treccani L , Rezwan K . Orientation of human osteoblasts on hydroxyapatite-based microchannels . Acta Biomaterialia 2012 1 ; 8 (1 ). 19 Charest JL , Bryant LE , Garcia AJ , King WP . Hot embossing for micropatterned cell substrates . Biomaterials 2004 8 ; 25 (19 ). 20 Curtis A , Wilkinson C . Topographical control of cells . Biomaterials 1997 12 ; 18 (24 ). 21 Ricci JL , Grew JC , Alexander H . Connective-tissue responses to defined biomaterial surfaces. I. Growth of rat fibroblast and bone marrow cell colonies on microgrooved substrates . Journal of Biomedical Materials Research Part A 2008 5 ; 85 (2 ). 22 den Braber ET , de Ruijter JE , Ginsel LA , von Recum AF , Jansen JA . Quantitative analysis of fibroblast morphology on microgrooved surfaces with various groove and ridge dimensions . Biomaterials 1996 11 ; 17 (21 ). 23 Loesberg WA , te Riet J , van Delft FC , Schon P , Figdor CG , Speller S , et al The threshold at which substrate nanogroove dimensions may influence fibroblast alignment and adhesion . Biomaterials 2007 9 ; 28 (27 ). 24 Kenar H , Kose GT , Hasirci V . Tissue engineering of bone on micropatterned biodegradable polyester films . Biomaterials 2006 2 ; 27 (6 ). 25 Tan J , Saltzman WM . Biomaterials with hierarchically defined micro- and nanoscale structure . Biomaterials 2004 8 ; 25 (17 ). 26 Yang S , Yang C , Lee T , Lui T . Effects of calcium-phosphate topography on osteoblast mechanobiology determined using a cytodetacher . Materials Science and Engineering C 2012 ; 32 (2 ). 22563143 27 Jiang L , Lu X , Leng Y , Qu S , Feng B , Weng J , et al Osteoblast behavior on TiO2 microgrooves prepared by soft-lithography and sol–gel methods . Materials Science and Engineering C 2012 ; 32 (4 ). 28 Madou M . Fundamentals of microfabrication: the science of miniaturization . Boca Raton : CRC Press ; 2002 . 29 Franssila S . Introduction to microfabrication . West Sussex : John Wiley & Sons Ltd ; 2010 . 30 le Digabel J , Ghibaudo M , Trichet L , Richert A , Ladoux B . Microfabricated substrates as a tool to study cell mechanotransduction . Medical and Biological Engineering and Computing 2010 10 ; 48 (10 ). 31 Nikkhah M , Edalat F , Manoucheri S , Khademhosseini A . Engineering microscale topographies to control the cell-substrate interface . Biomaterials 2012 7 ; 33 (21 ). 32 Barata D , Resmini A , Pereira D , Veldhuis SA , van Blitterswijk CA , ten Elshof JE , et al Surface micropatterning with zirconia and calcium phosphate ceramics by micromoulding in capillaries . Journal of Materials Chemistry B 2016 ; 4 (6 ). 33 Raimbault O , Benayoun S , Anselme K , Mauclair C , Bourgaded T , Kietzig AM , et al The effects of femtosecond laser-textured Ti-6Al-4V on wettability and cell response . Materials Science and Engineering C 2016 12 1 ; 69 . 34 Seyedjafari E , Soleimani M , Ghaemi N , Shabani I . Nanohydroxyapatite-coated electrospun poly(l-lactide) nanofibers enhance osteogenic differentiation of stem cells and induce ectopic bone formation . Biomacromolecules 2010 11 8 ; 11 (11 ). 35 Brunette DM . Fibroblasts on micromachined substrata orient hierarchically to grooves of different dimensions . Experimental Cell Research 1986 5 ; 164 (1 ). 36 Honda Y , Anada T , Kamakura S , Nakamura M , Sugawara S , Suzuki O . Elevated extracellular calcium stimulates secretion of bone morphogenetic protein 2 by a macrophage cell line . Biochemical and Biophysical Research Communications 2006 7 7 ; 345 (3 ). 37 Rosen V . BMP2 signaling in bone development and repair . Cytokine and Growth Factor Reviews 2009 Oct-Dec ; 20 (5–6 ). 38 Tsimbouri PM , Murawski K , Hamilton G , Herzyk P , Oreffo RO , Gadegaard N , et al A genomics approach in determining nanotopographical effects on MSC phenotype . Biomaterials 2013 3 ; 34 (9 ). 39 Barrere F , van Blitterswijk CA , de Groot K . Bone regeneration: molecular and cellular interactions with calcium phosphate ceramics . International Journal of Nanomedicine 2006 ; 1 (3 ). 40 Siggelkow H , Rebenstorff K , Kurre W , Niedhart C , Engel I , Schulz H , et al Development of the osteoblast phenotype in primary human osteoblasts in culture: comparison with rat calvarial cells in osteoblast differentiation . Journal of Cellular Biochemistry 1999 10 1 ; 75 (1 ). 41 Zhou F , Yuan L , Huang H , Chen H . Phenomenon of "contact guidance" on the surface with nano-micro-groove-like pattern and cell physiological effects . Chinese Science Bulletin 2009 ; 54 (18 ). 42 Czekanska EM , Stoddart MJ , Richards RG , Hayes JS . In search of an osteoblast cell model for in vitro research . European cells and materials 2012 7 9 ; 24 . 43 Lajeunesse D . Effect of 17 beta-estradiol on the human osteosarcoma cell line MG-63 . Bone and Mineral 1994 1 ; 24 (1 ). 44 Abbadia Z , Amiral J , Trzeciak MC , Delmas PD , Clezardin P . The growth-supportive effect of thrombospondin (TSP1) and the expression of TSP1 by human MG-63 osteoblastic cells are both inhibited by dexamethasone . FEBS Letters 1993 12 6 ; 335 (2 ). 45 Wiontzek M , Matziolis G , Schuchmann S , Gaber T , Krocker D , Duda G , et al Effects of dexamethasone and celecoxib on calcium homeostasis and expression of cyclooxygenase-2 mRNA in MG-63 human osteosarcoma cells . Clinical Experimental Rheumatology 2006 Jul-Aug ; 24 (4 ). 46 Sun L , Danoux CB , Wang Q , Pereira D , Barata D , Zhang J , et al Independent effects of the chemical and microstructural surface properties of polymer/ceramic composites on proliferation and osteogenic differentiation of human MSCs . Acta Biomaterialia 2016 6 16 . 47 Song W , Kawazoe N , Chen G . Dependence of Spreading and Differentiation of Mesenchymal Stem Cells on Micropatterned Surface Area . Journal of Nanomaterials 2011 ; 72 (20 ). 48 McBeath R , Pirone DM , Nelson CM , Bhadriraju K , Chen CS . Cell shape, cytoskeletal tension, and RhoA regulate stem cell lineage commitment . Developmental Cell 2004 4 ; 6 (4 ). 49 Kilian KA , Bugarija B , Lahn BT , Mrksich M . Geometric cues for directing the differentiation of mesenchymal stem cells . Proceedings of the National Academy of Sciences of the United States of America 2010 3 16 ; 107 (11 ). 50 Ergun A , Chung R , Ward D , Valdevit A , Ritter A , Kalyon DM . Unitary bioresorbable cage/core bone graft substitutes for spinal arthrodesis coextruded from polycaprolactone biocomposites . Annals of Biomedical Engineering 2012 5 ; 40 (5 ). 51 Ozkan S , Kalyon DM , Yu X . Functionally graded beta-TCP/PCL nanocomposite scaffolds: in vitro evaluation with human fetal osteoblast cells for bone tissue engineering . Journal of Biomedical Materials Research Part A 2010 3 1 ; 92 (3 ). 52 Erisken C , Kalyon DM , Wang H . A hybrid twin screw extrusion/electrospinning method to process nanoparticle-incorporated electrospun nanofibres . Nanotechnology 2008 4 23 ; 19 (16 ). 53 Chen X , Ergun A , Gevgilili H , Ozkan S , Kalyon DM , Wang H . Shell-core bi-layered scaffolds for engineering of vascularized osteon-like structures . Biomaterials 2013 11 ; 34 (33 ).

PMC005xxxxxx/PMC5003370.txt

==== Front PLoS Negl Trop DisPLoS Negl Trop DisplosplosntdsPLoS Neglected Tropical Diseases1935-27271935-2735Public Library of Science San Francisco, CA USA 2757120110.1371/journal.pntd.0004887PNTD-D-16-00243Research ArticleBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesArtificial Gene Amplification and ExtensionPolymerase Chain ReactionResearch and Analysis MethodsMolecular Biology TechniquesArtificial Gene Amplification and ExtensionPolymerase Chain ReactionMedicine and Health SciencesTropical DiseasesNeglected Tropical DiseasesChikungunya InfectionMedicine and Health SciencesInfectious DiseasesViral DiseasesChikungunya InfectionBiology and life sciencesOrganismsVirusesRNA virusesTogavirusesAlphavirusesChikungunya VirusBiology and Life SciencesMicrobiologyMedical MicrobiologyMicrobial PathogensViral PathogensTogavirusesAlphavirusesChikungunya VirusMedicine and Health SciencesPathology and Laboratory MedicinePathogensMicrobial PathogensViral PathogensTogavirusesAlphavirusesChikungunya VirusBiology and Life SciencesOrganismsVirusesViral PathogensTogavirusesAlphavirusesChikungunya VirusBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesArtificial Gene Amplification and ExtensionPolymerase Chain ReactionReverse Transcriptase-Polymerase Chain ReactionResearch and Analysis MethodsMolecular Biology TechniquesArtificial Gene Amplification and ExtensionPolymerase Chain ReactionReverse Transcriptase-Polymerase Chain ReactionResearch and analysis methodsExtraction techniquesRNA extractionPhysical SciencesPhysicsElectromagnetic RadiationLuminescenceFluorescenceBiology and Life SciencesImmunologyCross ReactivityMedicine and Health SciencesImmunologyCross ReactivityResearch and Analysis MethodsImmunologic TechniquesImmunoassaysEnzyme-Linked ImmunoassaysDevelopment of 2, 7-Diamino-1, 8-Naphthyridine (DANP) Anchored Hairpin Primers for RT-PCR Detection of Chikungunya Virus Infection DANP-Anchored RT-PCR for Detection of CHIKVChen Huixin 1Parimelalagan Mariya 2Takei Fumie 3Hapuarachchi Hapuarachchige Chanditha 4Koay Evelyn Siew-Chuan 5Ng Lee Ching 4Ho Phui San 2Nakatani Kazuhiko 6Chu Justin Jang Hann 1*1 Laboratory of Molecular RNA Virology and Antiviral Strategies, Department of Microbiology and Immunology, Yong Loo Lin School of Medicine, National University Health System, National University of Singapore, Singapore2 School of Applied Science, Republic Polytechnic, Singapore3 The National Defense Medical College, Tokorozawa, Saitama, Japan4 Environmental Health Institute, National Environment Agency, Singapore5 Molecular Diagnosis Centre, Department of Laboratory Medicine and Pathology, Yong Loo Lin School of Medicine, National University Health System, Level 3, National University Hospital Main Building, Singapore6 The Institute of Scientific and Industrial Research, Osaka University, Ibaraki, JapanWilliams Maya EditorNaval Medical Research Center, UNITED STATESThe authors have declared that no competing interests exist. Conceived and designed the experiments: HC MP FT HCH ESCK LCN PSH KN JJHC. Performed the experiments: HC MP. Analyzed the data: HC MP FT HCH ESCK LCN PSH KN JJHC. Contributed reagents/materials/analysis tools: FT HCH LCN ESCK PSH JJHC. Wrote the paper: HC MP HCH PSH JJHC. * E-mail: miccjh@nus.edu.sg; justin_chu@nuhs.edu.sg29 8 2016 8 2016 10 8 e000488716 2 2016 9 7 2016 © 2016 Chen et al2016Chen et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.A molecular diagnostic platform with DANP-anchored hairpin primer was developed and evaluated for the rapid and cost-effective detection of Chikungunya virus (CHIKV) with high sensitivity and specificity. The molecule 2, 7-diamino-1, 8-naphthyridine (DANP) binds to a cytosine-bulge and emits fluorescence at 450 nm when it is excited by 400 nm light. Thus, by measuring the decline in fluorescence emitted from DANP—primer complexes after PCR reaction, we could monitor the PCR progress. By adapting this property of DANP, we have previously developed the first generation DANP-coupled hairpin RT-PCR assay. In the current study, we improved the assay performance by conjugating the DANP molecule covalently onto the hairpin primer to fix the DANP/primer ratio at 1:1; and adjusting the excitation emission wavelength to 365/430 nm to minimize the background signal and a ‘turn-on’ system is achieved. After optimizing the PCR cycle number to 30, we not only shortened the total assay turnaround time to 60 minutes, but also further reduced the background fluorescence. The detection limit of our assay was 0.001 PFU per reaction. The DANP-anchored hairpin primer, targeting nsP2 gene of CHIKV genome, is highly specific to CHIKV, having no cross-reactivity to a panel of other RNA viruses tested. In conclusion, we report here a molecular diagnostic assay that is sensitive, specific, rapid and cost effective for CHIKV detection and can be performed where no real time PCR instrumentation is required. Our results from patient samples indicated 93.62% sensitivity and 100% specificity of this method, ensuring that it can be a useful tool for rapid detection of CHIKV for outbreaks in many parts of the world. Author Summary Chikungunya has reemerged as an important mosquito-borne infection with global health significance. Rapid diagnosis plays an important role in early clinical management of patients due to lack of a vaccine and effective treatment. Laboratory diagnosis is generally accomplished by blood tests to detect virus-specific antibodies but these antibodies are usually developed one week after infection, which misses the window of effective clinical management. On the other hand, although detecting the viral genome can be done in early stage of infection by real-time polymerase chain reaction (PCR) but it is costly to the patients. Here we utilized a fluorescent compound to improve the cost-efficiency of the molecular assay for diagnosis of Chikungunya virus infection. By testing on 77 serum samples, this improved assay has proven to be highly sensitive and specific towards Chikungunya virus. We believe that this research could benefit both clinicians and patients by providing early and accurate diagnosis. Ministry of Defence (SG)DIRP 2012 R182-000-210-232Chu Justin Jang Hann This study is supported by MINDEF Funding DIRP2011, R182-000-210-232. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper.Data Availability All relevant data are within the paper. ==== Body Introduction Chikungunya virus (CHIKV) is an arthropod-borne virus transmitted to humans primarily via the bite of an infected [1] Aedes agypti and Aedes albopictus mosquito. [2, 3] Currently, there are more than 40 countries including Africa, United States, European countries and Southeast Asian countries affected by chikungunya fever. [2] CHIKV is an enveloped positive-sense single stranded RNA virus belonging to Alphavirus genus of Togaviridae family. [4] The genome is approximately 11.8 Kb long, encoding four non-structural proteins (nsP1, nsP2, nsP3, nsP4) and five structural proteins (C, E3, E2, 6K and E1). [5] The clinical symptoms of chikungunya fever are similar to that of dengue fever which is caused by Dengue virus (DENV), an arthropod virus belonging to Flaviviriridae family transmitted by same vectors as CHIKV. [6] This may result in cases of misdiagnosis in places where both viruses co-exist. As there is no vaccine or specific therapeutic agent available for CHIKV infection, early diagnosis of CHIKV is crucial in preventing the collapse of health care system due to unprecedented number of cases usually encountered during CHIKV epidemics. [7] Virus isolation is classified as the gold standard in detection of CHIKV despite being a time-consuming process requiring 1–2 weeks to determine the presence of virus. The limitations associated with virus isolation resulted in the development of serological and molecular diagnostic methods that are rapid and less labour intensive. Enzyme-linked-immunosorbent assay (ELISA) and Immunochromatographic test (ICT) are examples of serological diagnostic assays which detect IgM and/or IgG antibodies that are specific to CHIKV present in patient sera. ELISA and ICT tests are inexpensive and easy to perform as they do not require handling live viruses. A four-fold increase in antibodies by comparing acute phase and convalescent phase serum samples is usually required to confirm CHIKV infection. IgM is detected on an average of two days after infection and persists for several weeks to three months, while IgG is detected in convalescent samples and may persist for years. [8] The outcome of having antibodies present in serum samples after recovery phase may deduce as false-positive detection. Blacksell and co-workers reported that commercially available antibody-based assays are not suitable for acute diagnosis of CHIKV as the results obtained showed ICT and ELISA kits having sensitivity of 1.9–3.9% and 3.9% respectively. [9] An alternative serological method of anti-CHIKV antibody detection has been reported to be used in commercial ELISA kits, but has shown cross-reactivity with other alphaviruses such as Ross River and O’ nyong-yong viruses as they are closely related serologically. [9] Thus, serological methods for CHIKV detection have been inefficient for acute phase diagnosis. [9–11] Recently, molecular diagnosis has been well established for rapid, highly sensitive and specific detection of CHIKV infection during the acute phase. Viral RNA is extracted from serum samples collected 1–7 days post-infection [8] were detected by primers targeting the conserved regions of Chikungunya genome specifically. In comparison, conventional RT-PCR appears to be a less sensitive and relatively more time-consuming process than TaqMan and SYBR Green I-based real-time RT-PCR assays. However, real-time RT-PCR assays require highly sophisticated instruments with yearly maintenance and calibration, restricting the utilization of such assays in places with poor financial and technical resources. [12] Previously, we have reported a novel diagnostic assay for CHIKV detection by adapting hairpin primers and fluorescent molecule, 2, 7-Diamino-1,8-naphthyridine (DANP), into a conventional PCR procedure. [13] In brief, DANP molecule contains a naphthyridine ring which enables it to bind specifically to a cytosine-bulge in a hairpin structure of the PCR primer by hydrogen bonds. [13] The binding of DANP molecule to DNA gives rise to a 400 nm excitation and 450 nm emission property to the bound DANP molecule. As PCR proceeds, the primer is incorporated into double stranded DNA and the hairpin is opened, causing the release of DNAP molecule and thereby decreasing the fluorescence intensity. [13] The utilization of DANP coupled hairpin PCR has also been demonstrated in a single-nucleotide polymorphism study of the cytochrome P450 gene 2C9*3 by Takei and colleagues. [14] However, the binding of DANP molecule to the hairpin-primer is in an equilibrium manner, so that excess DANP molecules must be added to ensure a detectable fluorescence intensity. Therefore, the background signal given off by unbound DANP molecules limits the sensitivity and consistency of the assay. In the present study, DANP molecule was covalently immobilized on the hairpin PCR primers containing C-G base-pairs directly after the C-bulge to quench the fluorescence emission, as shown in Fig 1A. As PCR progresses, the hairpin structure is opened up and the DANP molecule is moved to the outer surface of the double-stranded DNA molecule, away from cytosine-bulge, resulting in an increase in fluorescence emission at 430 nm when it is subjected to UV-light at 365 nm. Increments in fluorescence intensity can be picked up only if the viral RNA template is present in the reaction with negligible background signal as no excess DANP molecules were added to the reaction. The method is highly effective as it uses a conventional RT-PCR protocol followed by measurement of fluorescence signal intensity using a spectrophotometer. The assay is more rapid and cost-effective as compared to real-time PCR methods. The assay was also validated with CHIKV infected patient serum samples and healthy individual serum samples for its sensitivity and specificity. 10.1371/journal.pntd.0004887.g001Fig 1 A. Excitation-emission spectrum of DANP-DNA complexes. 365 nm UV-light is selected for excitation because only basal level of absorbance by DANP-C-bulge complex can be seen while the DANP-dsDNA complex absorbed substantially at this wavelength. Similarly, emission light at 430 nm is measured mainly because it generates the most significant difference when DANP-dsDNA and DANP-C-bulge complexes. B. Illustration of the chemical binding change happens to DANP molecule during PCR procedure. Materials and Methods Viruses CHIKV (GenBank accession No. FJ445502) was isolated from an infected patient during the CHIKV outbreak in Singapore in 2008. The virus was propagated in Aedes albopictus C6/36 cells. Briefly, cells were grown to about 80% confluency in T75 tissue culture flasks. Following removal of the growth media, virus inoculum was added to give a multiplicity of infection (MOI) of 0.1 PFU/cell. Flasks were incubated at 28°C for 1 hours with constant agitation at every 15 min interval. After the incubation, Rosewell Park Memorial Institute (RPMI) 1640 growth medium (Sigma-Aldrich Corp) supplemented with 2% FBS (Hyclone) was added and flasks were maintained at 28°C for about 3–5 days or until cells showed 80% cytophatic effects (CPE). The viral titers were determined by plaque forming assay. [13] Ross River virus (RRV), Sindbis virus (SINV), Kunjin virus (KUNV, MRM 61C strain), West Nile virus (WNV, Sarafend strain), Zika virus (ZIKV, MR 766 strain), DENV-1 (S144 strain), DENV-2 (New Guinea C strain), DENV-3 (Eden 130/05 strain), DENV-4 (S8976 strain), Influenza A virus subtype H1N1, H3N2, Poliovirus type 1 (PV1, Sabin strain), type 2 (PV2, Sabin strain), type 3 (PV3, Sabin strain), Human enterovirus 71 (HEV71, AF316321 strain), Coxsackie B2 virus (CB2), Coxsackie A16 virus (CA16, WHO strain) and Enteric cytopathic human orphan virus 7 (Echo7) were also used to examine the cross-reactivity of this assay. The ZIKV, DENV1-4, PV1-3, HEV71, CB2, CA16 and Echo7 viruses were maintained in the laboratory. The RRV, KUNV and WNV were kindly provided by Professor Mary Mah-Lee Ng, Department of Microbiology, National University of Singapore. The Influenza A viruses were kindly provided by Associate Professor Tan Yee Joo, Department of Microbiology, National University of Singapore. Ethics statement A set of 22 serum samples from CHIKV-infected patients, and 30 from uninfected individuals were collected at the National University Hospital, Singapore, with informed consent, to evaluate the clinical sensitivity and specificity of the DANP-anchored assay. All of the sera were confirmed as febrile illness associated with a positive result from the real-time RT-PCR. [15] This part of the study was performed in accordance with the National University of Singapore Institutional Review Board approved protocol (No. 10–234). Environmental Health Institute (EHI), National Environmental Agency of Singapore kindly provided a set of 25 serum samples from clinically-suspected patients in which the presence of CHIKV was confirmed by a real-time RT-PCR assay. [16] Written informed consent was given for all samples involved in this study. Viral RNA extraction CHIKV RNA was extracted from 140 μL of infected cell culture supernatants (3.6 X 10^7 PFU/mL) and serum samples using the QIAamp viral RNA mini kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. The RNA was eluted in a final volume of 50 μL of nuclease-free water and stored at −80°C until use. PCR Primer designing The full genomes of multiple geographically different strains of CHIKV from recent outbreaks were retrieved from GenBank and aligned using the ClustalX (version 2.1) [17] sequence alignment software. Primers were designed to target the highly-conserved nsP2 regions of CHIKV genome, as shown in Table 1. The primers were designed with hairpin (underlined sequences in Table 1) at the 5’ end to accommodate DANP molecule which is covalently conjugated to the thymine nucleotide (bolded sequences in Table 1) of the primers. 10.1371/journal.pntd.0004887.t001Table 1 DANP-anchored hairpin primer sequences used for detection of CHIKV. Primers Sequence DANP-ANCHORED F 5’ ATCATGCTTTTGCCATGATGACTAATCCGCCCTACCACG 3’ DANP-ANCHORED R 5’ ATCATGCTTTTGCCATGATGCATCCATTCAAGAGCAGCG 3’ RT-PCR conditions RT-PCR reactions were performed in C1000 thermal cycler (Bio-Rad, Hercules, CA). Reactions were optimized with a One Step RT-PCR kit (Biotech Rabbit, Hannover, Germany). Each reaction was performed in 25 μL total reaction mixture containing 12.5 μL of 2x reaction buffer, 0.2 μmol/L of each forward and reverse DANP hairpin primers, 1.25 μL of 20x RT-RI blend (reverse transcriptase and RNAse Inhibitor) and 1 μL of viral RNA. 10 μL from each total reaction volume was set aside while the remaining 15 uL of reaction was subjected to RT-PCR. The thermal profile was optimized as follows; reverse transcription step at 45°C for 20 minutes, activation of Taq polymerase at 95°C for 2 minutes, followed by 30 cycles of PCR cycling steps consisting of 95°C for 10 seconds, 60°C for 10 seconds and 72°C for 15 seconds. Fluorescence intensity measurement and native PAGE analysis In order to determine the fluorescence intensity, 10 μL of each reaction was diluted with 90 μL of nuclease-free water in each well of a white opaque flat-bottom 96-well plate. The fluorescence intensity from each well was scanned by Infinite® 200 PRO microplate reader (Tecan Trading AG, Switzerland) with an excitation filter at 365-nm and an emission filter at 430-nm. A sample positive for CHIKV infection was determined as the increment in fluorescence intensity after PCR was more than 100 arbitrary units (AU) as compared to background fluorescence in pre-PCR reaction mixture. For assay validation, all PCR products were analysed using 8% native polyacrylamide gel electrophoresis (PAGE), followed by ethidium bromide staining for two minutes. Gel images were captured using the GeneSnap software version 7.02 (Syngene, Cambridge, UK). PCR cycle number optimization In order to determine the number of PCR cycles that gives off the most significant increment in fluorescence intensity after PCR, the fluorescence intensity level was measured and compared after every five PCR cycles using CHIKV genomic RNA as positive control and nuclease-free water as negative control (NTC). Determination of limit of detection of the assay CHIKV RNA was extracted from 140 μL of infected cell culture supernatants with viral titre of 3.6 X 10^7 PFU/mL, using the QIAamp viral RNA mini kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. The viral RNA was eluted in a final volume of 50 μL of nuclease-free water and then serial diluted logarithmically until a final concentration of 1 X 10−5 PFU/μL. 1 μL of each of the serial diluted viral RNA samples was subjected to the DANP-anchored RT-PCR assay to determine the limit of detection of the assay. The RNA concentration ranges tested were 1.0 X 101 to 1.0 X 10−5 PFU/reaction (3.6 X 103 to 3.6 X 10−3 PFU/mL). Cross reactivity study RRV, SINV, KUNV, WNV, ZIKV, DENV1-4, Influenza H1N1, H3N2, PV1-3, HEV71, CB2, CA16 and Echo7 were used to examine the cross-reactivity of this assay. Viral RNA was extracted from 140 μL of each of the viruses and eluted in 50 μL of nuclease-free water. The RNA concentration were measured using NanoDrop ND2000 Spectrophotometer (NanoDrop Technologies, Inc., Wilmington, DE, USA) and only samples with at least 50 ng/μL were proceeded to cross reactivity study of the assay. Results PCR cycle optimization In order to determine the number of PCR cycles that gives off the most significant increment in fluorescence intensity after PCR, the fluorescence intensity level was measured and compared after every five PCR cycles using CHIKV genomic RNA as positive control and nuclease-free water as negative control (NTC). As shown in Fig 2, the difference in fluorescence intensity between before PCR and after PCR samples reached the maximum at 30 cycles. Due to the formation of primer dimer and non-specific PCR products in the NTC, the difference in fluorescence intensity began to narrow down after 30 cycles. Therefore, 30 cycles of PCR reaction was used in the rest of the study. 10.1371/journal.pntd.0004887.g002Fig 2 Optimization of number of PCR cycles. DANP-anchored RT-PCR is carried out with and without the presence of CHIKV RNA template and the fluorescence intensity is measured after every 5 PCR cycles from both before and after PCR reactions. The fluorescence intensity starts to increase significantly after 20 cycles when CHIKV RNA is present and reaches saturation after 30 cycles, while that of NTC also starts to increase slowly from 25 to 30 cycles and become more obvious afterwards. As a result, the maximum difference in fluorescence intensity can be achieved after 30 cycles of PCR reaction. Data are shown as means SEM of five experiments. ***P < 0.001, **P < 0.01 by multiple t-test. Primer validation To validate the suitability of hairpin primers and to verify the initial fluorescence intensity level, DANP-anchored hairpin RT-PCR procedure was carried out with and without CHIKV genomic RNA. All PCR products were analyzed by PAGE to determine the assay specificity. As indicated in Fig 3A, the specific PCR product of 296 bps can only be seen when CHIKV RNA is present. There was no significant change in fluorescence intensity between before/after PCR in NTC reactions (Fig 3B). In contrast, an increment of more than 2000 AU of fluorescence intensity was observed after 30 cycles of PCR in the presence of CHIKV RNA. 10.1371/journal.pntd.0004887.g003Fig 3 DANP-anchored RT-PCR primer was evaluate. A: Specific PCR product with expected size (296 bp) is observed only in reaction with CHIKV RNA template after 30 PCR cycles. Lane 1, DNA ladder (GeneRuler Ultra Low Range, Thermal Fisher Scientific, Waltham, Massachusetts, USA); lane 2, CHIKV RNA+ after 30 PCR cycles; lane 3, CHIKV RNA+ before PCR; lane 4, NTC after 30 PCR cycles; and lane 5, NTC before PCR. B: The fluorescence intensity from the PCR reactions with and without CHIKV RNA before and after 30 PCR cycles. Only the reaction with CHIKV RNA shows significant increment in fluorescence intensity (approximately 2000 AU) after 30 cycles of PCR reaction. Data are shown as means SEM of five experiments. ***P < 0.001 by Student’s t-test. Detection limit of the assay The detection limit of the DANP-anchored RT-PCR assay was determined through replicates of reactions, using serial logarithmic dilutions of the control CHIKV genomic RNA. Fig 4 shows the change in fluorescence intensity before and after 30 cycles of PCR reaction. A statistically significant increase of 120 AU was observed in 0.001 PFU per reaction, the lowest level of detection by the assay. 10.1371/journal.pntd.0004887.g004Fig 4 The detection limit of a DANP-anchored RT-PCR assay was determined using 10-fold serial diluted CHIKV genomic RNA. The detection limit for the assay was equivalent to 0.001 PFU per reaction of CHIKV. Data are shown as means SEM of three experiments. ***P < 0.001, **P < 0.01 by Student’s t-test. Cross-reactivity of the assay The cross-reactivity of the assay was determined by using a panel of other RNA viruses. RRV and SINV were used as representative members of Alphavirus; KUNV, WNV, ZIKV and DENV-1, DENV-2, DENV-3, and DENV-4 were used as representative members of Flavivirus in the cross-reactivity study. The remaining RNA viruses included H1N1, H3N2, Polio 1, Polio 2, Polio 3, HEV71, CB2, CA16 and ECHO7. Only CHIKV RNA samples demonstrated positive results, indicating the lack of cross-reactivity with other viruses tested (Fig 5). 10.1371/journal.pntd.0004887.g005Fig 5 Cross-reactivity of DANP-anchored RT-PCR was evaluated with a panel of positive-sense RNA viruses. There is no significant increase in fluorescence intensity found in any one of the viruses tested showing no cross-reactivity of the assay to these viruses. Data are shown as means SEM of three experiments. ***P < 0.001 by Student’s t-test. Sensitivity and specificity of the assay In order to evaluate the sensitivity and specificity of the present assay for clinical diagnosis, 47 serum samples obtained from patients with confirmed CHIKV infection during the acute phase and 30 serum samples from uninfected individuals were tested. The present assay demonstrated high sensitivity by picking up 44 of the 47 CHIKV cases (93.62% sensitivity; 95% CI, 81.44% to 98.37%). None of the 30 serum samples from uninfected individuals was false diagnosed as positive (100% detection specificity; 95% CI, 85.87% to 100%) (Table 2). 10.1371/journal.pntd.0004887.t002Table 2 Performance of DANP-ANCHORED on Serum samples. Type of Serum Sample No. of samples No. of samples diagnosed as positive Sensitivity† Specificity‡ CHIKV 47 44 93.62% (95% CI, 81.44% to 98.37%). 100% (95% CI, 85.87% to 100%) Healthy 30 0 † Number of positive specimens/(number of positive specimens + number of false-negative specimens) X 100%. ‡ Number of negative specimens/(number of negative specimens + number of false-positive specimens) X 100%. Discussion CHIKV has been relatively understudied as it was restricted to Africa and Asia countries. [18] Since 2005, CHIKV started to spread to countries in Indian Ocean and then globally. [19] It is estimated that > 1.5 million people were infected in India during the 2006 outbreak alone, [20] and, currently, Chikungunya fever has been documented in more than 40 countries. [21] The unprecedented worldwide spread of CHIKV was driven by international travel and the A226V mutation on the envelope protein 1, which better adapts the virus to Aedes albopictus. [3] Due to increased globalization and mosquito vectors expand to new areas, early diagnosis of CHIKV is critical in the absence of any licensed antiviral therapy and prophylaxis, especially in developing countries. Currently, the diagnosis of CHIKV largely relies on virus isolation, detection of specific antibody and nucleic acid. Virus isolation in tissue culture is time-consuming and technically complex that is limited in developing countries. Because of extensive cross-reaction between alphaviruses due to common antigens, serological assays often face the difficulty in differentiating commonly occurring alphaviruses. These drawbacks have made molecular assays the method of choice for diagnosis during acute phase of chikungunya fever. Molecular techniques based on the detection of genomic sequences by RT-PCR, nested RT-PCR, and real-time RT-PCR are rapid and sensitive and have replaced virus isolation as the new standard method for the detection of CHIKV in acute-phase serum samples, but the reagents and equipment are too costly for widespread use. In this regard, this DANP-anchored RT-PCR assay reported in this study is advantageous, because of its simplicity, rapidity, and cost-effectiveness. Only a standard conventional PCR procedure, with DANP hairpin primer used, and a fluorescence reading procedure is required without involving of sophisticated instrument or costly reagent. In comparison, previously, we have reported a novel DANP-coupled hairpin RT-PCR for rapid detection of CHIKV in the acute phase serum samples. PCR primers were designed specifically to target nsP2 gene of CHIKV with hairpin tag containing a cytosine-bulge. The DANP molecule binds to the C-bulge in its protonated form (DANPH+) before PCR reaction starts, resulting in fluorescence emission. During PCR amplification, the hairpin primer opens up and releases the DANP molecule resulting in a drop in fluorescence emission giving rise to a ‘turn-off’ system. CHIKV positive samples are determined by comparing the fluorescence intensity recorded before and after PCR process by subjecting the PCR products to UV-light and detecting the emitted fluorescence at 430 nm. Despite it was a rapid, sensitive, specific and cost effective assay; the optimization of DANP concentration due to background signal restricted its usage. In order to overcome the issue, in the present study, we covalently conjugated the DANP molecule onto the hairpin structure of the PCR primer. Therefore, the ratio between DANP molecule and primer is fixed at 1:1 and this standardization simplifies the optimization of the assay. In addition, by changing the reading spectrum from 400 nm excitation and 450 nm emission to 365 nm excitation and 430 nm emission, we managed not only to minimize the background signal but also give rise to a ‘turn-on’ system. In addition, we have also shortened the assay turnaround time from 90 minutes to 60 minutes including fluorescence reading, by cutting down the reverse transcription step duration and optimizing the PCR cycle number from 40 to 30. The detection limit of the assay was 0.001 PFU per reaction that is lower than that of the previous DANP coupled assay [13] and is comparable to real time RT-PCR assays developed by other groups [22, 23]. A side-by-side comparison of our assay with the abTES DEN 5 qPCR I Kit (Cat: 300152) from AIT biotech, a Taqman probe-based multiplex real time RT-PCR for DENV/CHIKV detection. Comparable limit of detection was noted. Despite the fact that the viremia load is usually above 4 log10 during the acute phase of CHIKV infection, [24] low detection limit of the assay enables us to detect CHIKV RNA even during late acute phase when the viral titers start to decline rapidly. More importantly, the present assay is not cross-reactive with a panel of RNA viruses that are co-circulating in endemic regions. Given that CHIKV is commonly misdiagnosed as DENV and vice versa, the outstanding specificity of our assay could benefit both clinicians and patients at the point-of-care by providing accurate diagnosis. ==== Refs References 1 Mohan A , Kiran DH , Manohar IC , Kumar DP . Epidemiology, clinical manifestations, and diagnosis of Chikungunya fever: lessons learned from the re-emerging epidemic . Indian journal of dermatology . 2010 ;55 (1 ):54 –63 . 10.4103/0019-5154.60355 20418981 2 Staples JE , Breiman RF , Powers AM . Chikungunya fever: an epidemiological review of a re-emerging infectious disease . Clinical infectious diseases: an official publication of the Infectious Diseases Society of America . 2009 ;49 (6 ):942 –8 .19663604 3 Tsetsarkin KA , Vanlandingham DL , McGee CE , Higgs S . A single mutation in chikungunya virus affects vector specificity and epidemic potential . PLoS pathogens . 2007 ;3 (12 ):e201 18069894 4 Strauss JH , Strauss EG . The alphaviruses: gene expression, replication, and evolution . Microbiological reviews . 1994 ;58 (3 ):491 –562 . 7968923 5 Khan AH , Morita K , Parquet Md Mdel C , Hasebe F , Mathenge EG , Igarashi A . Complete nucleotide sequence of chikungunya virus and evidence for an internal polyadenylation site . The Journal of general virology . 2002 ;83 (Pt 12 ):3075 –84 . 12466484 6 Ross RW . The Newala epidemic. III. The virus: isolation, pathogenic properties and relationship to the epidemic . The Journal of hygiene . 1956 ;54 (2 ):177 –91 . 13346078 7 Montero A . [Chikungunya fever—A new global threat ]. Medicina clinica . 2015 ;145 (3 ):118 –23 . 25087211 8 Burt FJ , Rolph MS , Rulli NE , Mahalingam S , Heise MT . Chikungunya: a re-emerging virus . Lancet . 2012 ;379 (9816 ):662 –71 . 10.1016/S0140-6736(11)60281-X 22100854 9 Blacksell SD , Tanganuchitcharnchai A , Jarman RG , Gibbons RV , Paris DH , Bailey MS , et al Poor diagnostic accuracy of commercial antibody-based assays for the diagnosis of acute Chikungunya infection . Clinical and vaccine immunology: CVI . 2011 ;18 (10 ):1773 –5 . 10.1128/CVI.05288-11 21865416 10 Prat CM , Flusin O , Panella A , Tenebray B , Lanciotti R , Leparc-Goffart I . Evaluation of commercially available serologic diagnostic tests for chikungunya virus . Emerging infectious diseases . 2014 ;20 (12 ):2129 –32 . 10.3201/eid2012.141269 25418184 11 Rianthavorn P , Wuttirattanakowit N , Prianantathavorn K , Limpaphayom N , Theamboonlers A , Poovorawan Y . Evaluation of a rapid assay for detection of IgM antibodies to chikungunya . The Southeast Asian journal of tropical medicine and public health . 2010 ;41 (1 ):92 –6 . 20578487 12 Sam IC , Chua CL , Chan YF . Chikungunya virus diagnosis in the developing world: a pressing need . Expert review of anti-infective therapy . 2011 ;9 (12 ):1089 –91 . 10.1586/eri.11.132 22114956 13 Chen H , Takei F , Koay ES , Nakatani K , Chu JJ . A novel DANP-coupled hairpin RT-PCR for rapid detection of Chikungunya virus . The Journal of molecular diagnostics: JMD . 2013 ;15 (2 ):227 –33 . 10.1016/j.jmoldx.2012.10.004 23313361 14 Takei F , Igarashi M , Hagihara M , Oka Y , Soya Y , Nakatani K . Secondary-structure-inducible ligand fluorescence coupled with PCR . Angewandte Chemie . 2009 ;48 (42 ):7822 –4 . 10.1002/anie.200902449 19746492 15 Pastorino B , Bessaud M , Grandadam M , Murri S , Tolou HJ , Peyrefitte CN . Development of a TaqMan RT-PCR assay without RNA extraction step for the detection and quantification of African Chikungunya viruses . Journal of virological methods . 2005 ;124 (1–2 ):65 –71 . 15664052 16 Ng LC , Tan LK , Tan CH , Tan SS , Hapuarachchi HC , Pok KY , et al Entomologic and virologic investigation of Chikungunya, Singapore . Emerging infectious diseases . 2009 ;15 (8 ):1243 –9 . 10.3201/eid1508.081486 19751586 17 Larkin MA , Blackshields G , Brown NP , Chenna R , McGettigan PA , McWilliam H , et al Clustal W and Clustal X version 2.0 . Bioinformatics . 2007 ;23 (21 ):2947 –8 . 17846036 18 Cecilia D , Kakade M , Alagarasu K , Patil J , Salunke A , Parashar D , et al Development of a multiplex real-time RT-PCR assay for simultaneous detection of dengue and chikungunya viruses . Archives of virology . 2015 ;160 (1 ):323 –7 . 10.1007/s00705-014-2217-x 25233940 19 Simon F , Savini H , Parola P . Chikungunya: a paradigm of emergence and globalization of vector-borne diseases . The Medical clinics of North America . 2008 ;92 (6 ):1323 –43 , ix 10.1016/j.mcna.2008.07.008 19061754 20 Taubitz W , Cramer JP , Kapaun A , Pfeffer M , Drosten C , Dobler G , et al Chikungunya fever in travelers: clinical presentation and course . Clinical infectious diseases: an official publication of the Infectious Diseases Society of America . 2007 ;45 (1 ):e1 –4 .17554689 21 Schwartz O , Albert ML . Biology and pathogenesis of chikungunya virus . Nature reviews Microbiology . 2010 ;8 (7 ):491 –500 . 10.1038/nrmicro2368 20551973 22 Santhosh SR , Parida MM , Dash PK , Pateriya A , Pattnaik B , Pradhan HK , et al Development and evaluation of SYBR Green I-based one-step real-time RT-PCR assay for detection and quantification of Chikungunya virus . Journal of clinical virology: the official publication of the Pan American Society for Clinical Virology . 2007 ;39 (3 ):188 –93 .17553740 23 Smith DR , Lee JS , Jahrling J , Kulesh DA , Turell MJ , Groebner JL , et al Development of field-based real-time reverse transcription-polymerase chain reaction assays for detection of Chikungunya and O'nyong-nyong viruses in mosquitoes . The American journal of tropical medicine and hygiene . 2009 ;81 (4 ):679 –84 . 10.4269/ajtmh.2009.09-0138 19815886 24 Johansson MA , Powers AM , Pesik N , Cohen NJ , Staples JE . Nowcasting the spread of chikungunya virus in the Americas . PloS one . 2014 ;9 (8 ):e104915 10.1371/journal.pone.0104915 25111394

PMC005xxxxxx/PMC5003371.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757096810.1371/journal.pone.0161919PONE-D-15-34394Research ArticleBiology and Life SciencesAnatomyReproductive SystemGenital AnatomyGonadsMedicine and Health SciencesAnatomyReproductive SystemGenital AnatomyGonadsBiology and Life SciencesBiochemistryLipidsFatty AcidsBiology and Life SciencesOrganismsAnimalsInvertebratesMolluscsBivalvesMusselsBiology and Life SciencesOrganismsPlantsAlgaePhytoplanktonDiatomsBiology and Life SciencesOrganismsAnimalsInvertebratesPlanktonPhytoplanktonDiatomsBiology and Life SciencesOrganismsPlantsAlgaePhytoplanktonBiology and Life SciencesOrganismsAnimalsInvertebratesPlanktonPhytoplanktonBiology and Life SciencesOrganismsProtistsDinoflagellatesPhysical SciencesMathematicsAlgebraLinear AlgebraEigenvaluesBiology and Life SciencesAnatomyBiological TissueMuscle TissueMedicine and Health SciencesAnatomyBiological TissueMuscle TissueSpatio-Temporal Variation in Effects of Upwelling on the Fatty Acid Composition of Benthic Filter Feeders in the Southern Benguela Ecosystem: Not All Upwelling Is Equal Upwelling Effects on Benthic Filter Feedershttp://orcid.org/0000-0002-6144-6650Puccinelli Eleonora 12*McQuaid Christopher David 1Noyon Margaux 131 Department of Zoology and Entomology, Rhodes University, Grahamstown, South Africa2 Department of Oceanography, Marine Research Institute, University of Cape Town, Rondebosch 7701, Cape Town, South Africa3 Marine Research Institute, Department of Biological Sciences, University of Cape Town, Rondebosch 7701, Cape Town, South AfricaMelzner Frank EditorHelmholtz-Zentrum fur Ozeanforschung Kiel, GERMANYCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: EP CDM MN. Data curation: EP. Formal analysis: EP. Funding acquisition: CDM. Investigation: EP. Methodology: EP CDM MN. Project administration: EP CDM MN. Resources: CDM. Supervision: CDM MN. Visualization: EP. Writing – original draft: EP CDM MN. Writing – review & editing: EP CDM MN. * E-mail: eleonorapuccinelli@gmail.com29 8 2016 2016 11 8 e01619195 8 2015 15 8 2016 © 2016 Puccinelli et al2016Puccinelli et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Variability in mesoscale nearshore oceanographic conditions plays an important role in the distribution of primary production and food availability for intertidal consumers. Advection of nutrient rich waters by upwelling usually allows the proliferation of diatoms, later replaced by dinoflagellates. We examined upwelling effects on the fatty acid (FA) signature of a benthic intertidal filter feeder to identify its response to pulsed variability in food availability. The study took place in two contrasting seasons and at two upwelling and two non-upwelling sites interspersed within the southern Benguela upwelling system of South Africa. We investigated the FA composition of the adductor muscles and gonads of the mussel Mytilus galloprovincialis to assess how FA are apportioned to the different tissues and whether this changes between upwelling and non-upwelling conditions. In situ temperature loggers used to identify upwelling conditions at the four sites indicated that such events occurred only at the upwelling centres and only in summer. Tissues differed strongly, with gonads presenting a higher proportion of essential FAs. This could reflect the faster turnover rate of gonad tissue or preferential retention of specific FA for reproductive purposes. FA composition did not vary as a direct function of upwelling, but there were strong dissimilarities among sites. Upwelling influenced mussel diets at one upwelling site while at the other, the expected signature of upwelling was displaced downstream of the core of upwelling. Condition Index (CI) and Gonad Index (GI) differed among sites and were not influenced by upwelling, with GI being comparable among sites. In addition, FA proportions were consistent among sites, indicating similar food quality and quantity over time and under upwelling and non-upwelling conditions. This suggests that the influence of upwelling on the west coast of South Africa is pervasive and diffuse, rather than discrete; while nearshore retention or advection of upwelled water is critical and site-specific so that the effects of upwelling differ even among sites categorised as upwelling centres. This work was supported by funding from the Andrew Mellon Foundation and the South African Research Chairs Initiative of the Department of Science and Technology and the National Research Foundation.McQuaid Christopher David This work was supported by funding from the Andrew Mellon Foundation and the South African Research Chairs Initiative of the Department of Science and Technology and the National Research Foundation awarded by Prof Christopher D. McQuaid. Data AvailabilityAll relevant data are within the paper and supporting information files.Data Availability All relevant data are within the paper and supporting information files. ==== Body Introduction Temporal and spatial variation in mesoscale nearshore oceanographic conditions plays an important role in the distribution of primary production [1,2], resulting in differences in the availability of resources for intertidal consumers [3,4]. These differences in food availability can strongly influence the distribution and metabolism of these organisms [5–7]. One feature that can influence food availability is represented by upwelling events, which bring deep, nutrient-rich waters into coastal ecosystems, promoting nearshore phytoplankton production [8–10]. Coastal upwelling supports many of the world’s most important pelagic fisheries [11] and previous studies have shown the importance of upwelling to benthic communities through the enhancement of primary production and food availability for primary consumers [12–14]. During periods of active upwelling, primary production usually exceeds 1 g carbon m-2 day-2 and can reach up to 10 g carbon m-2 day-2 [12]. The biological consequences associated with these events are critically important for consumers in costal environments. For instance, the reproductive peak of some invertebrates has been shown to coincide with the upwelling season [15,16], enabling planktotrophic larvae to benefit from the phytoplankton rich water. Upwelling events are highly variable in duration, intensity and frequency [17,18]. They are usually stronger and sometimes only occur during the spring and summer months [19–21], leading to marked seasonal shifts in primary production [22]. This variability has important biological consequences for the timing of energy input into the coastal system, with implications that propagate up the food chain, affecting trophic dynamics and thus ecosystem functioning [23,24]. Strong upwelling can reduce recruitment of benthic organisms due to the offshore export of larvae away from favourable settling grounds [25]. Likewise, coastal phytoplankton composition fluctuates seasonally in response to the nutrient input associated with the seasonality of upwelling [26], usually going through a succession of communities dominated first by diatoms and then dinoflagellates [27–29]. These changes at the base of the food web may cause structural changes at different scales of observation, from individuals to the whole food web, by influencing the richness and abundance of intermediate and higher trophic levels [30]. The eastern boundary current system off the west coast of Southern Africa is recognized as one of the world’s largest coastal upwelling systems [31]. It is characterized by the north flowing Benguela Current, of Antarctic origin [32,33], which is associated with strong, wind-driven upwelling events, that occur predominantly in the austral summer [34,19]. The natural seasonality of upwelling events on the west coast of South Africa provides the opportunity to investigate the relationship between temporal variability in nearshore oceanographic conditions and the diets of benthic populations. We used fatty acid (FA) analysis to examine the spatial-temporal scales at which upwelling effects are manifest in the diet of benthic filter feeders, comparing recognised upwelling centres with downstream sites, over time. The FA technique is recognized as a useful tool for understanding the trophic relationship between food source and consumer, providing information on ecosystem dynamics and functioning over a relatively short period of time [35], making it an appropriate tool for this type of study. Our main hypothesis was that filter feeder diets change directly in response to upwelling. Within this context, we investigated several aspects of these dynamics. First, we hypothesised that the FA composition of filter feeders and suspended organic matter (SPM) would be characterized by diatom FA trophic markers (FATM) at upwelling centres during, or shortly after upwelling events, while this signature should decrease downstream of these centres. Similarly, we expected the proportion of dinoflagellate biomarkers to be stronger in specimens and SPM downstream of upwelling sites, as observed by Allan et al. [36], or to increase in the period following an upwelling event, once the nutrient concentration in the water is reduced. Secondly, we expected specimens from upwelling centres to exhibit better body condition than those from non-upwelling areas as overall they are expected to encounter higher food concentrations and potentially accumulate lipids, as energy storage, for food-shortage periods. Lastly, we wanted to identify how FA composition differs between the various tissues of filter feeders in different months and in relation to upwelling. Materials and Methods This study was carried out in accordance with the requirements in the ‘‘permit for the purposes of a scientific investigation or practical experiment in term of section 83 of the Marine Living Resources Act, 1988 (Act no 18 of 1998)”. This permit was approved by the Chief Director of Fisheries Research and Development; Department of Agriculture, Forestry and Fishery, Republic of South Africa (Permit ref. no: RES2012/05 and RES2013/09). Study area The study was conducted along the South African west coast (Fig 1, 34.4°–32.19°S°, 17.52°–18.27°E). The Benguela Current flows south-north along this coast and samples were collected at four sites identified by Xavier et al. as upwelling or non-upwelling centres [37]: sites 1 (Llandudno) and 3 (Paternoster) were categorised a priori as upwelling sites, and sites 2 (Bloubergstrand) and 4 (Elandsbaai) as non-upwelling or downstream sites (Fig 1). Sampling was carried out on four occasions: twice in austral summer (10–11th December 2012 and 8–9th February 2013), during the upwelling season, and twice in winter (12–13th June and 8–9th July 2013), during the non-upwelling season. 10.1371/journal.pone.0161919.g001Fig 1 Map of the study area on the west coast of South Africa showing the sampling sites in upwelling (black) and non-upwelling (white) areas. Upwelling characterization To characterize upwelling frequency and intensity, sea temperature was recorded in situ at each site from December 2012 to July 2013, using four temperature loggers at each site deployed intertidally during low tide. The loggers were composed of an iButton (model DS 1922L Dallas Maxim, CA, USA, Thermochron high resolution, − 40°C to + 85°C with an accuracy of ± 0.0625°C) covered with teflon and glued with two-component epoxy (Alcolin rapid-epoxy) onto perspex plates which were screwed onto the rock surface. All loggers were programmed to measure temperature every 30 min and were replaced every three months due to memory limitations. The iButtons were programmed using the software ColdChain Thermo Dynamics. Maximum tidal range on this coast is approximately two meters and the locations of loggers differed slightly among sites, but all were within the same 20 cm tidal range and deployed next to mussel beds. Hourly data of predicted tidal height at the sampling sites were provided from the website XTide: harmonic tide clock and tide predictor (http://www.flaterco.com/xtide/xtide.html), and were used to separate temperature data for periods of immersion and emersion of the loggers. Upwelling events in this region can be defined as a daily decline in mean temperature (ΔT) of ≥ 5°C [34,38], while a drop in daily mean temperature of 1–4°C was defined as a weak cooling event. The number of successive days after each cooling event in which the temperature remained constant (ΔT ≤ 5°C) from the initial drop was also recorded. The results of the temperature loggers were used to confirm the a priori characterization of sites as either upwelling or non-upwelling. Specifically, we used temperature data from 5 days prior a sampling event to categorize upwelling (48 measurements per day per data logger). Sampling Fatty acids To investigate the influence of upwelling on filter feeder diets and metabolism, we studied the adductor muscles and gonads of the mussel Mytilus galloprovincialis. The adductor muscle was chosen due to its low turnover rates, making it representative of a time-integrated diet [39], while the composition of the gonad provides insight into the reproductive state. On each sampling occasion and at each site, six haphazardly selected individuals were collected, dissected and processed for FA analyses. In addition, to assess if food quality differed among sites or varied among sampling occasions, three replicates of 5 L of seawater were collected from the shore to evaluate the SPM. Water samples were filtered gently (< 5 cm Hg vacuum) onto pre-combusted (450°C) GF/F filters (0.7 μm pore size and 47 mm diameter). The tissue and SPM samples were then flash frozen in liquid nitrogen and transferred to a– 80°C freezer until processing. Condition and gonad indices To understand the possible effects of upwelling on the condition of mussels, the condition index (CI) and the gonad index (GI) of M. galloprovincialis were measured from specimens at each site on each of the four sampling occasions. For the CI, 20 haphazardly selected adult mussels (4–5 cm shell length) were collected and kept frozen at– 20°C until processing. The shells and the soft body tissues of each mussel was dissected and dried at 60°C for 48 h. CI was also calculated for the samples used for the FA analyses (providing six additional individuals per site). For these samples, the soft parts were freeze dried for 24 h as drying at 60°C would degrade the FA. No differences in the CI were recorded between specimens dried with the two different methods (ANOVA, p > 0.05). Consequently all replicates were pooled for the CI analyses (n = 26 per site). The CI was then calculated as a percentage of the dry soft tissue weight over the dry shell weight, following Davenport and Chen [40]: CI=dry soft tissue weightdry shell weightX 100 The GI was measured only for the samples used for the FA analyses (n = 6 per site). The gonads were dissected and freeze dried for 24 h. The GI was calculated as a percentage, by dividing the dry weight of the gonad by the total body dry weight following Williams and Babcock [41]: GI=dry gonad weightdry total body weight X 100 Fatty acid analyses Samples were lyophilized (VirTis BenchTop K) for 24 h, then stored at -80°C until total lipids were extracted and trans-esterified using a modified Indarti et al [42] one-step procedure within 6 months of collection. A known amount of tissue (between 25 and 50 mg of dry weight) for each sample was homogenized into a 4 mL fresh solution of a mixture of methanol, concentrated sulphuric acid and chloroform containing 0.01% of an anti-oxidant, BHT (butylated hydroxytoluene) (1.7/0.3/2.0 v/v/v), and closed under nitrogen. The extraction and transesterification reactions occurred at 100°C for 30 minutes. The FA methyl esters (FAME) formed were then stored at -80°C until Gas Chromatography (GC) analyses. FAME composition of each sample was determined by GC (Agilent Technologies 7890A) equipped with a ZB-Waxplus capillary column (ZB-Waxplus 320 column), with helium as the carrier gas at a flow rate of 1.664 ml min−1. The injector was at a temperature of 250°C. The flame ionization detector was set at 260°C, and the oven was initially set at 70°C. After 1 min, the oven temperature was increased by 40°C min-1 until 170°C and then raised to 250°C at a rate of 2.5°C min−1 and held for 4.5 min. Peaks were integrated using GC ChemStation software (Agilent Technologies, version B.04.02), identified by comparison with retention times of external known standards (37 component fatty acid methyl ester mix Supelco, marine PUFA no. 1 Supelco, menhaden oil PUFA no. 3, bacterial acid methylesters mix Supelco), as well as by mass spectrometry analyses (Agilent Technologies 7000 GC/MS Triple Quad; Agilent Mass Hunter (MS), version B.05.00,) using the NIST library. Each FA was measured as a proportion of the total FA (TFA) composition (% by weight of TFA) and peak areas were corrected according to the FID response to FA chain length [43]. FA are reported using a shorthand notation of A:Bwx, where A indicates the number of carbon atoms, B is the number of double bonds and x indicates the position of the first double bond relative to the terminal methyl group [44]. Data analysis Upwelling characterisation To test the a priori characterization of upwelling or non-upwelling conditions at the four sites, we used temperature data from 5 days prior to each sampling event at each site. Because in the month of December we sampled on the day of deployment of the data loggers, we decided not to include these data in the analyses that tested for upwelling effects, and for that month only to evaluate differences in FA composition, CI and GI of mussels among sites (see details below). To test for differences in temperature among sites over time we ran an analysis of variance (ANOVA) with a mixed model design consisting of the factors: month (three levels, fixed and crossed with the other factors), upwelling (two levels, fixed and crossed with month) and site (two levels, random and nested in upwelling). In the event of significant results, Tukey HSD post hoc tests were performed. The violation of homogeneity of variances was considered to be acceptable because ANOVA is relatively robust to heterogeneous variances for large designs such as the one in this study [45]. Analyses were performed using STATISTICA v12 (StatSoft). Fatty acids To investigate the FA composition of mussels and SPM in relation to upwelling, we used the categorization of upwelling and non-upwelling conditions based on the temperature logger results, which indicated upwelling only at sites 1 and 3 and only in February. For June and July, all sites were considered as representing non-upwelling conditions (see Results and Table A in S1 File). Thus we evaluated the effect of upwelling for the month of February only, using a mixed model design consisting of the factors: upwelling (two levels, fixed and crossed with tissue), site (two levels, random and nested in upwelling), and tissue (two levels, fixed and crossed with all the other factors). To test for differences among sites in June and July, we used a similar design without the factor upwelling, as conditions at all sites were considered to be non-upwelling. For this analysis, the factors were: site (four levels, random and crossed with month and tissue), month (two levels, random and crossed with site and tissue) and tissue (two levels, fixed and crossed with site and month). With the present tools we were not able to assess whether upwelling occurred in December or at which sites, thus we decided to analyse this month separately from the others and we only tested for the factor site (four levels, random). Similarly, the experimental design to test for FA differences in SPM in February comprised the factors: upwelling (two levels, fixed) and site (two levels, random and nested in upwelling). The design for the SPM in June and July consisted of the factors: site (four levels, random and crossed with month) and month (two levels, random and crossed with site); while in December the analysis comprised only the factor site (four levels, random). A Multivariate Permutation Analysis [46] of Bray-Curtis dissimilarities was used to assess differences among factors. Each term in the analysis was tested using > 9999 permutations as the relevant permutable units [47]. We used Principal Component Analysis (PCA), of square root transformed data, to explore differences in FA signatures among specimens [48]. For clarity, in the eigenvector plot we show only FA which had a coefficient higher than 0.5 for one of the axes. The combined results of the PCA and SIMPER (similarity percentage, PRIMER) were used to identify the FA responsible for differences among groups of samples, as the number of FA used (variables) was important and sometimes difficult to interpret based only on the PCA. Only FA forming more than 1% of TFA were included in the analyses (i.e. 31 FA used). The analyses were conducted using the PRIMER v6 and PERMANOVA+ [48,49]. Condition Index and Gonad Index Two-way ANOVA was used to test for an effect of upwelling on the CI of M. galloprovincialis in February. This included the factors: upwelling (two levels, fixed) and site (two levels, random and nested in upwelling). A second design was used to evaluate significant differences in CI of specimens among sites for June and July consisting of: month (two levels, random) and site (four levels, random); while a third design was used for December which comprised the factor site only (four levels, random). The same designs were used to test for significant effects on the GI of specimens used for the FA analyses. The violation of homogeneity of variances was considered to be acceptable because ANOVA is relatively robust to heterogeneous variances for large designs such as the one used in this study [45]. Results Upwelling characterization Onshore water temperature showed pronounced variability over the study period and marked differences among sites (Fig 2). From December to February conditions at the upwelling sites 1 and 3 (Llandudno and Paternoster) were colder than at their downstream sites 2 and 4 (Bloubergstrand and Elandsbaai; ANOVA, p < 0.01) by an average of 4.3 and 2.4°C respectively. ANOVA showed no significant differences among sites in June and July (p > 0.05). Periods of cooling events occurred at all sites, but with differences between the two categories of sites during the upwelling season. Sites 1 and 3 experienced three and six events of strong upwelling, respectively. One of these was protracted, running from 31st Dec and involving low temperatures (10–12°C) that lasted until the 13th of February (ΔT 8°C) at site 1 and until 30th of January (ΔT 5°C) at site 3. The other upwelling events were shorter (6–10 days) and less intense (ΔT 5°C), occurring in February and March. At sites 2 and 4, cooling events involved a drop in temperature of about 4°C with the exception of site 4, where a decrease of 6 and then 7°C was observed on the 31st December and 16th February, respectively. These events were less prolonged compared to upwelling events. Based on these analyses we confirmed the occurrence of upwelling at sites 1 and 3 (identified a priori as upwelling sites) in February, with no upwelling during winter (June and July) at these sites or at the non-upwelling sites, 2 and 4, in any sampling month. The protracted period of low seawater temperature that was observed at site 1 in January/ February suggested that upwelling occurred throughout the period of time prior the sampling event in February. Thus, we decided to classify site 1 as an upwelling site for the February collection. Additionally, these results allowed us to treat all the sites as experiencing non-upwelling conditions during winter (see Table A in S1 File). The lack of information on water temperature prior to sampling in December, due to the deployment of the data loggers on the same day as sample collection, did not allow us to categorize sites in either upwelling or non-upwelling. Thus, the samples for December were not tested for upwelling effects. 10.1371/journal.pone.0161919.g002Fig 2 Mean daily in situ sea temperatures at four sites on the South African west coast derived from onshore loggers for the duration of the investigation. The arrows indicate the four sampling events. Upwelling events were defined by a daily decline in mean temperature (ΔT) of ≥ 5°C. Fatty acid composition Food source No effect of upwelling on the SPM was found in this study, however PERMANOVA showed a significant effect of month and the interaction between month and site for June and July (both p < 0.01) and significant site effects in both December and February. Generally, SPM had higher proportions of saturated FA (SFA, 40–60%) than monounsaturated FA (MUFA, 20–45%) and polyunsaturated FA (PUFA, 10–25%) in all months (Table 1). The main SFA were represented by 16:0 (14–33%), followed by 18:0 (6–15%) and 14:0 (3–10%). The MUFA with the highest values were 18:1w9 (3–26%), 18:1w7 (1–11%) and 16:1w7 (12–12%), while among PUFA, only 18:2w6 and 20:3w3 counted for more than 5% of TFA at any specific site or sampling event (Table 1). Despite PERMANOVA highlighted dissimilarities, PCA and SIMPER showed no obvious pattern among sites (S1 Fig). Importantly, the differences observed in SPM did not correspond to the pattern observed in consumer tissues. 10.1371/journal.pone.0161919.t001Table 1 Total fatty acid composition of suspended particulate matter (SPM) collected during a) December and February and b) June and July across four sites along the South African west coast. The values are percentages expressed as mean ± standard deviation (n = 3 per site). A December February 1 2 3 4 1 2 3 4 14:0 5.28 ± 1.22 5.93 ± 0.54 3.90 ± 0.34 3.89 ± 1.04 3.06 ± 0.13 3.28 ± 1.09 3.04 ± 0.07 2.75 ± 1.00 14:1w5 0.99 ± 0.95 1.04 ± 1.12 0.29 ± 0.26 0.44 ± 0.36 0.21 ± 0.26 0.74 ± 0.82 1.53 ± 0.27 0.61 ± 0.23 16:0 20.82 ± 5.28 26.49 ± 2.33 33.01 ± 1.32 33.09 ± 0.86 20.01 ± 7.17 19.21 ± 1.41 22.41 ± 0.53 17.34 ± 5.80 16:1w7 6.36 ± 4.87 11.58 ± 1.66 2.65 ± 0.68 3.22 ± 1.68 4.01 ± 0.89 2.68 ± 0.27 6.19 ± 0.07 1.31 ± 0.53 16:1w5 3.55 ± 2.61 0.33 ± 0.21 1.84 ± 1.02 2.91 ± 0.83 0.76 ± 0.61 3.04 ± 2.07 1.96 ± 0.42 6.19 ± 1.81 17:1w7 0.77 ± 0.59 1.22 ± 0.35 0.38 ± 0.02 0.49 ± 0.21 0.46 ± 0.14 0.30 ± 0.18 0.62 ± 0.14 0.82 ± 0.36 18:0 11.36 ± 3.80 8.92 ± 1.26 11.22 ± 1.54 9.06 ± 1.46 14.43 ± 7.45 9.77 ± 4.85 7.08 ± 0.73 7.74 ± 2.24 18:1w9 9.55 ± 3.35 18.78 ± 8.83 20.07 ± 8.26 26.32 ± 4.84 18.44 ± 16.74 16.87 ± 4.26 22.23 ± 1.57 11.11 ± 4.30 18:1w7 3.31 ± 2.19 1.69 ± 2.07 4.28 ± 4.56 0.91 ± 0.22 10.44 ± 10.28 2.30 ± 0.69 2.27 ± 2.02 3.94 ± 0.61 18:1w5 0.99 ± 0.58 0.76 ± 0.13 0.74 ± 0.34 0.57 ± 0.17 0.42 ± 0.21 0.24 ± 0.07 0.62 ± 0.02 1.99 ±1.75 18:2w6 1.34 ± 0.51 2.12 ± 1.28 2.35 ± 1.07 2.89 ± 0.77 8.12 ± 4.91 5.69 ± 3.34 9.60 ± 0.42 2.02 ± 0.54 18:4w3 1.22 ± 1.05 0.91 ± 1.57 0.76 ± 0.48 0.82 ± 0.51 0.04 ± 0.08 2.87 ± 1.06 2.24 ± 0.71 4.99 ± 2.15 20:0 1.46 ± 1.32 0.69 ± 0.21 1.18 ± 0.04 1.16 ± 0.26 1.76 ± 2.53 0.96 ± 0.79 0.60 ± 0.23 2.03 ± 2.58 20:1w11 4.50 ± 7.62 1.03 ± 0.68 0.52 ± 0.09 0.94 ± 0.61 6.02 ± 9.47 4.37 ± 4.02 2.88 ± 0.37 3.13 ± 2.35 20:1w9 3.60 ± 2.03 1.97 ± 1.08 1.35 ± 0.82 0.95 ± 0.19 2.43 ± 3.30 3.99 ± 1.29 1.03 ± 1.66 5.62 ± 3.90 20:1w7 1.18 ± 1.22 0.80 ± 0.60 0.34 ± 0.33 0.64 ± 0.33 1.85 ± 2.10 0.17 ± 0.23 0.17 ± 0.29 0.54 ± 0.48 20:3w3 1.97 ± 2.20 1.20 ± 0.67 1.65 ± 0.94 0.74 ± 0.38 0.63 ± 0.26 3.54 ± 2.76 2.36 ± 0.66 5.84 ± 1.99 20:5w3 4.84 ± 4.31 0.25 ± 0.27 1.04 ± 0.74 1.89 ± 1.31 0.33 ± 0.54 4.25 ± 3.94 1.24 ± 0.14 2.10 ± 0.93 22:0 0.91 ± 1.18 0.50 ± 0.07 0.76 ± 0.40 0.42 ± 0.03 0.45 ± 0.26 3.41 ± 2.48 0.54 ± 0.06 6.00 ± 1.83 22:1w9 3.80 ± 2.37 2.68 ± 0.52 3.14 ± 1.36 1.69 ± 1.19 1.35 ± 0.58 4.15 ± 2.43 2.51 ± 0.57 1.60 ± 1.38 22:2w6 2.23 ± 2.98 2.27 ± 3.50 0.38 ± 0.66 0.66 ± 0.75 0.66 ± 0.48 0.61 ± 0.49 0.78 ± 0.14 1.17 ± 0.38 22:5w3 2.26 ± 2.74 1.39 ± 0.81 2.03 ± 1.14 1.51 ± 0.83 0.85 ± 0.42 3.72 ± 1.25 1.91 ± 0.54 4.53 ± 1.27 22:6w3 2.60 ± 1.67 2.14 ± 2.31 2.99 ± 1.04 1.29 ± 0.39 0.97 ± 0.81 1.70 ± 1.03 1.29 ± 0.23 2.89 ± 1.10 BAME 5.12 ± 2.83 5.31 ± 0.83 3.11 ± 0.70 3.51 ± 1.87 2.30 ± 0.69 2.15 ± 1.05 4.88 ± 0.24 3.76 ± 1.84 ΣSFA 44.95 ± 3.03 47.84 ± 3.11 53.19 ± 0.40 51.13 ± 2.25 42.01 ± 7.93 38.77 ± 6.64 38.55 ± 0.44 39.61 ± 7.58 ΣMUFA 38.60 ± 6.58 41.88 ± 2.05 35.60 ± 1.37 39.07 ± 2.15 46.40 ± 5.56 38.85 ± 3.44 42.02 ± 1.55 36.85 ±2.33 ΣPUFA 16.45 ± 3.59 10.28 ± 4.95 11.21 ± 1.23 9.80 ± 2.03 11.60 ± 4.85 22.38 ± 3.53 19.43 ± 1.52 23.53 ±5.37 B June July 1 2 3 4 1 2 3 4 14:0 3.87 ± 0.89 10.50 ± 0.48 7.96 ± 2.50 6.84 ± 0.40 4.75 ± 0.89 3.34 ± 0.48 2.72 ± 1.14 2.89 ± 0.55 14:1w5 0.18 ± 0.27 0.67 ± 0.20 0.53 ± 0.31 0.72 ± 0.33 0.45 ± 0.27 0.49 ± 0.20 0.21 ± 0.14 0.50 ± 0.25 16:0 29.17 ± 5.55 31.73 ± 3.47 28.95 ± 5.26 31.35 ± 2.35 28.40 ± 5.55 14.12 ± 3.47 20.02 ± 17.55 17.17 ± 2.96 16:1w7 4.23 ± 2.00 1.43 ± 1.15 2.44 ± 1.95 6.44 ± 6.65 3.99 ± 2.00 3.09 ± 1.15 8.25 ± 9.43 1.56 ± 0.39 16:1w5 1.51 ± 2.58 1.93 ± 0.47 2.03 ± 2.20 1.35 ± 1.17 5.45 ± 2.58 5.36 ± 0.47 4.05 ± 0.87 4.19 ± 0.65 17:1w7 0.28 ± 0.04 0.31 ± 0.09 0.44 ± 0.39 0.96 ± 1.17 0.15 ± 0.04 0.25 ± 0.09 0.29 ± 0.16 0.31 ± 0.32 18:0 5.98 ± 3.15 12.42 ± 1.51 10.68 ± 3.96 14.76 ± 7.56 12.50 ± 3.15 8.19 ± 1.51 15.01 ± 5.78 7.30 ± 4.11 18:1w9 18.13 ± 2.50 12.26 ± 1.97 10.28 ± 10.11 2.82 ± 2.67 12.95 ± 2.50 9.66 ± 1.97 10.20 ± 3.03 12.24 ± 4.18 18:1w7 3.08 ± 1.47 3.15 ± 0.22 5.22 ± 2.63 2.52 ± 1.68 5.29 ± 1.47 4.31 ± 0.22 4.36 ± 0.12 9.56 ± 8.02 18:1w5 1.78 ±1.75 0.07 ± 1.65 1.66 ± 1.01 0.57 ± 0.75 3.61 ± 1.75 3.79 ± 1.65 3.42 ± 1.12 2.78 ± 1.90 18:2w6 3.26 ± 1.01 2.25 ± 2.76 1.73 ± 1.47 0.84 ± 0.69 3.74 ± 1.01 9.63 ± 2.76 4.40 ± 3.62 22.06 ± 2.83 18:4w3 3.61 ± 2.75 0.77 ± 0.67 0.53 ± 0.68 2.22 ± 0.75 2.17 ± 2.75 1.99 ± 0.67 2.41 ± 1.84 1.08 ± 0.64 20:0 7.25 ± 1.55 1.32 ± 3.18 1.82 ± 1.26 1.75 ± 0.38 1.87 ± 1.55 4.43 ± 3.18 1.25 ± 0.16 2.98 ± 2.10 20:1w11 3.21 ± 0.04 0.69 ± 1.40 1.87 ± 0.44 0.79 ± 0.23 0.52 ± 0.04 3.08 ± 1.40 3.01 ± 2.32 1.23 ± 0.54 20:1w9 1.55 ± 0.24 0.85 ± 2.27 1.30 ± 0.85 1.20 ± 1.29 0.37 ± 0.24 4.14 ± 2.27 1.77 ± 0.53 0.79 ± 0.09 20:1w7 0.38 ± 0.43 0.22 ± 1.36 0.60 ±0.22 1.53 ± 1.86 0.39 ± 0.43 3.07 ± 1.36 0.63 ± 0.62 0.69 ± 0.67 20:3w3 0.57 ± 0.49 0.95 ± 0.42 0.87 ± 0.78 0.95 ± 1.09 0.66 ± 0.49 0.67 ± 0.42 0.75 ± 0.32 0.33 ± 0.13 20:5w3 1.54 ±0.56 1.40 ± 2.49 1.60 ± 0.57 1.33 ± 1.01 1.03 ± 0.56 3.48 ± 2.49 1.66 ± 0.77 1.49 ± 0.45 22:0 1.38 ± 0.10 2.12 ± 0.01 3.00 ± 1.96 1.00 ± 0.14 0.67 ± 0.10 0.48 ± 0.01 0.42 ± 0.15 1.25 ± 1.22 22:1w9 1.96 ± 2.04 4.00 ± 2.56 2.86 ± 1.63 2.27 ± 3.28 2.40 ± 2.04 4.54 ± 2.56 2.92 ± 1.75 2.04 ± 0.73 22:2w6 0.26 ± 0.11 1.14 ± 0.27 1.21 ± 0.67 1.86 ± 2.19 0.37 ± 0.11 1.02 ± 0.27 0.41 ± 0.11 0.29 ± 0.30 22:5w3 0.95 ± 0.47 1.25 ± 1.30 2.92 ± 1.42 2.14 ± 1.83 1.30 ± 0.47 1.52 ± 1.30 0.71 ± 0.29 0.49 ± 0.20 22:6w3 1.96 ± 0.57 2.48 ± 1.66 3.34 ± 2.30 2.64 ± 2.24 1.19 ± 0.57 2.62 ± 1.66 2.03 ± 0.74 1.19 ± 0.24 BAME 3.90 ± 0.68 6.10 ± 2.29 6.16 ± 2.56 11.14 ± 7.16 5.78 ± 0.68 6.72 ± 2.29 9.09 ± 5.60 5.60 ± 1.54 ΣSFA 51.56 ± 6.17 64.20 ± 6.90 58.56 ± 8.55 66.85 ± ## 53.98 ± 6.17 37.29 ± 6.90 48.52 ± 16.53 37.19 ± 7.68 ΣMUFA 36.29 ± 9.30 25.56 ± 6.11 29.25 ± ### 21.16 ± ## 35.56 ± 9.30 41.79 ± 6.11 39.11 ± 13.89 35.89 ± 5.05 ΣPUFA 12.15 ± 3.51 10.24 ± 8.31 12.19 ± 4.93 11.99 ± 9.07 10.46 ± 3.51 20.92 ± 8.31 12.37 ± 3.10 26.92 ±2.77 Only FA > 1% are displayed. PUFA = Polyunsaturated Fatty Acids, MUFA = Monounsaturated Fatty Acids, SFA = Saturated Fatty Acids, BAME = Bacterial Fatty Acids. Consumers A total of 28 and 29 FA contributing > 1% to TFA were found in the adductor muscles and the gonads respectively, but their compositions differed (Tables B and C in S1 File). The major FA found in adductor muscles were 16:0 (18–23%), 18:0 (5–7%), 22:2 Non-methylene-interrupted FA (NMI; 4–5%), 20:2NMI1 (5–7%), 20:5w3 (10–13%) and 22:6w3 (12–20%); while those in gonad tissues were 16:0 (20–23%), 16:1w7 (3–7%), 18:0 (3–7%), 20:5w3 (12–27%) and 22:6w3 (10–26%). The main source of difference amongst all samples was between the two types of tissues (PERMANOVA, p < 0.001). SIMPER highlighted that the main differences between the tissues were due to higher proportions of 18:1w9, diatom trophic markers (TM, 16:1w7 and 20:5w3) and dinoflagellate TM (18:4w3 and 22:6w3) in gonads; and higher proportions of 18:0, 20:0, 20:2NMI1, 20:4w6 and 22:2NMI1 in the adductor muscles, explaining 55% of the FA differences between the two tissues (S2 Fig). Gonads also showed a higher proportion of essential FA (EFA; 46% of TFA) than muscles (26–35% of TFA) at all sites and across all months. In particular, gonads had twice the proportion of 20:5w3 compared to the adductor muscles. Considering the strong dissimilarities between gonads and adductor muscles, the FA compositions of these tissues were investigated separately in the remaining analyses. Consumers: adductor muscle Adductor muscle samples were characterized by high proportions of PUFA throughout the year (48–57%), followed by SFA (31–42%) and MUFA (10–16%; Table 2). PERMANOVA showed that in December the factor site was significant (Table 3), indicating that the FA signatures of the mussels differed across sites. Similarly, during the month of February, the only source of variation among muscle samples was the factor site, with no significant effect of the factor upwelling (Table 3). 10.1371/journal.pone.0161919.t002Table 2 Multiple mean comparisons of the main fatty acid trophic markers of the adductor muscle and gonad of Mytilus galloprovincialis among the four sites in a) December and February and b) June and July. The values for ΣSFA, ΣMUFA, ΣPUFA and ΣEFA are percentages expressed as mean ± standard deviation (n = 6 per site). A December February 1 2 3 4 1 2 3 4 Adductor muscle ΣSFA 34.53+1.06 35.02+0.67 34.07+1.54 34.78+0.97 33.00+1.20 33.29+0.32 31.44+2.19 31.67+1.29 ΣMUFA 10.88+0.97 10.15+0.72 11.85+2.31 11.76+1.14 16.33+2.59 15.17+2.19 12.76+2.42 10.82+2.34 ΣPUFA 54.59+1.24 54.83+0.86 54.07+2.63 53.46+1.18 50.67+2.22 51.54+2.11 55.81+3.84 57.52+3.51 ΣEFA 32.91+0.66 33.47+1.25 36.76+2.60 33.80+2.69 30.08+2.16 29.20+3.03 33.67+2.34 26.61+1.53 20:5w3/22:6w3 0.73 0.85 0.59 0.66 1.02 1.22 0.63 0.73 Gonad ΣSFA 34.82+3.15 31.00+2.48 34.54+2.76 31.78+2.00 31.61+2.54 33.58+1.67 33.76+5.40 31.76+1.24 ΣMUFA 13.66+4.63 15.14+2.76 11.02+4.96 17.72+4.82 18.40+4.69 15.46+4.18 12.13+5.02 17.97+3.74 ΣPUFA 51.52+3.15 53.86+2.36 54.44+4.93 50.51+4.14 49.99+3.19 50.96+3.74 54.11+1.85 50.27+3.70 ΣEFA 40.11+5.12 40.96+2.91 45.07+8.36 36.95+4.52 37.65+4.06 39.54+4.95 41.30+4.74 36.39+3.38 20:5w3/22:6w3 1.77 1.74 0.58 1.23 1.73 2.84 0.66 1.49 B June July 1 2 3 4 1 2 3 4 Adductor muscle ΣSFA 35.80+2.26 34.82+1.21 35.72+2.49 37.40+3.00 41.42+6.98 35.16+1.48 37.22+4.28 38.00+4.46 ΣMUFA 11.98+1.58 11.77+1.29 12.24+2.31 10.63+0.93 10.60+0.52 11.66+2.18 11.52+2.44 11.84+1.97 ΣPUFA 52.23+1.74 53.41+0.76 52.04+1.26 51.97+3.59 47.98+6.85 53.18+2.49 51.26+2.88 50.16+3.63 ΣEFA 33.31+3.01 35.89+2.37 35.91+1.39 33.77+2.84 28.34+7.87 35.94+4.13 35.89+3.99 31.60+3.43 20:5w3/22:6w3 0.80 1.26 0.61 0.49 0.76 1.15 0.61 0.52 Gonad ΣSFA 30.96+1.90 32.18+1.18 33.04+2.12 33.02+0.65 31.16+7.09 34.40+2.70 34.11+1.67 32.67+1.38 ΣMUFA 14.39+3.69 14.64+2.64 16.37+3.84 12.66+3.14 13.51+2.68 12.24+3.82 10.94+3.24 13.09+3.39 ΣPUFA 54.65+2.94 53.18+2.28 50.58+4.18 54.32+3.19 55.33+4.55 53.37+1.48 54.95+1.67 54.23+2.75 ΣEFA 42.43+4.32 40.16+1.95 39.48+4.78 41.63+4.33 40.41+2.60 41.66+3.13 46.56+4.36 40.49+4.79 20:5w3/22:6w3 1.27 1.68 0.61 0.49 1.12 1.48 0.68 0.57 Only FA > 1% were used for the analyses. PUFA = Polyunsaturated Fatty Acids, MUFA = Monounsaturated Fatty Acids, SFA = Saturated Fatty Acids, EFA = Essential Fatty Acids (20:4w6, 20:5w3 and 22:6w3). 10.1371/journal.pone.0161919.t003Table 3 PERMANOVA results on the fatty acid composition of the adductor muscles of Mytilus galloprovincialis at four sites and across four month on the South African west coast. December February June- July df MS Pseudo-F p df MS Pseudo-F p df MS Pseudo-F p Si 3 74.09 4.40 0.000 *** Up 1 173.65 1.01 0.669 Month 1 40.08 1.23 0.282 Res 20 16.84 Si (Up) 2 171.47 7.01 0.000 *** Si 3 268.35 8.24 0.000 *** Res 20 24.454 Month x Si 3 33.21 1.02 0.439 Res 40 32.58 Up = Upwelling, Si = Site, df = degrees of freedom, MS = mean square; * p < 0.05; ** p < 0.01; *** p < 0.001. In December, sites 1 and 2 were different from each other and from all the other sites (PERMANOVA post hoc pair-wise test, p < 0.05) by being characterized by SFA, 20:4w6 and 22:2w6 in the case of site 1, while site 2 had high percentages of diatom TM and 22:2NMI1. Sites 3 and 4 did not differ from each other and they exhibited a high proportion of 18:1w9, dinoflagellate TM and 20:1w11. These FA contributed to 55% of the dissimilarity among sites (SIMPER). In February, only the two non-upwelling sites, 2 and 4, were significantly different from each other, while sites 1 and 3 did not differ (PERMANOVA post hoc pair-wise test and PCA). SIMPER showed that sites 1, 2 and 3 had higher proportions of 14:0 and diatom TM than site 4 (20% of TFA), which was enriched in, 20:2NMI1, 22:2w6 and 20:4w3 (17% of TFA; Table B in S1 File). This is partially in agreement with the ratio 20:5w3/ 22:6w3, which discriminates between diatoms (> 1) and dinoflagellates (< 1) ([50,51]) and indicated a predominance of diatoms in mussel diets (values > 1) in February at sites 1 and 2. For June and July, PERMANOVA highlighted a significant effect of the factor site only (Table 3). All sites were different from each other in both months, with the exception of sites 1 and 4 which did not differ. As for December, specimens at site 2 were typified by diatom TM, site 3 by dinoflagellate TM, while sites 1 and 4 were characterized by 18:1w9, 20:1w9, 20:2NMI1, 20:4w6 and 22:2w6 (SIMPER, Fig 3). The ratio 20:5w3/ 22:6w3 again showed a dominance of dinoflagellates at all sites with the exception of site 2 in June and July when it was diatom dominated. 10.1371/journal.pone.0161919.g003Fig 3 PCA conducted on the adductor muscles of Mytilus galloprovincialis collected at rocky shore sites, on the South African west coast in a) December, c) February e) June and July. Each symbol represents a single replicate. b, d, f) Eigenvalues for each of the factors (fatty acids). The circle corresponds to eigenvalues of −1 to 1. For clarity, only fatty acids with eigenvalues > 0.5 are shown. Consumers: gonad As for the adductor muscle, gonad FA compositions included a high proportion of PUFA at all sites and months (50–55% of TFA), followed by SFA (30–35%) and MUFA (10–15%; Table 2). The 20:5w3/22:6w3 ratio was variable among sites: at sites 1 and 2 the ratio was > 1 across all months; at site 3 it was < 1, while at site 4 it was > 1 during the summer months and decreased to < 1 in the winter period. PERMANOVA showed that the only significant factor for the analyses performed was site, with no effect of upwelling in February, indicating that patterns in the FA composition of gonads differed across sites and that these differences remained constant among months (Table 4). Axes 2 of the PCA showed a partial separation between samples of sites 1–2 and sites 3–4 in December, February and June- July (S3 Fig). This pattern was not confirmed by the PERMANOVA pairwise test, however, which showed that only sites 3 and 4 did not differ from each other in all months (p > 0.05). Thus sites were different among each other, however, we could not identify a clear pattern of separation (S3 Fig). 10.1371/journal.pone.0161919.t004Table 4 PERMANOVA results on the fatty acid compositions of gonads of Mytilus galloprovincialis at four sites and across four month on the South African west coast in relation to the different factors. December February June- July df MS Pseudo-F p df MS Pseudo-F P df MS Pseudo-F p Si 3 267.79 2.96 0.004 ** Up 1 188.08 0.82 1 Month 1 121.41 2.32 0.070 Res 20 90.479 Si (Up) 2 228.69 4.53 0.004 ** Si 3 355.57 6.79 0.000 *** Res 20 50.512 Month x Si 3 64.39 1.23 0.273 Res 38 52.34 Up = Upwelling, Si = Site, df = degrees of freedom, MS = mean square; * p < 0.05; ** p < 0.01; *** p < 0.001. Indices Condition Index Month did not affect CI in winter, nor did upwelling have a significant effect in February (ANOVA, p > 0.05 for both analyses), however, there were significant effects of the factor site in all analyses (ANOVA, p < 0.001) and the interaction between site and month in June-July (ANOVA, p < 0.05). Tukey HSD tests showed that site 1 had the lowest CI (4.05 ± 1.20; p < 0.001, Fig 4a), while site 2 had CI values that were intermediate between those of site 1 and sites 3–4. These latter sites were statistically different only in December, when CI values at site 3 were significantly higher than at the other three sites (12.80 ± 3.20; p < 0.01). In addition, CI at site 2 increased from the summer to the winter months (5.55 ± 1.9–9.05 ± 2.70). The results indicate a clear south-north gradient of increasing CI from sites 1 to 3, followed by a decrease at site 4, a pattern that was consistent across all months. 10.1371/journal.pone.0161919.g004Fig 4 a) Condition Index and b) Gonad Index of Mytilus galloprovincialis at the four sites and across the four months sampled (n = 26 and n = 6 for CI and GI, respectively). Sites are arranged from south to north (left to right). Values are means, the error bars indicate standard deviations. Letters indicate homogenous groups within months (p > 0.05). Gonad Index The three analyses indicated that the only significant effects were of site and its interaction with month in the winter months (ANOVA, p < 0.01 for all analyses), with no upwelling effect in February. In December, GI at site 3 was significantly higher than the other three sites (Tukey HSD, p < 0.05), which did not differ from each other. In February and June the four sites did not differ (Tukey HSD, p > 0,05), while in July sites 3 and 4 had higher GI values than sites 1, and site 2 was not different from the other sites (Tukey HSD, p < 0.01; Fig 4b). Although differences among sites were not always significant, the spatial pattern was identical to that for CI, with a trend of increasing values from site 1 to site 3 followed by a slight decrease at site 4. Discussion Using FA techniques, we aimed to assess the effects of upwelling on the dietary signatures of benthic filter feeders in the Benguela upwelling region, by comparing sites that are considered to be strong upwelling centres with downstream sites, which may be influenced only indirectly by upwelling. The temperature loggers confirmed the categorisation of sites 1 and 3 as upwelling centres [37], but with upwelling occurring only during summer, predominantly in January and February, as previously shown [34,19] and in situ temperatures indicated non-upwelling conditions at all sites during the winter sampling months from April to July. Throughout summer, from December to March, the upwelling sites were colder and were characterized by frequent, intense upwelling events, with drops in mean daily temperatures of as much as 8°C. Site 2 clearly behaved as a non-upwelling site, while site 4 had temperatures that were on average higher than at the upwelling sites, but it did experience two weak upwelling events during the study. This reflects the inter-annual variability in upwelling frequency and intensity that has been shown for the Benguela and other upwelling systems [52,53]. We anticipated a stronger diatom FA signature at upwelling sites than at downstream sites, and expected the difference to be even more pronounced during the upwelling season. The SPM data did not support this hypothesis, with no differences in the SPM FA composition among sites or sampling events. SPM is an instantaneous measure, however, reflecting the FA composition of the food available for filter feeders at the moment of the sampling, while the FA turn-over rate in mussels is approximately one month [35]. Consequently, the SPM collected on the same day as the mussels does not necessary represent what the organisms were feeding on a few days/weeks before. We did not, therefore, expect a strong link between the FA composition of the SPM and the mussels, but we did expect some differences in the SPM signatures in relation to upwelling. The spatial and temporal homogeneity of the SPM even when temperature data indicated the occurrence of upwelling, was unexpected and suggests either that upwelling influenced SPM conditions in a homogeneous way across the study region, or that changes in primary producers were too rapid to be detected with our approach. Although there was a predominance of phytoplankton in the diets of our specimens, the results for adductor muscle indicated idiosyncrasy among sites, with no significant effect of upwelling. Specimens from site 3 were characterized by dinoflagellate TM throughout the investigation, and the ratio of 20:5w3/22:6w3 indicated a generally higher contribution of dinoflagellates than diatoms (ratio < 1 [50,51]) at sites 1, 3 and 4, while the presence of diatom TM in all months and a ratio of > 1 in all sampling events indicated a predominance of diatoms in the diets of mussels at site 2. The presence of the cold, nutrient-rich Benguela Current on the South African west coast promotes high primary production dominated by phytoplankton [32,54,55], explaining the strong phytoplankton FA signatures at all sites observed for M. galloprovincialis. Gonad has rapid turnover rates, so that its FA signatures change more rapidly than those of the more conservative adductor muscle [56]. Gonad showed marked variability in the 20:5w3/22:6w3 ratio and in the FA composition among sites and months, but in this case, there was no clear pattern. Although both CI and GI showed strong variability among sites, with no significant effects of the factors upwelling or month, both exhibited a marked geographic gradient. In all months, values for both indices increased from south to north, from site 1 to site 3, slightly dropping at site 4, with site 1 showing the lowest CI across all months and site 3 the highest. This pattern is intriguing as temperature data indicated strong upwelling at site 1 during summer, and we expect higher phytoplankton concentrations during upwelling events [7,57]. The low CI suggests, however, that the conditions at this site were less favourable for M. galloprovincialis than at the other sites. The low water temperatures experienced by animals at site 1 are unlikely to be responsible for the low CI measured. The average seawater temperature for the study period was approximately 13–15°C with minimum values down to 9–10°C. Some studies have shown that temperatures < 13°C are suboptimal for M. galloprovincialis [58,59], but another showed that the thermal limits of this species were between 5–27°C [60]. M. galloprovincialis is well represented in the intertidal zone of the west coast of South Africa; it has displaced the indigenous species to become the dominant mussel along this coast [61] and is well-adapted to cold environments [62,60]. More plausibly, the results may reflect low food availability at this site as has been observed in other systems [63,64]. Field et al [65] conducted daily measurements for 30 days at a site just 2 km from our site 1 and showed low nearshore chlorophyll a concentrations during upwelling driven by strong offshore winds. Consequently, phytoplankton blooms supported by this nutrient injection develop over the course of several days farther offshore and are advected to downstream areas. In contrast, the non-upwelling site 4 had CI values comparable to site 3, which had the highest CI on three of the four sampling events. This presumably reflects topographic effects, as site 4, Elandsbaai, is located in a bay, where chlorophyll a and mussel growth are likely to be greater due to retention of particles [66]. CI thus seems to reflect not just the effects of upwelling on primary production, but the effects of local hydrodynamics which determine whether enhanced productivity is retained near or advected away from sites. GI showed an identical pattern to CI, but values at site 1 were comparable to the other sites. This suggests that animals at site 1 invested as heavily in reproduction as the other populations despite experiencing unfavourable conditions and low CI values. Investment in reproduction has been shown to be independent of body condition and feeding regime in other invertebrates such as clams [67] and squid [68]. Importantly the GI in mussels is closely related to the reproductive phase: the GI will be low if an individual has just spawned and high if it is collected immediately before spawning. According to van Erkom Schurink & Griffiths [69], M. galloprovincialis in South Africa has a peak spawning period in summer (November- December) and one in winter (April- June), though the timing of mussel spawning can change markedly [70]. Although we have no information on the reproductive state of the sampled populations or whether they were reproductively synchronised, the results were consistent across the months sampled, indicating consistency in the pattern observed. The clearest results highlighted strong differences in the FA compositions of gonad and the more stable adductor muscle tissues, with gonad showing a higher proportion of EFA (i.e. 20:4n-6, 20:5n-3 and 22:6n-3; [71]) across all months and sites. EFA are critical for the maintenance of cell membrane structure and function. They are also precursors of bioactive compounds that are essential for survival, growth and reproduction [72,73], however they cannot be synthetized de novo in sufficient quantities and must be acquired through the diet [74,75]. Previous studies have suggested preferential retention of specific FA in the gonad, in order to ensure reproduction and to improve offspring survival [76–78]. Increased concentrations of specific FA in the gonads during reproduction have been shown for a range of animals including octopuses [79], fishes [80] and bivalves [81]. In situ temperature loggers used to identify upwelling conditions at the four sites indicated that such events occurred only at the upwelling centres and only in austral summer. The high proportion of EFA we found, especially of 20:5w3, suggests preferential retention of EFA in the gonads, presumably for reproductive purposes. The contrast between GI and CI results supports this interpretation. Although both GI and CI were consistently low at site 1, the differences were only significant for CI, suggesting that investment of resources into reproduction was a priority, maintained even at the cost of adult condition. The strong dissimilarities between gonad and muscle may also reflect the very different turnover rates of the two tissue types, with gonad changing FA signatures faster than adductor muscle [56,82]. Upwelling or cooling events occurred in late December-February and adductor muscle samples from February were characterized by a high proportion of diatom TM. This matches the one month FA turn-over period found for mussels [35], reflecting the effects of upwelling from the previous month. However, this effect was not detectable in gonads, which showed differences in FA signatures, but without a clear pattern. We found no direct effect of upwelling on the FA composition of muscle or gonad tissues. However, strong dissimilarities were found among sites which were mostly consistent across the four months sampled. Adductor muscle at non-upwelling site 2 were characterised by diatom TM, at the upwelling site 3 they were enriched in dinoflagellate TM, while at sites 1 and 4 muscle samples generally had high proportions of PUFA, though no specific FATM were identified. These results partially contradict our expectations of stronger phytoplankton TM at upwelling centres, with the two upwelling sites behaving in opposite ways. Site 3 conformed to the general conception of a late stage of upwelling on three of the four sampling events, being characterised by dinoflagellate TM (post diatom blooms), but site 1 was not typified by any distinct phytoplankton TM. Likewise, site 2, which is supposedly a downstream, non-upwelling site, had a strong diatom signature. Similar disparities were found with the condition indices. Site 3 had the highest CI and GI among all sites and in all months, while site 1 showed the lowest. However, in situ temperature data indicated that, during the upwelling season, both upwelling sites were exposed to long periods of upwelled water, while the two non-upwelling sites, with two exceptions at site 4, experienced cooling events that were less frequent, less protracted and less intense. The effects of upwelling can be interpreted in two ways. At upwelling sites, nutrients and particles can be upwelled close to the coast and then moved offshore with the surface water, resulting in phytoplankton-poor waters inshore [65,83–85]. Offshore incubation of these enriched waters can be followed by onshore advection (either to the same point on the shore or farther downstream) during upwelling relaxation or downwelling. Thus, the retention time of the newly upwelled water can be too short to allow bloom development at the upwelling site itself, while phytoplankton blooms are detectable downstream of the upwelling centre, so that downstream sites can exhibit higher phytoplankton concentrations than the upwelling centres themselves. Other studies have indicated that upwelling stimulates phytoplankton and macrophyte growth at the site of upwelling, which suggests a very localised effect [86,1]. In our study, the first scenario seems to apply to the comparison of sites 1 and 2, and the second to sites 3 and 4, presumably reflecting differences in nearshore hydrodynamics in the two regions. For instance, local hydrodynamics may result in offshore advection of water at site 1 but retention at site 3. Although we could not classify sites as either upwelling or non-upwelling in the month of December, the FA results conformed to expectations based on such categorisation, showing dinoflagellate TM at site 3 and diatom TM at site 2, as a site downstream of site 1. This agrees with the fact that upwelling occurs during the summer months in the Southern Benguela area and that the sites we chose a priori as upwelling or non-upwelling sites for the comparison conformed to this classification. Our results partially contradict the findings of Xavier et al [37]. Upwelling has long been identified as a key process influencing the advection of planktonic larvae to/from the nearshore [87,88] and, using the same sites as us, Xavier et al [37] found higher recruitment of the mussel M. galloprovincialis at upwelling centres than at the downstream sites. The discrepancy between the two investigations may reflect inter-annual variability in upwelling intensity and frequency as is known to occur in this area [52,53], or may more simply indicate that nutrient levels, phytoplankton response and the advection of mussel recruits all respond to upwelling with very different lag periods. Another important aspect to highlight is that the differences observed among sites were consistent in both the upwelling and non-upwelling seasons, suggesting that differences in the conditions at these sites perpetuate across the upwelling and non-upwelling seasons. Adductor muscle tissue had a strong phytoplankton FA signature at all sites in February, during the upwelling season, suggesting that a very strong upwelling event occurred during the previous weeks, as was shown by the temperature data. However, even if upwelled waters were present only during the summer months, the tissues of mussels showed similar proportions of PUFA-MUFA-SFA over time, indicating that the specimens were exposed to similar food in both austral summer and winter and under both upwelling and non-upwelling conditions. The fact that the factor upwelling was not significant in any of the analyses, suggests that the entire study area is affected by upwelling, indicating that the influence of upwelling in the southern Benguela system is pervasive, rather than discrete. This partially contrasts with a similar study conducted over a larger scale with more widely separated sites on the same coast [89]. On comparing sites spread across larger scales (approximately 400 km of the west coast of South Africa), the effects of upwelling were found to be discrete, with mussels and barnacles at upwelling sites having different FA and stable isotope signatures from specimens at non-upwelling sites in the same biogeographic province [89]. The discrepancy between the two studies presumably reflects their different spatial resolutions, as well as their geography. One study was conducted over 400 km of coastline, including an area close to the Luderitz upwelling cell in Namibia, which is characterised by frequent and intense upwelling events throughout the year, whereas the other encompassed 200 km of coast where upwelling occurs only seasonally [19,32,90]. Thus, depending on the spatial scale of resolution and the nature of upwelling, its effects can be either discrete or diffuse. We tend to characterise upwelling as a binary phenomenon, but clearly, the intensity and frequency of upwelling events are critical, as are the effects of nearshore hydrodynamics, which can either retain nutrient enriched waters and the associated primary production locally or export them to downstream areas. Overall, we demonstrated upwelling at sites categorised a priori as upwelling centres, but only during the upwelling season. Despite the seasonal differences in upwelling conditions at these sites compared to the non-upwelling sites, we found no clear differences in the FA of mussels at the two types of sites, indicating that the effects of nutrient enrichment were diffuse across the study area. Critically, we found that upwelling centres behaved differently, presumably reflecting the nature of retention and advection of upwelled water. Supporting Information S1 Fig a) PCA conducted on the fatty acid composition of suspended organic matter (SPM) collected from intertidal rocky shores at four sites during the four sampling events. Each symbol represents a single replicate. Upwelling did not have an effect on the fatty acid signatures of samples from February and thus we show a single PCA with the samples of all months together. b) Eigenvalues for each of the factors (fatty acids). The circle corresponds to eigenvalues of −1 to 1. For clarity, only fatty acids with eigenvalues > 0.5 are shown. (TIF) Click here for additional data file. S2 Fig a) PCA conducted on the adductor muscle and gonad tissues of Mytilus galloprovincialis collected at intertidal rocky shores at each site during the four sampling events. Each symbol represents a single replicate. b) Eigenvalues of each of the factors (fatty acids). The circle corresponds to eigenvalues of −1 to 1. For clarity, only fatty acids with eigenvalues > 0.5 are shown. (TIF) Click here for additional data file. S3 Fig PCA conducted on the fatty acid composition of gonad tissue of Mytilus galloprovincialis collected from intertidal rocky shores at four sites during a) December c) February e) June and July. Each symbol represents a single replicate. b), d) and f) Eigenvalues of each of the factors (fatty acids). The circle corresponds to eigenvalues of −1 to 1. For clarity, only fatty acids with eigenvalues > 0.5 are shown. (TIF) Click here for additional data file. S1 File Supplementary material. (DOCX) Click here for additional data file. This work was supported by funding from the Andrew Mellon Foundation and the South African Research Chairs Initiative of the Department of Science and Technology and the National Research Foundation of South Africa. Fatty acid processing was accomplished at the Fatty Acid Facility at Rhodes University provided by the National Research Foundation of South Africa and Rhodes University. Special thanks to the Coastal Research Group of Rhodes University, Grahamstown, South Africa. We gratefully acknowledge the contributions of anonymous referees whose comments greatly improved the manuscript. ==== Refs References 1 Wieters EA . Upwelling control of positive interactions over mesoscales: a new link between bottom-up and top-down processes on rocky shores . Mar Ecol Prog Ser . 2005 ;301 : 43 –54 . 2 Blanchette CA , Broitman BR , Gaines SD . Intertidal community structure and oceanographic patterns around Santa Cruz Island, CA, USA . Mar Biol . 2006 ;149 : 689 –701 . 10.1007/s00227-005-0239-3 3 Menge BA , Daley BA , Wheeler PA , Dahlhoff E , Sanford E , Strub PT . Benthic–pelagic links and rocky intertidal communities: bottom-up effects on top-down control? Proc Natl Acad Sci . 1997 ;94 : 14530 –14535 . 9405647 4 McQuaid CD , Payne AIL . Regionalism in marine biology: the convergence of ecology, economics and politics in South Africa . South Afr J Sci . 1998 ;94 : 433 –436 . 5 Connell JH . The consequences of variation in initial settlement vs. post-settlement mortality in rocky intertidal communities . J Exp Mar Biol Ecol . 1985 ;93 : 11 –45 . 10.1016/0022-0981(85)90146-7 6 Raimondi PT . Patterns, mechanisms, consequences of variability in settlement and recruitment of an intertidal barnacle . Ecol Monogr . 1990 ;60 : 283 –309 . 10.2307/1943059 7 Figueiras FG , Labarta U , Reiriz MJF . Coastal upwelling, primary production and mussel growth in the Rías Baixas of Galicia In: Vadstein O , Olsen Y , editors. Sustainable Increase of Marine Harvesting: Fundamental Mechanisms and New Concepts . Springer Netherlands ; 2002 pp. 121 –131 . Available: http://link.springer.com/chapter/10.1007/978-94-017-3190-4_11 8 Bosman AL , Hockey P a. R , Siegfried WR . The influence of coastal upwelling on the functional structure of rocky intertidal communities . Oecologia . 1987 ;72 : 226 –232 . 10.1007/BF00379273 28311545 9 Nielsen KJ . Nutrient loading and consumers: agents of change in open-coast macrophyte assemblages . Proc Natl Acad Sci . 2003 ;100 : 7660 –7665 . 10.1073/pnas.0932534100 12796509 10 Smith JR , Fong P , Ambrose RF . Spatial patterns in recruitment and growth of the mussel Mytilus californianus (Conrad) in southern and northern California, USA, two regions with differing oceanographic conditions . J Sea Res . 2009 ;61 : 165 –173 . 10.1016/j.seares.2008.10.009 11 Cury P , Roy C . Optimal environmental window and pelagic fish recruitment success in upwelling areas . Can J Fish Aquat Sci . 1989 ;46 : 670 –680 . 10.1139/f89-086 12 Ryther JH . Photosynthesis and fish production in the sea . Science . 1969 ;166 : 72 –76 . 10.1126/science.166.3901.72 5817762 13 Bustamante RH , Branch GM , Eekhout S , Robertson B , Zoutendyk P , Schleyer M , et al Gradients of intertidal primary productivity around the coast of South Africa and their relationships with consumer biomass . Oecologia . 1995 ;102 : 189 –201 . 10.1007/BF00333251 28306874 14 Sanford E , Menge BA . Spatial and temporal variation in barnacle growth in a coastal upwelling system . Mar Ecol Prog Ser . 2001 ;209 : 143 –157 . 15 Ventura CRR , Falcao APC , Santos JS , Fiori CS . Reproductive cycle and feeding periodicity in the starfish Astropecten brasiliensis in the Cabo Frio upwelling ecosystem (Brazil) . Invertebr Reprod Dev . 1997 ;31 : 135 –141 . 10.1080/07924259.1997.9672571 16 Skinner LF , Macharet HKL , Coutinho R . Influence of upwelling and tropical environments on the breeding development of the intertidal barnacle Tetraclita stalactifera (Lamarck, 1818) . Braz J Oceanogr . 2011 ;59 : 349 –356 . 10.1590/S1679-87592011000400005 17 Adamec D , O’Brien JJ . The seasonal upwelling in the Gulf of Guinea due to remote forcing . J Phys Oceanogr . 1978 ;8 : 1050 –1060 . 10.1175/1520-0485(1978)008<1050:TSUITG>2.0.CO;2 18 Lewis RK . Seasonal upwelling along the south-eastern coastline of South Australia . Mar Freshw Res . 1981 ;32 : 843 –854 . 19 Field JG , Shillington FA . Variability of the Benguela Current system The sea: the global coastal ocean: interdisciplinary regional studies and syntheses . Harvard University: Harvard University Press ; 2006 pp. 835 –863 . 20 García-Reyes M , Largier JL . Seasonality of coastal upwelling off central and northern California: new insights, including temporal and spatial variability . J Geophys Res Oceans . 2012 ;117 : C03028 10.1029/2011JC007629 21 Corbisier TN , Petti MAV , Soares LSH , Muto EY , Bromberg S , Valiela I . Trophic structure of benthic communities in the Cabo Frio upwelling system (southeastern Brazilian shelf): a temporal study using stable isotope analysis . Mar Ecol Prog Ser . 2014 ;512 : 23 –38 . 10.3354/meps10947 22 Soares LSH , Muto EY , Lopez JP , Clauzet GRV , Valiela I . Seasonal variability of δ13C and δ15N of fish and squid in the Cabo Frio upwelling system of the southwestern Atlantic . Mar Ecol Prog Ser . 2014 ;512 : 9 –21 . 10.3354/meps10948 23 Stenseth NC , Llope M , Anadón R , Ciannelli L , Chan K-S , Hjermann DØ , et al Seasonal plankton dynamics along a cross-shelf gradient . Proc R Soc Lond B Biol Sci . 2006 ;273 : 2831 –2838 . 10.1098/rspb.2006.3658 24 Chenillat F , Rivière P , Capet X , Di Lorenzo E , Blanke B . North Pacific Gyre Oscillation modulates seasonal timing and ecosystem functioning in the California Current upwelling system . Geophys Res Lett . 2012 ;39 : L01606 10.1029/2011GL049966 25 Smith JR , Fong P , Ambrose RF . Spatial patterns in recruitment and growth of the mussel Mytilus californianus (Conrad) in southern and northern California, USA, two regions with differing oceanographic conditions . J Sea Res . 2009 ;61 : 165 –173 . 10.1016/j.seares.2008.10.009 26 De Léo FC , Pires-Vanin AMS . Benthic megafauna communities under the influence of the South Atlantic Central Water intrusion onto the Brazilian SE shelf: a comparison between an upwelling and a non-upwelling ecosystem . J Mar Syst . 2006 ;60 : 268 –284 . 10.1016/j.jmarsys.2006.02.002 27 Humborg C , Conley DJ , Rahm L , Wulff F , Cociasu A , Ittekkot V . Silicon retention in river basins: far-reaching effects on biogeochemistry and aquatic food webs in coastal marine environments . AMBIO J Hum Environ . 2000 ;29 : 45 –50 . 10.1579/0044-7447-29.1.45 28 Martin-Jézéquel V , Hildebrand M , Brzezinski MA . Silicon metabolism in diatoms: implications for growth . J Phycol . 2000 ;36 : 821 –840 . 10.1046/j.1529-8817.2000.00019.x 29 Tilstone G , Míguez B , Figueiras F , Fermín E . Diatom dynamics in a coastal ecosystem affected by upwelling: coupling between species succession, circulation and biogeochemical processes . Mar Ecol Prog Ser . 2000 ;205 : 23 –41 . 10.3354/meps205023 30 Cury P , Bakun A , Crawford RJM , Jarre A , Quiñones RA , Shannon LJ , et al Small pelagics in upwelling systems: patterns of interaction and structural changes in “wasp-waist” ecosystems . ICES J Mar Sci J Cons . 2000 ;57 : 603 –618 . 10.1006/jmsc.2000.0712 31 Chavez FP , Messié M . A comparison of eastern boundary upwelling ecosystems . Prog Oceanogr . 2009 ;83 : 80 –96 . 10.1016/j.pocean.2009.07.032 32 Andrews WRH , Hutchings L . Upwelling in the Southern Benguela Current . Prog Oceanogr . 1980 ;9 : 1 –81 . 10.1016/0079-6611(80)90015-4 33 Verheye Dua F , Lucas MI . Southern Benguela frontal region. 1. hydrology, phytoplankton and bacterioplankton . Mar Ecol Prog Ser . 1988 ;47 : 271 –280 . 34 Shannon LV , Hutchings L , Bailey GW , Shelton PA . Spatial and temporal distribution of chlorophyll in southern African waters as deduced from ship and satellite measurements and their implications for pelagic fisheries . South Afr J Mar Sci . 1984 ;2 : 109 –130 . 10.2989/02577618409504363 35 Pirini M , Manuzzi MP , Pagliarani A , Trombetti F , Borgatti AR , Ventrella V . Changes in fatty acid composition of Mytilus galloprovincialis (Lmk) fed on microalgal and wheat germ diets . Comp Biochem Physiol B Biochem Mol Biol . 2007 ;147 : 616 –626 . 10.1016/j.cbpb.2007.04.003 17482494 36 Allan EL , Ambrose ST , Richoux NB , Froneman PW . Determining spatial changes in the diet of nearshore suspension-feeders along the South African coastline: Stable isotope and fatty acid signatures . Estuar Coast Shelf Sci . 2010 ;87 : 463 –471 . 10.1016/j.ecss.2010.02.004 37 Xavier BM , Branch GM , Wieters EA . Abundance, growth and recruitment of Mytilus galloprovincialis on the west coast of South Africa in relation to upwelling . Mar Ecol Prog Ser . 2007 ;346 : 189 –201 . 10.3354/meps07007 38 Lutjeharms JRE , Cooper J , Roberts M . Upwelling at the inshore edge of the Agulhas Current . Cont Shelf Res . 2000 ;20 : 737 –761 . 10.1016/S0278-4343(99)00092-8 39 Gorokhova E , Hansson S . An experimental study on variations in stable carbon and nitrogen isotope fractionation during growth of Mysis mixta and Neomysis integer . Can J Fish Aquat Sci . 1999 ;56 : 2203 –2210 . 10.1139/cjfas-56-11-2203 40 Davenport J , Chen X . A comparison of methods for the assessment of condition in the mussel (Mytilus edulis L.) . J Molluscan Stud . 1987 ;53 : 293 –297 . 41 Williams JR , Babcock RC . Assessment of size at maturity and gonad index methods for the scallop Pecten novaezelandiae . N Z J Mar Freshw Res . 2005 ;39 : 851 –864 . 42 Indarti E , Majid MIA , Hashim R , Chong A . Direct FAME synthesis for rapid total lipid analysis from fish oil and cod liver oil . J Food Compos Anal . 2005 ;18 : 161 –170 . 10.1016/j.jfca.2003.12.007 43 Ackman RG . The gas chromatograph in practical analyses of common and uncommon fatty acids for the 21st century . Anal Chim Acta . 2002 ;465 : 175 –192 . 10.1016/S0003-2670(02)00098-3 44 Budge SM , Iverson SJ , Koopman HN . Studying trophic ecology in marine ecosystems using fatty acids: a primer on analysis and interpretation . Mar Mammal Sci . 2006 ;22 : 759 –801 . 10.1111/j.1748-7692.2006.00079.x 45 Underwood AJ . Experiments in ecology: their logical design and interpretation using analysis of variance . Cambridge University Press ; 1997 . 46 Anderson MJ . A new method for non-parametric multivariate analysis of variance . Austral Ecol . 2001 ;26 : 32 –46 . 10.1111/j.1442-9993.2001.01070.pp.x 47 Anderson M , Braak CT . Permutation tests for multi-factorial analysis of variance . J Stat Comput Simul . 2003 ;73 : 85 –113 . 10.1080/00949650215733 48 Clarke KR , Gorley RN . Primer V6: user manual—tutorial . Plymouth Marine Laboratory ; 2006 . 49 Anderson M, Gorley R, Clarke K. PERMANOVA + for PRIMER: guide to software and statistical methods. PRIMER-E; 2008. 50 Budge SM , Parrish CC . Lipid biogeochemistry of plankton, settling matter and sediments in Trinity Bay, Newfoundland. II. Fatty acids . Org Geochem . 1998 ; 1547 –1559 . 10.1016/S0146-6380(98)00177-6 51 Dalsgaard J , St. John M , Kattner G , Müller-Navarra D , Hagen W . Fatty acid trophic markers in the pelagic marine environment Advances in Marine Biology . Academic Press ; 2003 pp. 225 –340 . Available: http://www.sciencedirect.com/science/article/pii/S0065288103460057 52 Meeuwis JM , Lutjeharms JRE . Surface thermal characteristics of the Angola-Benguela front . South Afr J Mar Sci . 1990 ;9 : 261 –279 . 10.2989/025776190784378772 53 Demarcq H . Trends in primary production, sea surface temperature and wind in upwelling systems (1998–2007) . Prog Oceanogr . 2009 ;83 : 376 –385 . 10.1016/j.pocean.2009.07.022 54 Pitcher GC , Brown PC , Mitchell-Innes BA . Spatio-temporal variability of phytoplankton in the southern Benguela upwelling system . South Afr J Mar Sci . 1992 ;12 : 439 –456 . 10.2989/02577619209504717 55 Shannon LV , Nelson G . The Benguela: large scale features and processes and system variability The South Atlantic . Springer Berlin Heidelberg ; 1996 pp. 163 –210 . Available: http://link.springer.com/chapter/10.1007/978-3-642-80353-6_9 56 Ezgeta-Balić D , Lojen S , Dolenec T , Žvab Rožič P , Dolenec M , Najdek M , et al Seasonal differences of stable isotope composition and lipid content in four bivalve species from the Adriatic Sea . Mar Biol Res . 2014 ;10 : 625 –634 . 10.1080/17451000.2013.833338 57 Basterretxea G , Arístegui J . Mesoscale variability in phytoplankton biomass distribution and photosynthetic parameters in the Canary-NW African coastal transition zone . Mar Ecol Prog Ser . 2000 ;197 : 27 –40 . 10.3354/meps197027 58 Suchanek TH , Geller JB , Kreiser BR , Mitton JB . Zoogeographic distributions of the sibling species Mytilus galloprovincialis and M. trossulus (Bivalvia: Mytilidae) and their hybrids in the North Pacific . Biol Bull . 1997 ;193 : 187 10.2307/1542764 28575597 59 Anestis A , Pörtner HO , Karagiannis D , Angelidis P , Staikou A , Michaelidis B . Response of Mytilus galloprovincialis (L.) to increasing seawater temperature and to marteliosis: metabolic and physiological parameters . Comp Biochem Physiol A Mol Integr Physiol . 2010 ;156 : 57 –66 . 10.1016/j.cbpa.2009.12.018 20045485 60 Tagliarolo M , McQuaid CD . Sub-lethal and sub-specific temperature effects are better predictors of mussel distribution than thermal tolerance . Mar Ecol Prog Ser . 2015 ;535 : 145 –159 . 10.3354/meps11434 61 Robinson T , Griffiths C , McQuaid C , Rius M . Marine alien species of South Africa—status and impacts . Afr J Mar Sci . 2005 ;27 : 297 –306 . 10.2989/18142320509504088 62 Fly EK , Hilbish TJ . Physiological energetics and biogeographic range limits of three congeneric mussel species . Oecologia . 2012 ;172 : 35 –46 . 10.1007/s00442-012-2486-6 23064978 63 Camacho AP , Labarta U , Beiras R . Growth of mussels (Mytilus edulis galloprovincialis) on cultivation rafts: influence of seed source, cultivation site and phytoplankton availability . Aquaculture . 1995 ;138 : 349 –362 . 10.1016/0044-8486(95)01139-0 64 Pampanin DM , Volpato E , Marangon I , Nasci C . Physiological measurements from native and transplanted mussel (Mytilus galloprovincialis) in the canals of Venice. Survival in air and condition index . Comp Biochem Physiol A Mol Integr Physiol . 2005 ;140 : 41 –52 . 10.1016/j.cbpb.2004.10.016 15664311 65 Field JG , Griffiths CL , Linley EA , Carter RA , Zoutendyk P . Upwelling in a nearshore marine ecosystem and its biological implications . Estuar Coast Mar Sci . 1980 ;11 : 133 –150 . 10.1016/S0302-3524(80)80037-5 66 Archambault P , McKindsey CW , Bourget E . Large-scale shoreline configuration influences phytoplankton concentration and mussel growth . Estuar Coast Shelf Sci . 1999 ;49 : 193 –208 . 10.1006/ecss.1999.0481 67 Giese AC , Hart MA , Smith AM , Cheung MA . Seasonal changes in body component indices and chemical composition in the pismo clam Tivela stultorum . Comp Biochem Physiol . 1967 ;22 : 549 –561 . 10.1016/0010-406X(67)90617-2 68 Maxwell MR , Hanlon RT . Female reproductive output in the squid Loligo pealeii: multiple egg clutches and implications for a spawning strategy . Mar Ecol Prog Ser . 2000 ;199 : 159 –170 . 69 van Erkom Schurink C , Griffiths C . A comparison of reproductive cycles and reproductive output in four southern African mussel species . Mar Ecol Prog Ser . 1991 ;76 : 123 –134 . 10.3354/meps076123 70 Porri F , McQuaid C , Radloff S . Spatio-temporal variability of larval abundance and settlement of Perna perna: differential delivery of mussels . Mar Ecol Prog Ser . 2006 ;315 : 141 –150 . 10.3354/meps315141 71 Alkanani T , Parrish CC , Thompson RJ , McKenzie CH . Role of fatty acids in cultured mussels, Mytilus edulis, grown in Notre Dame Bay, Newfoundland . J Exp Mar Biol Ecol . 2007 ;348 : 33 –45 . 10.1016/j.jembe.2007.02.017 72 Parrish CC , Abrajano TA , Budge SM , Helleur RJ , Hudson ED , Pulchan K , et al Lipid and phenolic biomarkers in marine ecosystems: analysis and applications In: Wangersky PJ , editor. Marine Chemistry . Springer Berlin Heidelberg ; 2000 pp. 193 –223 . Available: http://link.springer.com/chapter/10.1007/10683826_8 73 Sushchik NN . Role of essential fatty acids in trophometabolic interactions in the freshwater ecosystems (a review) . Zhurnal Obshcheĭ Biol . 2008 ;69 : 299 –316 . 74 Arts MT , Ackman RG , Holub BJ . “Essential fatty acids” in aquatic ecosystems: a crucial link between diet and human health and evolution . Can J Fish Aquat Sci . 2001 ;58 : 122 –137 . 10.1139/f00-224 75 Kelly JR , Scheibling RE . Fatty acids as dietary tracers in benthic food webs . Mar Ecol Prog Ser . 2012 ;446 : 1 –22 . 10.3354/meps09559 76 Pollero RJ , Irazú CE , Brenner RR . Effect of sexual stages on lipids and fatty acids of Diplodon delodontus . Comp Biochem Physiol Part B Comp Biochem . 1983 ;76 : 927 –931 . 10.1016/0305-0491(83)90414-5 77 Soudant P , Moal J , Marty Y , Samain JF . Impact of the quality of dietary fatty acids on metabolism and the composition of polar lipid classes in female gonads of Pecten maximus (L.) . J Exp Mar Biol Ecol . 1996 ;205 : 149 –163 . 10.1016/S0022-0981(96)02608-1 78 Kattner G , Hagen W . Lipids in marine copepods: latitudinal characteristics and perspective to global warming In: Kainz M , Brett MT , Arts MT , editors. Lipids in Aquatic Ecosystems . Springer New York ; 2009 pp. 257 –280 . Available: http://link.springer.com/chapter/10.1007/978-0-387-89366-2_11 79 Estefanell J , Socorro J , Izquierdo M , Roo J . Effect of two fresh diets and sexual maturation on the proximate and fatty acid profile of several tissues in Octopus vulgaris: specific retention of arachidonic acid in the gonads . Aquac Nutr . 2014 ;41 10.1111/anu.12163 80 Blanchard G , Druart X , Kestemont P . Lipid content and fatty acid composition of target tissues in wild Perca fluviatilis females in relation to hepatic status and gonad maturation . J Fish Biol . 2005 ;66 : 73 –85 . 10.1111/j.0022-1112.2005.00578.x 81 Besnard J-Y , Lubet P , Nouvelot A . Seasonal variations of the fatty acid content of the neutral lipids and phospholipids in the female gonad of Pecten maximus L . Comp Biochem Physiol Part B Comp Biochem . 1989 ;93 : 21 –26 . 10.1016/0305-0491(89)90210-1 82 Napolitano GE , Ackman RG , Silva-Serra MA . Incorporation of dietary sterols by the sea scallop Placopecten magellanicus (Gmelin) fed on microalgae . Mar Biol . 1993 ;117 : 647 –654 . 10.1007/BF00349777 83 Brown PC , Field JG . Factors limiting phytoplankton production in a nearshore upwelling area . J Plankton Res . 1986 ;8 : 55 –68 . 10.1093/plankt/8.1.55 84 Wieters EA , Kaplan DM , Navarrete SA , Sotomayor A , Largier JL , Nielsen K , et al Alongshore and temporal variability in chlorophyll a concentration in Chilean nearshore waters . Mar Ecol Prog Ser . 2003 ;249 : 93 –105 . 85 Roughan M , Mace AJ , Largier JL , Morgan SG , Fisher JL , Carter ML . Subsurface recirculation and larval retention in the lee of a small headland: A variation on the upwelling shadow theme . J Geophys Res Oceans . 2005 ;110 : 0 –27 . 10.1029/2005JC002898 86 Nielsen KJ , Navarrete SA . Mesoscale regulation comes from the bottom-up: intertidal interactions between consumers and upwelling . Ecol Lett . 2004 ;7 : 31 –41 . 10.1046/j.1461-0248.2003.00542.x 87 Farrell TM , Bracher D , Roughgarden J . Cross-shelf transport causes recruitment to intertidal populations in central California . Limnol Oceanogr . 1991 ;36 : 279 –288 . 10.4319/lo.1991.36.2.0279 88 Wing SR , Largier J , Botsford L , Quinn J . Settlement and transport of benthic invertebrates in an intermittent upwelling region . Limnol Oceanogr . 1995 ;40 : 316 . 89 Puccinelli E , Noyon M , McQuaid C . Hierarchical effects of biogeography and upwelling shape the dietary signatures of benthic filter feeders . Mar Ecol Prog Ser . 2016 ; 10.3354/meps11567 90 Shannon LV , Mostert SA , Walters NM , Anderson FP . Chlorophyll concentrations in the southern Benguela Current region as determined by satellite (Nimbus-7 coastal zone colour scanner) . J Plankton Res . 1983 ;5 : 565 –583 . 10.1093/plankt/5.4.565

PMC005xxxxxx/PMC5003372.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757130310.1371/journal.pone.0161961PONE-D-16-11619Research ArticleBiology and Life SciencesAnatomyHeadEyesMedicine and Health SciencesAnatomyHeadEyesBiology and Life SciencesAnatomyOcular SystemEyesMedicine and Health SciencesAnatomyOcular SystemEyesBiology and Life SciencesAnatomyOcular SystemOcular AnatomyOptic DiscMedicine and Health SciencesAnatomyOcular SystemOcular AnatomyOptic DiscMedicine and Health SciencesOphthalmologyVisual ImpairmentsMyopiaPhysical SciencesPhysicsClassical MechanicsDeformationPhysical SciencesPhysicsClassical MechanicsDamage MechanicsDeformationMedicine and Health SciencesOphthalmologyEye DiseasesGlaucomaBiology and Life SciencesAnatomyOcular SystemOcular AnatomyChoroidMedicine and Health SciencesAnatomyOcular SystemOcular AnatomyChoroidMedicine and Health SciencesOphthalmologyEye DiseasesResearch and Analysis MethodsMathematical and Statistical TechniquesStatistical MethodsRegression AnalysisLinear Regression AnalysisPhysical SciencesMathematicsStatistics (Mathematics)Statistical MethodsRegression AnalysisLinear Regression AnalysisAssociation of Myopic Optic Disc Deformation with Visual Field Defects in Paired Eyes with Open-Angle Glaucoma: A Cross-Sectional Study Myopic Optic Disc Deformation and Visual Field DefectSawada Yu 1*Hangai Masanori 2Ishikawa Makoto 1Yoshitomi Takeshi 11 Department of Ophthalmology, Akita University Graduate School of Medicine, Akita, Japan2 Department of Ophthalmology, Saitama Medical University, Saitama, JapanHejtmancik James Fielding EditorNational Eye Institute, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: YS. Data curation: YS MI TY. Formal analysis: YS. Investigation: YS MI TY. Methodology: YS MH. Project administration: YS. Resources: YS MI TY. Software: YS. Supervision: MH TY. Validation: YS TY. Visualization: YS. Writing – original draft: YS. Writing – review & editing: YS MH TY. * E-mail: sawadayu@doc.med.akita-u.ac.jp29 8 2016 2016 11 8 e016196121 3 2016 15 8 2016 © 2016 Sawada et al2016Sawada et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Purpose To examine the association of myopia with the visual field (VF) defects in open-angle glaucoma (OAG) using paired eyes to eliminate the effect of unknown confounding factors that are diverse among individuals. Methods One hundred eighteen eyes of 59 subjects with myopia (spherical equivalent [SE] ≥ -2 diopter [D] and axial length ≥ 24.0 mm) whose intra-ocular pressure between paired eyes was similar and the mean deviation (MD) of the Humphrey VF test differed by more than 6 dB were included. Refractive errors (SE, axial length) and parameters associated with the papillary and parapapillary myopic deformation (tilt ratio, torsion angle, and β-zone parapapillary atrophy [PPA] area without Bruch’s membrane) were measured in each eye. The paired eyes were divided into worse and better eyes according to the MD of the VF, and parameters were compared between them. Further, multiple linear regression analysis was performed to examine the correlation of the difference in various parameters with the MD difference between paired eyes. Results The SE of all eyes was -6.39 ± 2.15 D (mean ± standard deviation) and axial length was 26.42 ± 1.07 mm. MD of the worse and better VF eyes were -13.56 ± 6.65 dB and -4.87 ± 5.32 dB, respectively. Eyes with worse VFs had significantly greater SE, axial length, tilt ratio, and PPA area without Bruch’s membrane than those with better VFs (all P < 0.05). In multiple linear regression analysis, the difference of the MD between paired eyes was significantly correlated with the difference in the tilt ratio and PPA area without Bruch’s membrane. Conclusion The myopic papillary and parapapillary deformations, but not refractive error itself, were related to the worse VF in paired eyes with OAG. This suggests that myopia influences the severity of the glaucomatous VF defects via structural deformation. The authors received no specific funding for this work. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction A growing body of evidence supports the idea that myopia is an independent risk factor for the development of glaucoma. Epidemiological studies reported a high prevalence of myopia among glaucoma patients [1, 2]. A recent systematic review and meta-analysis demonstrated an odds ratio of 1.88 for all myopia and 1.77 for low myopia (spherical equivalent [SE] > -3 diopter [D]) in open-angle glaucoma (OAG) [3]. Myopic eyes present characteristic features of the optic disc region including tilt and torsion of the optic disc and parapapillary atrophy (PPA) [1, 4–7]. These morphological changes are considered to be associated with an increased susceptibility to the stress of glaucoma. Because glaucoma is likely to be a multifactorial disease and the extent that each factor contributes to its development is different in each subject, it is difficult to estimate the contribution of myopia alone. Thus, the influence of myopia on the glaucomatous damage remains unknown. In addition, the myopic deformation of the optic disc and surrounding region varies considerably among individuals even when the refractive errors are similar. Therefore, evaluating myopic deformation and its contribution to the glaucomatous damage among different individuals requires caution. Comparison of the paired eyes of the same individual has been used in clinical studies because it can eliminate the effects of factors that vary among individuals [8–10]. Within each individual, the paired eyes have the same variables of sex, age, inheritance, and systemic diseases including vascular conditions and diabetes mellitus. They also have the same cerebrospinal fluid pressure. The lamina cribrosas of the paired eyes are reported to have similar configuration, thickness, and material properties [11, 12], which indicate a similar susceptibility to glaucomatous stress. This similarity led us to hypothesize that if the degree of myopia is different in paired eyes of a glaucoma patient with asymmetric visual field (VF) damage, then the VF difference between eyes might be attributed to their myopic difference. The paired-eye study might be an effective method to evaluate the influence of myopia on glaucomatous damage. To our knowledge, there have not been any earlier studies that investigated the difference of myopia in paired eyes with asymmetric glaucoma. The purpose of this study was to determine the association of myopia with the VF defects in OAG using paired eyes to eliminate the effect of unknown confounding factors that are diverse among individuals. Materials and Methods This was a cross-sectional study that was approved by the Review Board of Akita University Graduate School of Medicine. All patients signed an informed consent before participation in the study, and all procedures were in accordance with the tenets of the Declaration of Helsinki. Study Subjects The patients were recruited from the outpatient clinic of Akita University Graduate School of Medicine from June 2011 to July 2015. Each patient underwent a comprehensive ophthalmic assessment, including refraction test, measurement of the best-corrected visual acuity, measurement of central corneal thickness (CCT) and axial length by ultrasound pachymetry (Tomey Corporation, Nagoya, Japan), Goldmann applanation tonometry, slit-lamp biomicroscopy, gonioscopy, dilated fundus stereoscopic examination, color fundus stereo photography (Canon, Tokyo, Japan), spectral domain-optical coherent tomography (SD-OCT) (Spectralis, Heidelberg Engineering GmbH, Heidelberg, Germany), and standard automated perimetry (Humphrey Field Analyzer II 750; 30–2 Swedish interactive threshold algorithm; Carl Zeiss Meditec, Dublin, CA, USA). The intraocular pressure (IOP) was evaluated as the untreated baseline and IOP at the VF examination. The untreated IOP was determined as the average of at least two measurements before the use of IOP lowering medications. The subject inclusion criteria were as follows: (1) OAG patients with an open iridocorneal angle, glaucomatous optic disc changes such as localized or diffuse rim thinning or retinal nerve fiber defects, and glaucomatous VF defects corresponding to the glaucomatous structural changes. Glaucomatous VF defects were defined as glaucoma hemifield test results outside normal limits, or the presence of at least 3 contiguous non-edge test points within the same hemifield on the pattern deviation plot at < 5%, with at least one of these points at < 1%, which was confirmed on two consecutive reliable tests (fixation loss rate, ≤ 20%; false-positive and false-negative error rates, ≤ 15%). (2) SE was ≥ -2 D and axial length was ≥ 24.0 mm. (3) To minimize the effect of media opacity, the corrected visual acuity was ≥ 20/30. (4) The difference between paired eyes in mean deviation (MD) of the Humphrey VF test was > 6 dB [13, 14]. (5) To ensure that the IOPs of the paired eyes were as close to identical as possible, the difference in the IOPs between paired eyes was ≤ 1 mmHg in untreated baseline and at the VF examination [9]. (6) To ensure that the CCTs of the paired eyes were as close to identical as possible, the difference in the CCT between paired eyes was ≤ 10 μm. Both eyes of each subject had to satisfy the criteria to be included in this paired-eye study. Subjects were excluded according to the following criteria: (1) Potential subjects with intraocular disease, ocular injury, previous intraocular surgery except uncomplicated cataract extraction and glaucoma surgery. (2) Congenital optic disc abnormalities. (3) Eyes with extremely high myopia due to the difficulty in identifying Bruch’s membrane (BM) opening in SD-OCT images and increased risk of myopic macular changes that might affect the VF. Measurement of the Optic Disc Tilt and Torsion Optic disc tilt and torsion were measured on stereo fundus photographs by two independent observers (Y.S. and T.Y.) who were masked to the patients’ clinical information. Tilt ratio was defined as the ratio between the longest and shortest diameters of the optic disc [6, 15]. Optic disc torsion was defined as the deviation of the long axis of the optic disc and the vertical meridian [15]. The vertical meridian was identified as a vertical line 90 degrees from a horizontal line connecting the fovea and the center of the optic disc. The angle between the long axis of the optic disc and vertical meridian was defined as the torsion angle [15]. A positive torsion value indicated an inferotemporal torsion, and a negative value indicated a superonasal torsion. The absolute values of the torsion angle were used in the analysis to avoid compensation of the positive and negative torsion values. Assessment of the PPA Area The PPA area was assessed in the infrared fundus image shown in the display window of the Spectralis viewer [16, 17]. This enabled us to observe the fundus image and the OCT scan image simultaneously, and pinpoint the location identified in the OCT image on the fundus image. Magnification difference was corrected by adjusting for the curvature of the cornea. Radial scan OCT was performed from the center of the optic disc and included 48 B-scan images, 3.8 degrees apart, in each eye. Each B-scan image was constructed from the 42 averaged frames. Before assessment, all infrared fundus images were compared with the stereo fundus photographs, and the identification of the β-zone PPA margin and optic disc margin in both images were confirmed. The margin of the β-zone PPA and optic disc were defined as the border between low and high reflectivity in the infrared fundus images. Images were magnified sufficiently to be able to define the borders. The margins were manually delineated using the built-in caliper tool of the OCT system [18]. The β-zone PPA area was defined as the area that subtracted disc area from the area that delineated β-zone PPA margin. The β-zone PPA area was divided into the area with Bruch’s membrane (BM) (PPA+BM) and that without BM (PPA-BM). The PPA-BM area was obtained as follows. The termination of the BM was identified in 12 equidistant radial SD-OCT images (Fig 1), and both sides of the termination were plotted on the scan line. Twenty-four BM terminations plotted on the 12 scan lines were manually delineated, and the extent was defined as the BM opening area. The PPA-BM area was obtained by subtracting the disc area from the BM opening area. The PPA+BM area was obtained by subtracting the PPA-BM area from the β-zone PPA area. When the image quality of the radial B-scan used for identifying BM termination was suboptimal, a neighboring image was used. When the quality of both neighboring images was suboptimal, the eye was excluded. 10.1371/journal.pone.0161961.g001Fig 1 Delineation of Bruch’s membrane (BM) opening area. The central figure is an infrared fundus image of an optic disc with 12 equidistant radial SD-OCT scans. SD-OCT images are presented around the fundus image and placed besides the scan lines that have the identical numbers. The termination of BM was detected in each OCT images (yellow dots), and both sides of the termination were plotted on the scan line. Twenty-four BM terminations plotted on the 12 scan lines were manually delineated, and the extent was defined as the BM opening area (yellow line). Data Analysis The paired eyes were divided into worse and better eyes according to the MDs of the Humphrey VF test performed within three months of the OCT examination. The comparison was performed between worse and better VF eyes for the refractive errors (i.e., SE and axial length) and parameters associated with the myopic deformation of the optic disc region (i.e., tilt ratio, torsion angle, and PPA-BM area). Comparisons were performed using paired t tests. Multiple linear regression analysis was performed to examine the correlation between paired-eye differences in MDs and various parameters. Inter-observer reproducibility of the measurement of tilt ratio, torsion angle, β-zone PPA area, and PPA-BM area was assessed in 30 randomly selected eyes, and the intraclass correlation coefficients (ICC) with 95% confidence intervals (CI) were calculated. The level of significance was set at P < 0.05. The statistical analyses, with two-sided P-values, were performed using SPBS ver. 9.66 [19]. Results Among 129 subjects enrolled, the followings were excluded (some for multiple reasons): IOP difference between paired eyes more than 1 mmHg (n = 39), VF difference between eyes less than 6 dB (n = 16), CCT difference between eyes more than 10 μm (n = 11), unreliable VF test (n = 14), unclear OCT images (n = 6), congenital optic nerve abnormalities (n = 5), and epiretinal membrane affecting the VF (n = 3). The remaining 59 subjects with 118 eyes were included in the analysis. All of the subjects were Japanese. The ICCs for the measurement of the tilt ratio and torsion angle were 0.962 (95% CI, 0.933–0.992) and 0.956 (95% CI, 0.936–0.988) respectively. The ICCs for the measurement of the β-zone PPA area and PPA-BM area were 0.970 (95% CI, 0.950–0.982) and 0.946 (95% CI, 0.920–0.968) respectively. For all eyes, the mean SE was -6.39 ± 2.15 D, and the axial length was 26.42 ± 1.07 mm (Table 1). The mean MD was -9.22 ± 7.38 dB. The measured parameters were compared between paired eyes with worse and better VFs (Table 1). The IOP parameters and CCT were similar between two groups. The MD was -13.56 ± 6.65 dB in the eyes with worse VFs and -4.87 ± 5.32 dB in the eyes with better VFs (P < 0.0001). In the eyes with worse VFs, the SE was greater, and the axial length was longer in the eyes with worse VFs than for eyes with better VFs (P = 0.0085 and 0.0262, respectively). The tilt ratio was higher in the eyes with worse VFs (P < 0.0001), while torsion angle did not differ between the two groups. The PPA-BM area and PPA+BM area were greater in the eyes with worse VFs than in the better eyes (both P < 0.0001). 10.1371/journal.pone.0161961.t001Table 1 Subject Demographics and Comparison of Parameters between Eyes with Worse and Better Visual Fields. Comparison between Eyes with Worse and Better VF Parameters All Eyes (n = 118) Worse VF (n = 59) Better VF (n = 59) P Value Sex (male/female) 37/22 Age (yrs) 54.5±13.6 IOP Untreated (mmHg) 19.3±3.3 19.4±3.3 19.3±3.4 0.4179     VF examination (mmHg) 14.3±2.0 14.4±2.1 14.3±2.0 0.2548 CCT (μm) 522.7±28.4 522.7±28.7 522.8±28.7 0.9408 Mean deviation (dB) -9.22±7.38 -13.56±6.65 -4.87±5.32 <0.0001 Pattern standard deviation (dB) 10.00±4.81 12.52±3.40 7.47±4.73 <0.0001 Spherical equivalent (Diopter) -6.39±2.15 -6.62±2.18 -6.12±2.12 0.0085 Axial length (mm) 26.42±1.07 26.50±1.11 26.33±1.05 0.0262 Tilt ratio 1.33±0.26 1.38±0.30 1.28±0.21 <0.0001 Torsion angle (degree) 10.1±5.3 10.6±5.8 9.5±4.8 0.1156 PPA area without BM (mm2) 0.60±0.63 0.70±0.67 0.50±0.58 <0.0001 PPA area with BM (mm2) 1.53±0.68 1.67±0.72 1.39±.0.62 <0.0001 Statistical analysis was performed using paired t test. Data are shown as mean ± standard deviation. Statistically significant values are shown in bold. VF = visual field; IOP = intraocular pressure; CCT = central corneal thickness; PPA = parapapillary atrophy; BM = Burch’s membrane. In the multiple linear regression analysis, the differences in tilt ratio, PPA-BM area, and PPA+BM area were significantly correlated with the MD differences between paired eyes (P < 0.05 for tilt ratio and PPA-BM area, P < 0.01 for PPA+BM area), while the same correlation was not found for the axial length (Table 2). 10.1371/journal.pone.0161961.t002Table 2 Multiple Linear Regression Analysis with Difference of MD between Paired Eyes as Dependent Variable. Parameters Model 1 Model 2 Model 3 Model 4 IOP untreated -0.0653 -0.0486 -0.0548 -0.0393     VF examination -0.0221 -0.0583 -0.0391 -0.0029 CCT -0.1259 -0.1344 -0.1381 -0.1044 Axial length 0.0487 Tilt ratio 0.2703* Torsion angle 0.1431 PPA area without BM 0.2617* PPA area with BM 0.4066† 0.3583† 0.4081† 0.3766† Adjusted R 0.3545 0.4460 0.3816 0.4408 P value 0.0319 0.0045 0.0193 0.0052 Independent variables are the differences of the parameters between paired eyes. Values are the standard regression coefficient. Axial length, tilt ratio, torsion angle, PPA area without BM were included in the separate models because of the high correlation among them. Statistically significant values are shown in bold. * P < 0.05 † P <0.01 Adjusted multiple correlation coefficients (R) are shown in each model with P values. MD = mean deviation; IOP = intraocular pressure; VF = visual field; CCT = central corneal thickness; PPA = parapapillary atrophy; BM = Bruch’s membrane. A representative case was shown in Fig 2. In the eyes of a 47-year-old female OAG patient, the myopic refractive errors and myopic deformation of the optic disc region were greater in the right eye with worse VF than in the left eye with better VF. 10.1371/journal.pone.0161961.g002Fig 2 A representative case of myopic open-angle glaucoma patient with asymmetric visual fields (VF). In this 47-year-old female, the right eye with the worse VF had greater myopic refractive errors, tilt ratio, and PPA-BM area than the left eye with the better VF. Humphrey VF data are shown in the gray scale and pattern deviation plot. Solid blue lines are the vertical meridians that are 90 degrees from the line connecting the fovea to the center of the optic disc (dotted blue lines). Yellow dotted lines are the BM openings. SE = spherical equivalent; MD = mean deviation; PSD = pattern standard deviation. Discussion We evaluated differences in myopic parameters between paired eyes in OAG patients with asymmetric VF defects. The eyes with worse VFs had greater myopic refractive errors and optic disc deformations than those with better VFs based on univariate analysis. Based on multiple linear regression analyses, MD difference between paired eyes was correlated with the differences in tilt ratio, PPA-BM area, and PPA+BM area. Tilt ratio and PPA-BM area are the parameters that present myopic deformation of the optic disc region [7, 20]. The current study, which demonstrated association of tilt ratio and PPA-BM area with the worse VFs, suggested that the susceptibility to the glaucomatous stress was greater in the eyes with severer myopic deformations. The PPA+BM area is speculated to be associated with glaucoma [20]. Our results that showed association between larger PPA+BM area and worse VF defects supports this speculation. In this study, we used the paired-eye comparisons to estimate the influence of myopia on the glaucomatous damage. This approach minimized the effects of other factors diverse among individuals. Having equal IOP between eyes is particularly important for this purpose because IOP is known to play a role in the development and progression of glaucoma [8, 9, 14]. As used in a previous report, we employed the strictest definition of equal IOP between eyes, in which the IOP difference was ≤ 1 mmHg [9]. This definition was employed for both untreated baseline and IOP at the VF examination to eliminate the effect of IOP as much as possible. CCT is another factor that could affect the development of glaucoma [21]; therefore, we included only paired eyes with similar CCTs. Myopia is accepted as one of the risk factors for the development of glaucoma [1–3], while its influence on the progression of glaucoma remains controversial [22–24]. A recent meta-analysis summarized studies that investigated prognostic factors of OAG and reported that myopic refractive error was not likely to be associated with the VF progression [24]. We reviewed previous studies that reported the absence of association between myopia and VF progression, and noticed that these studies employed refractive errors (i.e., SE or axial length) as myopic parameters [22, 23]. Although our study investigated the severity of the VF defect, our findings were compatible with these studies in that the refractive error was not associated with the VF status. The differences in refractive errors were not correlated with the differences in the MDs between paired eyes. Instead, the differences in the parameters that presented myopic deformations (e.g., tilt ratio and PPA-BM area) were significantly correlated with the MD difference. This implies that not refractive error itself, but papillary and parapapillary myopic deformations may influence the severity of the VF in OAG. These findings let us hypothesize two possible processes that caused VF differences between the paired eyes in this study. In the first hypothesis, the eyes with greater myopic deformation developed more VF damage during the childhood progression of myopia, and the VF difference remained after the cessation of the myopic progression. A recent study by Kim et al. demonstrated progressive optic disc tilting with the enlargement of PPA in children who exhibited myopic shift [7]. The structural parameters that were altered during the myopic shift in their study were exactly the same as the ones associated with worse MD in the current study. Doshi et al. reported on a group of young Chinese males who presented non-progressive glaucomatous damage regardless of the use of IOP lowering therapy [25]. The majority of their subjects was myopic and had tilted discs. They speculated that the tilting and associated deformation of the optic discs might develop during the progression of myopia, and these changes could make eyes susceptible to the axonal loss characteristic of the glaucomatous optic neuropathy. Once myopic progression halted and the scleral tension decreased, then the glaucomatous damage would stabilize. In the second hypothesis, the VF damage progressed after the initial damage created during the development of myopia, and the progression speed was affected by the degree of the myopic deformation of the optic disc region. The parapapillary sclera is considered to be the main stress-bearing tissue of the globe [26]. When optic disc tilt and myopic deformation of the surrounding region develops, it may alter the biomechanics of the parapapillary sclera [27]. With this change, the stretched and flattened temporal part of the optic disc may become the focus of the glaucomatous stress at any level of IOP. In the eyes having a glaucomatous pathology, the optic discs with greater myopic deformation might gain a greater susceptibility to the glaucomatous stress, and it could accelerate the axonal loss and consequent VF damage. It may be more practical to think that the second hypothesis applies to a certain subset of eyes because the VF defect severity and its difference between paired eyes were significant in this study (mean MD of the worse eyes was -13.56 dB and better eyes was -4.87 dB). These differences are not likely to be created only during the development of myopia in childhood. Further longitudinal studies are required to test these hypotheses. The current study has several limitations. The clinically perceived optic disc margin is not a homogeneous structure because it is composed of some aspect of Bruch’s membrane and the border tissue external to the Bruch’s membrane opening [28]. This might have affected the accuracy of the measurement of the disc area, and subsequent calculation of the PPA-BM area. Also, we employed strict inclusion criteria to assess the influence of myopia on the VF severity, including similar IOPs, CCTs, and asymmetric VFs between paired eyes. This might have excluded a certain portion of potential subjects, and it made the sample size smaller. Further study is needed to examine whether or not the findings are applicable to the general population. In conclusion, our study showed that of the two eyes of OAG patients, the eyes with greater degrees of myopic deformation presented worse VFs than the eyes with less deformation. The paired-eye design of this study helped rule out the effect of various confounding factors. This finding suggests that myopia has an influence on the VF damage in OAG eyes by way of papillary and parapapillary structural deformation. Supporting Information S1 File Data of paired eyes with better and worse visual fields. Data of the paired eyes with better and worse visual fields of each patient are presented. (CSV) Click here for additional data file. The authors would like to thank Katsuyuki Murata and Toyoto Iwata, Department of Environmental Health Sciences, Akita University Graduate School of Medicine, for their valuable assistance with statistic computing. ==== Refs References 1 Mitchell P , Hourihan F , Sandbach J , Wang JJ (1999 ) The relationship between glaucoma and myopia: the Blue Mountain Eye Study . Ophthalmology 106 :2010 –2015 . 10519600 2 Suzuki Y , Iwase A , Araie M , Yamamoto T , Abe H , Shirato S , et al (2006 ) Risk factors for open-angle glaucoma in a Japanese population . Ophthalmology 113 :1613 –1617 . 16828504 3 Marcus MW , Vries MM , Montolio FGJ , Jansonius NM (2011 ) Myopia as a risk factor for open-angle glaucoma: A systematic review and meta-analysis . Ophthalmology 118 :1989 –1994 . 10.1016/j.ophtha.2011.03.012 21684603 4 Jonas JB , Gusek GC , Naumann GO . (1988 ) Optic disc morphometry in high myopia . Graefes Arch Clin Exp Ophthalmol 226 :587 –590 . 3209086 5 Vongphanit J , Mitchell P , Wang JJ (2002 ) Population prevalence of tilted optic disks and the relationship of this sign to refractive error . Am J Ophthalmol 133 :679 –685 . 11992866 6 Samarawickrama C , Mitchell P , Tong L , Gazzard G , Lim L , Wong TY , et al (2011 ) Myopia-related optic disc and retinal changes in adolescent children from Singapore . Ophthalmology 118 :2050 –2057 . 10.1016/j.ophtha.2011.02.040 21820741 7 Kim TW , Kim M , Weinreb RN , Woo SJ , Park KH , Hwang JM (2011 2) Optic disc change with incipient myopia of childhood . Ophthalmology 119 :21 –26 . 10.1016/j.ophtha.2011.07.051 21978594 8 Cartwright MJ , Anderson DR (1988 ) Correlation of asymmetric damage with asymmetric intraocular pressure in normal-tension glaucoma (low-tension glaucoma) . Arch Ophthalmol 106 :898 –900 . 3390051 9 Crichton A , Drance SM , Douglas GR , Schulzer M (1989 ) Unequal intraocular pressure and its relation to asymmetric visual field defects in low-tension glaucoma . Ophthalmology 96 :1312 –1314 . 2779999 10 Nicolela MT , Drance SM , Rankin SJA , Buckley AR , Walman BE (1996 ) Color Doppler imaging in patients with asymmetric glaucoma and unilateral visual field loss . Am J Ophthalmol 121 :502 –510 . 8610793 11 Sawada Y , Ishikawa M , Sato N , Yoshitomi T (2011 ) Optic nerve head morphology assessed by laser scanning tomography in normal Japanese subjects . J Glaucoma 20 :445 –451 . 10.1097/IJG.0b013e3181f3eb20 20852439 12 Seo JH , Kim TW , Weinreb RN (2014 ) Lamina cribrosa depth in healthy eyes . Invest Ophthalmol Vis Sci 55 :1241 –1250 . 10.1167/iovs.13-12536 24474269 13 Plange N , Kaup M , Arend O , Remky A (2006 ) Asymmetric visual field loss and retrobulbar haemodynamics in primary open-angle glaucoma . Graefe’s Arch Clin Exp Ophthalmol 244 :978 –983 . 16411100 14 Haefliger IO , Hitchings RA (1990 ) Relationship between asymmetry of visual field defects and intraocular pressure difference in an untreated normal (low) tension glaucoma population . Acta Ophthalmol 68 :564 –567 . 2275352 15 Park HYL , Lee K , Park CK (2012 ) Optic disc torsion direction predicts the location of glaucomatous damage in normal-tension glaucoma patients with myopia . Ophthalmology 119 :1844 –1851 . 10.1016/j.ophtha.2012.03.006 22595297 16 Hassenstein A , Meyer CH (2009 ) Clinical use and research applications of Heidelberg retina angiography and spectral-domain optical coherence tomography–a review . Clin Experiment Ophthalmol 37 :130 –143 .19338610 17 Kim M , Kim TW , Weinreb RN , Lee EJ (2013 ) Differentiation of parapapillary atrophy using spectral-domain optical coherence tomography . Ophthalmology 120 :1790 –1797 . 10.1016/j.ophtha.2013.02.011 23672970 18 Hayashi K , Tomidokoro A , Lee KY , Konno S , Saito H , Mayama C , et al (2012 ) Spectral-domain optical coherence tomography of β-zone peripapillary atrophy: Influence of myopia and glaucoma . Invest Ophthalmol Vis Sci 53 :1499 –1505 . 10.1167/iovs.11-8572 22323471 19 Murata K , Yano E (2002 ) Medical Statistics for Evidence-Based Medicine with SPBS Users' Guide [in Japanese] . Tokyo : Nankodo Publishers :1 –198 . 20 Jonas JB , Jonas SB , Jonas RA , Holbach L , Dai Y , Sun X , et al (2012 ) Parapapillary atrophy: histological gamma zone and delta zone . PLOS One 7 : e47237 10.1371/journal.pone.0047237 23094040 21 Dueker DK , Singh K , Lin SC , Fechtner RD , Minckler DS , Samples JR , et al (2007 ) Corneal thickness measurement in the management of primary open-angle glaucoma: a report by the American Academy of Ophthalmology . Ophthalmology 114 :1779 –1787 . 17822980 22 Sohn SW , Song JS , Kee C (2010 ) Influence of the extent of myopia on the progression of normal-tension glaucoma . Am J Ophthalmol 149 :831 –838 . 10.1016/j.ajo.2009.12.033 20231010 23 Lee JY , Sung KR , Han S , Na JH (2015 ) Effect of myopia on the progression of primary open-angle glaucoma . Invest Ophthalmol Vis Sci 56 :1775 –1781 . 10.1167/iovs.14-16002 25698704 24 Ernest PJ , Schouten JS , Beckers HJ , Hendrikse F , Prins MH , Webers CA (2013 ) An evidence-based review of prognostic factors for glaucomatous visual field progression . Ophthalmology 120 :512 –519 . 10.1016/j.ophtha.2012.09.005 23211636 25 Doshi A , Kreidl KO , Lombardi L , Sakamoto DK , Singh K (2007 ) Nonprogressive glaucomatous cupping and visual field abnormalities in young Chinese males . Ophthalmology 114 :472 –479 . 17123617 26 Fazio MA , Grytz R , Bruno L , Girard MJ , Gardiner S , Girkin CA , et al (2012 ) Regional variations in mechanical strain in the posterior human sclera . Invest Ophthalmol Vis Sci 53 :5326 –5233 . 10.1167/iovs.12-9668 22700704 27 Coudrillier B , Tian J , Alexander S , Myers KM , Quingley HA , Nguyen TD (2012 ) Biomechanics of the human posterior sclera: Age- and glaucoma-related changes measured using inflation testing . Invest Ophthalmol Vis Sci 53 :1714 –1728 . 10.1167/iovs.11-8009 22395883 28 Reis ASC , Sharpe GP , Yang H , Nicolela MT , Burgoyne CF , Chauhan BC (2012 ) Optic disc margin anatomy in patients with glaucoma and normal controls with spectral domain optical coherence tomography . Ophthalmology 119 :738 –747 . 10.1016/j.ophtha.2011.09.054 22222150

PMC005xxxxxx/PMC5003373.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757141710.1371/journal.pone.0161339PONE-D-16-16859Research ArticleBiology and life sciencesOrganismsVirusesDNA virusesPapillomavirusesHuman PapillomavirusBiology and Life SciencesMicrobiologyMedical MicrobiologyMicrobial PathogensViral PathogensPapillomavirusesHuman PapillomavirusMedicine and Health SciencesPathology and Laboratory MedicinePathogensMicrobial PathogensViral PathogensPapillomavirusesHuman PapillomavirusBiology and Life SciencesOrganismsVirusesViral PathogensPapillomavirusesHuman PapillomavirusBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesArtificial Gene Amplification and ExtensionPolymerase Chain ReactionResearch and Analysis MethodsMolecular Biology TechniquesArtificial Gene Amplification and ExtensionPolymerase Chain ReactionResearch and Analysis MethodsMathematical and Statistical TechniquesStatistical MethodsMeta-AnalysisPhysical SciencesMathematicsStatistics (Mathematics)Statistical MethodsMeta-AnalysisBiology and life sciencesOrganismsVirusesDNA virusesPapillomavirusesHPV-16Biology and Life SciencesMicrobiologyMedical MicrobiologyMicrobial PathogensViral PathogensPapillomavirusesHPV-16Medicine and Health SciencesPathology and Laboratory MedicinePathogensMicrobial PathogensViral PathogensPapillomavirusesHPV-16Biology and Life SciencesOrganismsVirusesViral PathogensPapillomavirusesHPV-16Medicine and Health SciencesUrologyGenitourinary InfectionsHuman Papillomavirus InfectionMedicine and Health SciencesInfectious DiseasesSexually Transmitted DiseasesHuman Papillomavirus InfectionMedicine and Health SciencesInfectious DiseasesViral DiseasesHuman Papillomavirus InfectionResearch and Analysis MethodsResearch DesignCase-Control StudiesPeople and PlacesGeographical LocationsAsiaIndiaPeople and PlacesGeographical LocationsAsiaThe Magnitude of the Association between Human Papillomavirus and Oral Lichen Planus: A Meta-Analysis The Association between HPV and OLPMa Junxian 1Zhang Jinshan 2http://orcid.org/0000-0003-1857-9623Zhang Yan 1*Lv Tingting 1Liu Jie 11 Department of Rheumatology and Immunology, Tangdu hospital, The Fourth Military Medical University, Xi'an, China2 Department of Human Anatomy and Histology and Embryology, The Fourth Military Medical University, Xi'an, ChinaMeyers Craig EditorPenn State University School of Medicine, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: YZ Data curation: JM JZ Formal analysis: JM JL TL Funding acquisition: YZ JZ Methodology: JZ Software: JZ Supervision: YZ JZ Validation: YZ Writing – original draft: JM JZ Writing – review & editing: YZ JZ. * E-mail: zyhphc@fmmu.edu.cn29 8 2016 2016 11 8 e016133926 4 2016 3 8 2016 © 2016 Ma et al2016Ma et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Background The role of human papilloma virus (HPV) in oral lichen planus (OLP) is controversial. Objectives The primary aim of the current study is to calculate the pooled risk estimates of HPV infection in OLP when compared with healthy controls. Methods Bibliographic searches were conducted in three electronic databases. Articles on the association between HPV and OLP were selected from case-control studies or cross-sectional studies, following predefined criteria. Pooled data were analyzed by calculating odds ratios (OR) and 95% confidence interval (CI). Results Of the 233 publications identified, 22 case-control studies met the inclusion criteria. Collectively, 835 cases and 734 controls were available for analysis. The summary estimate showed that OLP patients have significantly higher HPV prevalence (OR: 6.83; 95% CI: 4.15–11.27) than healthy controls. In subgroup analyses, the association of HPV and OLP varied significantly by geographic populations. The ORs ranged from 2.43 to 132.04. The correlation of HPV and erosive-atrophic oral lichen planus (EA-OLP) (OR: 9.34) was comparable and well above that of HPV and non-EA-OLP (OR: 4.32). Among HPV genotypes, HPV 16 showed an extremely strong association with OLP (OR: 11.27), and HPV 18 showed a relatively strong one (OR: 6.54). Conclusion In conclusion, a significant association was found between HPV and OLP. The strength of the association varied across geographic populations, clinical types of OLP, and HPV genotypes. The results suggest that HPV might play an important causal role in OLP and in its malignant to progression. http://dx.doi.org/10.13039/501100001809National Natural Science Foundation of China81273280http://orcid.org/0000-0003-1857-9623Zhang Yan http://dx.doi.org/10.13039/501100001809National Natural Science Foundation of China81272176Zhang Jinshan This work was supported by the National Natural Science Fund project of China, no. 81273280 and 81272176 (http://www.nsfc.gov.cn/). Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Oral lichen planus (OLP) is a common chronic autoimmune disorder, which may present epithelial thickening or atrophy with or without ulceration [1]. Clinically, there are six different types: papule, reticular, plaque, atrophic, erosive and bullous. The most common type is reticular. All types of OLP can be pooled in 2 clinical groups: erosive-atrophic forms (EA-OLP), including erosive, atrophic, bullous and mixed EA variants; and non-erosive-atrophic forms (non-EA-OLP), involved papule, reticular, plaque and mixed non-EA variants. EA-OLP is more prone to malignant transformation than non-EA-OLP [2]. The prevalence of HPV in OLP has been reported to range from 0.5 to 2.2%, varying according to geographic location [3]. Although the etiology of OLP is still unknown, it is generally accepted to be a T-cell-mediated inflammatory disease [1]. The reaction of these specific CD8+ T cells is similar to what occurs during a viral infection, in which a virus can act as a cytoplasmic antigen or induce the expression of host cell proteins, resulting in a differing host cell protein profile [4]. In this way, exploring the possibility of viral involvement in the pathogenesis of OLP is improving. As early as 1987, the association between HPV and OLP was reported [5]. In one study published prior to 1998, 107 OLP samples were studied, and 23% were HPV-positive. The first three most prevalent HPV types were HPV 6, 11, and 16. Then 1929 normal oral mucosa samples were tested for HPV DNA, and 11% were found to be positive [6]. The rate of HPV in OLP was twice as high as in normal cases. On the basis of these studies, the case-control studies of HPV and OLP correlation were increasingly growing. However, results have been conflicting: some studies seemed to show that HPV did play an important role in OLP [7–10], but others disagreed [11–13]. There have been some reviews of the HPV and OLP, but most of them are qualitative analyses. Very few reviews have included quantitative analysis. There was one meta-analysis of HPV in oral carcinoma and oral potentially malignant disorders (OPMD). OLP was a sub-analysis of OMPD, and the OR of the link between HPV and OLP was 5.12 [14]. The results indicated a strong association, but the articles included were limited, a more comprehensive meta-analysis is needed. Two English databases (PubMed and Web of Science) and one Chinese database (CNKI) were screened in order to collect more articles. Finally, 22 case-control studies were included to perform this systematic review and meta-analysis. The primary aim of the current paper was to clarify the magnitude of the association between HPV and OLP, and to get a further understanding of it concerning with different geographical populations, different OLP clinical types and HPV genotypes. Materials and Methods Data sources A systematic search was performed in PubMed (www.ncbi.nlm.nih.gov/pubmed), Web of Science (SCI) (http://webofknowledge.com/) and CNKI (http://epub.cnki.net/) to screen relevant literature, until the last search update on September 21, 2015. The search terms were (lichen planus or oral lichen planus or LP or OLP) and (human papillomavirus or HPV). The reference lists of selected papers were searched to identify additional studies. Data extraction The screening process was carried out by two investigators (J. Ma and J. Zhang) independently under the same criteria (Fig 1). Disagreements about eligibility were settled by consensus with a third investigator (Y. Zhang). The following details were recorded from each study: first author, publication date, country, clinical type of OLP, number of OLP patients and healthy controls, test methods, and HPV genotypes (Table 1). 10.1371/journal.pone.0161339.g001Fig 1 Flow chart of literature searches for this meta-analysis. 10.1371/journal.pone.0161339.t001Table 1 Characteristics of the 22 included studies in this meta-analysis. Reference Nation Clinical type OLP(n/N) Control(n/N) Detection method HPV genotypes Pol CA et al., 2015 India 21/30 0/30 IHC HPV16 Arirachakaran P et al., 2013 Thailand 1/37 0/37 PCR NA Yildirim B et al., 2011 Turkey 14/65 0/15 IHC HPV16 Debanth S et al., 2009 India 6/6 3/35 PCR HPV16, 18, 31, 33, 35, 39, 45, 51,52, 56, 58, 59,68 Fehér E et al., 2009 Hungary 39/119 3/72 PCR HPV6.11.16.18.31.33 Razavi SM et al., 2009 Iran 9/29 1/14 PCR HPV18 Szarka K et al., 2009 Hungary 39/119 3/72 PCR NA EA-OLP 26/61 3/72 non-EA-OLP 13/58 3/72 Yu Hong et al., 2007 China 33/57 3/20 PCR HPV16 Cianfriglia F et al., 2006 Italy 3/15 2/10 ISH NA Giovannelli L et al., 2006 Italy 12/49 11/49 PCR NA Campisi G et al., 2004 Italy 14/71 5/90 PCR HPV16.18.31.6 EA-OLP 9/44 5/90 non-EA-OLP 5/27 5/90 Ma Jian et al., 2003 China 26/30 5/18 PCR NA OFlatharta C et al., 2003 Ireland 10/38 0/20 PCR HPV16 Giovannelli L et al., 2002 Italy 9/34 5/90 PCR NA Li Hui et al., 2000 China 9/30 3/40 PCR NA Sand L et al., 2000 Sweden 6/22 0/12 NA NA Qi Yanchun et al., 1999 China 9/30 3/40 PCR NA Lei Lei et al., 1997 China 9/22 2/10 PCR NA EA-OLP 3/4 2/10 non-EA-OLP 6/18 2/10 Vespe M et al., 1997 Germany 3/7 0/33 PCR HPV16.18.31 Boyd AS et al., 1996 USA 11/13 0/10 PCR NA Cox M et al., 1993 UK 3/4 3/5 PCR HPV16 Maitland NJ et al., 1987 UK 7/8 5/12 PCR HPV16 n: numbers of HPV positive subjects; N: numbers of total subjects; IHC: immunohistochemical staining; PCR: polymerase chain reaction; ISH: in situ hybridization; NA: not available; EA-OLP: erosive-atrophic oral lichen planus; non-EA-OLP: non-erosive-atrophic oral lichen planus. Selection criteria Studies included in the meta-analysis had to meet the following criteria: (i) address the relationship between HPV and OLP; (ii) full-text papers; and (iii) original case-control or cross-sectional studies. Studies were excluded if they included any of the following: (i) reviews; (ii) incomplete data; and (iii) republished articles or reused data. Quality assessment Results of observational studies are greatly influenced by the design of the research. To guarantee the quality of the meta-analysis, each included paper was carefully assessed using the standard proposed by Chalmers et al. [15], including selection of subjects, comparability between groups and outcome presented. The selected studies were scored on an ordinal star scale from 1 to 9, with higher scores representing higher quality. Statistical analysis The meta-analysis was conducted using RevMan5, a copyrighted freeware developed by the Cochrane Collaboration, for preparing and maintaining reviews (http://www.cochrane-net.org/revman). Heterogeneity among different studies was measured by calculating χ2 and I2, if P < 0.05 and I2 > 50% were considered statistically significant. The primary outcome was reported as pooled odds ratios (OR) with 95% confidence intervals (CI), calculated using the random-effects model when heterogeneity existed or the fixed-effects model when no heterogeneity was found. To investigate potential for publication bias, the funnel plot asymmetry of the included studies was evaluated. Subgroup analysis was undertaken for geographic differences, clinical types of OLP, and HPV genotypes. Sensitivity analysis Low-quality studies (five stars or fewer) were excluded to investigate potential selection bias. Articles with different methods of HPV detection were analyzed to assess detection bias. Results Overall information on the studies included The process of search strategy is showed in Fig 1. From the 233 articles identified through an initial research, 144 were excluded for lack of relevance to OLP. After further reading, 15 papers were excluded for reviews, 25 papers were excluded without case-control or cross-sectional design, and 16 studies with insufficient data were excluded. Eleven papers were both collected in PubMed and SCI database. In the end, 22 articles met the criteria. The main information of the 22 selected studies is listed in Table 1. The analysis covered 12 countries, 5 of Asia, specifically China [16–20], India [9, 10], Iran [21], Thailand [22], and Turkey [23]; 6 in Europe, specifically Italy [12, 13, 24, 25], Hungary [26, 27], England [5, 11], Germany [8], Ireland [28], and Sweden [29]; 1 in North American, specifically the United States [7]. Results of quality assessment On the basis of the criteria described previously [15], 20 studies were evaluated as superior quality, with scores of six or more stars, and 2 studies were of low quality, with five stars or fewer (Table 2). Nine articles lacked a definition or diagnosis of OLP [5, 7, 11, 12, 16–20]. Three papers did not clearly explain the selection criteria for the controls [8, 11, 25], and 1 paper used diseased patients as controls [7]. Seventeen papers either did not give information about the sex of subjects or the sexes between cases and controls were incomparable [5, 7, 8, 10–13, 17–21, 23, 24, 26, 28, 29]. 10.1371/journal.pone.0161339.t002Table 2 Assessment of quality and score for 22 studies included. Reference Selection of subjects Comparability Outcome Score (1) (2) (3) Age Sex Living area Race Pol CA et al., 2015 * * * * * * ** 8 Arirachakaran P et al., 2013 * * * * * * * ** 9 Yildirim B et al., 2011 * * * * * ** 7 Debanth S et al., 2009 * * * * * * * ** 9 Fehér E et al., 2009 * * * * * ** 7 Razavi SM et al., 2009 * * * * * ** 7 Szarka K et al., 2009 * * * * * * * ** 9 Yu Hong et al., 2007 * * * * * * ** 8 Cianfriglia F et al., 2006 * * * * ** 6 Giovannelli L et al., 2006 * * * * * ** 7 Campisi G et al., 2004 * * * * * ** 7 Ma Jian et al., 2003 * * * * ** 6 OFlatharta C et al., 2003 * * * * * ** 7 Giovannelli L et al., 2002 * * * * * ** 7 Li Hui et al., 2000 * * * * ** 6 Sand L et al., 2000 * * * * * ** 7 Qi Yanchun et al., 1999 * * * * ** 6 Lei Lei et al., 1997 * * * * * ** 7 Vespe M et al., 1997 * * * * ** 6 Boyd AS et al., 1996 * * * ** 5 Cox M et al., 1993 * * * ** 5 Maitland NJ et al., 1987 * * * * ** 6 (1) Was there a specific definition of the diagnosis of this disease in the article? (2) Were the selection criteria for the patients in the study specifically described? (3) How representative was the control group with respect to the source population of cases enrolled? Comparability: Were the groups comparable with respect to age, sex, living area or race? The “Outcome” item was scored for double asterisks (**) in cells; others were scored for a single asterisk (*). Meta-analysis As shown in Fig 2, the studies had a total of 835 patients and 734 controls, of which 293 cases (35.09%) and 57 controls (7.77%) were HPV positive. The χ2 and I2 were 35.81 (P = 0.02) and 41%, respectively, suggesting heterogeneity. The random-effects model was used to analyze the data. The pooled OR was 6.83 (95%CI: 4.15–11.27), and the overall effect Z value was 7.54 (P < 0.00001), which indicated a strong association between HPV and OLP. The visual examination of the symmetry of the funnel plot did not suggest a large publication bias (Fig 2). 10.1371/journal.pone.0161339.g002Fig 2 Forest and funnel plots of 22 included studies on the association between HPV and OLP. Subgroup analyses We further conducted subgroup analyses of all included studies based on geographic population, OLP clinical types and HPV genotypes respectively, to determine the influencing factors that may impact the overall results. Association of HPV with OLP in terms of geographic population The 22 articles included 12 countries in Asia, Europe, and North America. The association varied significantly by geographic population (Table 3). In Asia (OR: 9.37), it was strongest in India (OR: 132.04), followed by Turkey (OR: 8.73), China (OR: 6.58), Iran (OR: 5.85), and Thailand (OR: 3.08). There were 11 studies that reported the association of OLP and HPV (OR: 5.15) in Europe, including Germany (OR: 52.11), Ireland (OR: 15.11), Hungary (OR: 11.21), Sweden (OR: 9.85), England (OR: 5.34) and Italian (OR: 2.43). The relationship between OLP and HPV was especially significant in the U.S. study (OR: 96.60). 10.1371/journal.pone.0161339.t003Table 3 Association of HPV with OLP in terms of geographic population. Region Geographical Articles OLP Control OR (95% CI) population involved (n/N) (n/N) Asia Asia 10 137/336 20/259 9.37[5.38,16.30] India 2 27/36 3/65 132.04[14.92,1168.20] Turkey 1 14/65 0/15 8.73[0.49,154.84] China 5 86/169 16/128 6.58[3.44,12.60] Iran 1 9/29 1/14 5.85[0.66,51.79] Thailand 1 1/37 0/37 3.08[0.12,78.14] Europe Europe 11 145/486 37/465 5.15[2.56,10.34] Germany 1 3/7 0/33 52.11[2.30, 1182.85] Ireland 1 9/29 1/14 15.11[0.84,272.68] Hungary 2 78/238 6/144 11.21[3.32,37.89] Sweden 1 6/22 0/12 9.85[0.51,191.74] England 2 10/12 8/17 5.34[0.91,31.45] Italy 4 38/169 23/239 2.43[1.38,4.27] North American America 1 11/13 0/10 96.60[4.14,2252.15] n, positive number of subjects; N, total number of subjects Association of HPV with OLP in terms of clinical type Two types of OLP were involved, EA-OLP and non-EA-OLP. There were only 3 studies mentioned [20, 24, 27], including Hungary (OR: 17.09 vs 6.64), Italy (OR: 4.37 vs 3.86), and China (OR: 12.00 vs 2.00). As shown in Fig 3, the prevalence of HPV differed significantly between the more risky EA-OLP (OR: 9.34; 95%CI: 4.25–20.56) and non-EA-OLP (OR: 4.32; 95%CI: 1.89–9.87). 10.1371/journal.pone.0161339.g003Fig 3 Forest plots of the association between HPV and OLP in terms of clinical type. (a) EA-OLP and (b) non-EA-OLP. Association of HPV with OLP in terms of HPV genotype The association of OLP and HPV varied across different HPV genotypes. Among the 22 included articles, HPV16 was the most frequently reported genotype, and HPV18 was the second. ORs of the correlation between HPV16 and OLP from 1.83 to 138.05, the pooled OR of 6 papers was 11.27 (95%CI: 4.17–30.43) (Fig 4A). Only Razavi SM et al. [21] and Sand L et al. [29] reported the association between HPV18 and OLP (pooled OR: 6.54) (Fig 4B). 10.1371/journal.pone.0161339.g004Fig 4 Forest plots of the association between HPV and OLP in terms of HPV genotypes. (a) HPV16 and (b) HPV18. Sensitivity analysis The exclusion of low-quality studies did not change the summary estimate significantly, OR was 6.60 (Fig 5). Comparing results across detection methods was not considered feasible, given that only 2 studies used IHC, 1 used ISH, and 19 used PCR. 10.1371/journal.pone.0161339.g005Fig 5 Forest plots of the association between HPV and OLP in high-quality studies. Discussion OLP is a chronic inflammatory mucocutaneous disease whose pathogenesis is still the object of much speculation. Different mechanisms by which OLP may develop have been hypothesized in recent years. The association between OLP and viral infections is one of the most controversial positions. The most widely studied viruses in OLP are HPV and hepatitis-C (HCV) [30]. HPVs are small, double-stranded, and circular DNA viruses. There are approximately 40 genotypes known to infect the oral cavity and urogenital tract [31]. The incidence of HPV oral infection has increased in recent decades, and the infection rate seems to be associated with age and gender [32]. Given the U.S. Annual Report on cancer in 2013, the incidence of HPV-positive oropharyngeal (OP) cancers has increased proportionally, while the total incidence of cancer has recently declined [33]. The latest analysis of U.S. cancer registry data showed that the number of HPV-positive OP cancers diagnosed each year may exceed that of invasive cervical cancers by 2020 [32]. In this way, the association of HPV with OLP and oral cancer has received more attention. This meta-analysis showed a significantly strong association between OLP and HPV. The pooled OR showed that OLP patients have about a 7-fold higher risk of HPV infection than controls, which was consistent with results reported by Syrjänen et al. in 2011 [14]. Besides, we found that the correlation of HPV and OLP vary not only according to geographic populations, but also clinical types of OLP and HPV genotypes. This is the most comprehensive meta-analysis of this issue ever performed. The current study showed there to be wide variations in the association of OLP and HPV with regard to different geographic populations, which was consistent with results reported by Lodi G et al. [34]. Among the 12 countries included, the ORs of HPV/OLP association ranged from 1.00 to 138.05. Studies performed in Italy (OR: 1.12, 1.00) and the U.K. (OR: 2.00) did not show any significant relationship between HPV and OLP [11–13]. In contrast, patients with OLP in India (OR: 132.04), the United States (OR: 96.60), Germany (OR: 52.11), Ireland (OR: 15.11), and Hungary (OR: 11.21) showed an extremely strong association. In this way, efforts into exploring the correlation between HPV and OLP should focus mainly on epidemiological studies of different populations. The most common clinical forms of OLP are EA-OLP and non-EA-OLP. In the current meta-analysis, the association between HPV and EA-OLP (OR: 9.34) was comparable and stronger than the association of HPV and non-EA-OLP (OR: 4.32). The risk of progression to malignancy for non-EA-OLPs was approximately 0.5%, but in EA-OLPs it was at least 3.5–4.0% during similar follow-up periods [27]. All this adds up to a hypothesis that differences in HPV prevalence may influence OLP malignant potential. This could shed light on the possibility of a potential prognostic application of HPV detection in OPMD. Here, the association strength between HPV 16 and OLP was extremely high (OR: 11.27), and HPV types16 has been classified as high-risk (HR) type and reported to be associated with malignancy [35]. Besides, HPV16 has also been reported to be the most common infectious HPV genotype in oral squamous cell carcinoma (OSCC). Its prevalence was 16%. OLP is an OPMD of the oral mucosa with a transformation rate of 0–6.25% [2, 34], so whether HPV 16 has a causal role in OLP malignancy, a long-term follow up is needed to determine. Recent data from case-control and meta-analytic studies suggested that HPV is a causal factor of the development of several types oropharyngeal and oral cavity squamous carcinomas [36–39]. However, only small proportion of individuals who become infected with HPV will develop OSCC. A recent study showed that patient individual susceptibility to HPV infection, and other biomarkers may be related to OSCC development [40]. Other studies showed that the prognosis, age of onset, and incidence for men and women were different between HPV-positive and HPV-negative oral cancer patients [33]. In this way, determining the HPV infection status of HPV-associated OPMD and OSCC patients may be important to prognosis, treatment, and prevention strategies. The major limitations of the current meta-analysis are the detection/diagnostic bias and selection bias of the case-control studies that were included. The 22 articles evaluated in this meta-analysis used three kinds of HPV detection methods: PCR, ISH and IHC. PCR is exquisite sensitive and widespread used, it is generally considered as the gold standard in most meta-analysis [14]. However, PCR detection is prone to false positive results because of contamination may occur during sampling, processing and PCR protocols [7]. ISH can be more sensitive in cases in which only a few cells in the sample tissue contain high copy numbers of the virus [41], but its results can influenced by the quality of the sample (e.g. frozen or fixed) [42]. In this way, detection and selection biases limit the usefulness of these results due to possible errors of these kinds. In future work, a uniform research standard should be established to perform a more precise comparison of the results. Conclusion In conclusion, this systematic review and meta-analysis showed a significant association between HPV and OLP. All sub-analyses strongly and consistently indicated this association. Because of the lack of prospective cohort studies, it was not possible to take a position on the relationship between HPV infection and oral malignancy. More prospective cohort studies are needed to formally confirm the role of HPV as an etiological agent of OLP, and a uniform research standard should be established to produce get more convincing results. Supporting Information S1 PRISMA Checklist PRISMA 2009 Checklist. (DOC) Click here for additional data file. We thank LetPub (www.letpub.com) for its linguistic assistance during the preparation of this manuscript. ==== Refs References 1 Epstein JB , Wan LS , Gorsky M , Zhang L . Oral lichen planus: progress in understanding its malignant potential and the implications for clinical management . Oral Surg Oral Med Oral Pathol Oral Radiol Endod . 2003 ; 96 (1 ):32 –7 . 12847441 2 Gorsky M , Epstein JB . Oral lichen planus: malignant transformation and human papilloma virus: a review of potential clinical implications . Oral Surg Oral Med Oral Pathol Oral Radiol Endod . 2011 ; 111 (4 ):461 –4 . 10.1016/j.tripleo.2010.11.007 21330164 3 McCartan BE , Healy CM . The reported prevalence of oral lichen planus: a review and critique . J Oral Pathol Med . 2008 ; 37 (8 ):447 –53 . 10.1111/j.1600-0714.2008.00662.x 18624932 4 Abbas AK , Lichtman AH . Cellular and Molecular Immunology . 5th edition chapter 15. Philadelphia, Pa, USA : WB Saunders ; 2004 pp. 345 –366 . 5 Maitland NJ , Cox MF , Lynas C , Prime SS , Meanwell CA , Scully C . Detection of human papillomavirus DNA in biopsies of human oral tissue . Br J Cancer . 1987 ; 56 (3 ):245 –50 . 2822070 6 Syrjänen S , Syrjänen K HPV infections of the oral mucosa, Chapter 17 In: Syrjänen K , Syrjänen S , eds. Papillomavirus infections in human pathology . J. Wiley &Sons : New York ; 2000 pp. 379 –412 . 7 Boyd AS , Annarella M , Rapini RP , Adler-Storthz K , Duvic M . False-positive polymerase chain reaction results for human papillomavirus in lichen planus. Potential laboratory pitfalls of this procedure . J Am Acad Dermatol . 1996 ; 35 (1 ):42 –6 . 8682962 8 Vesper M , Riethdorf S , Christoph E , Ruthke A , Schmelzle R , Loning T . Detection of human papillomavirus (HVP)-DNA in oral manifestation of lichen planus . Mund Kiefer Gesichtschir . 1997 ; 1 (3 ):146 –9 . 9410622 9 Debanth S , Singh PA , Mehrotra R , Singh M , Gupta SC , Pandya S , et al Human papillomavirus infection and premalignant lesions of the oral cavity: A cross-sectional study in Allahabad, North India . Asia-Pac J Clin Onco . 2009 ; 5 (2 ):111 –8 . 10 Pol CA , Ghige SK , Gosavi SR . Role of human papilloma virus-16 in the pathogenesis of oral lichen planus—an immunohistochemical study . Int Dent J . 2015 ; 65 (1 ):11 –4 . 10.1111/idj.12125 25256345 11 Cox M , Maitland N , Scully C . Human herpes simplex-1 and papillomavirus type 16 homologous DNA sequences in normal, potentially malignant and malignant oral mucosa . Eur J Cancer B Oral Oncol . 1993 ; 29B (3 ):215 –9 . 8298425 12 Cianfriglia F , Di Gregorio DA , Cianfriglia C , Marandino F , Perrone DR , Vocaturo A . Incidence of human papillomavirus infection in oral leukoplakia. Indications for a viral aetiology . J Exp Clin Cancer Res . 2006 ; 25 (1 ):21 –8 . 16761614 13 Giovannelli L , Campisi G , Colella G , Capra G , Di Liberto C , Caleca MP , et al Brushing of oral mucosa for diagnosis of HPV infection in patients with potentially malignant and malignant oral lesions . Mol Diagn Ther . 2006 ;10 (1 ):49 –55 . 16646577 14 Syrjänen S , Lodi G , von Bultzingslowen I , Aliko A , Arduino P , Campisi G , et al Human papillomaviruses in oral carcinoma and oral potentially malignant disorders: a systematic review . Oral Dis . 2011 ; 17 Suppl 1 :58 –72 . 10.1111/j.1601-0825.2011.01792.x 21382139 15 Chalmers TC , Smith HJ , Blackburn B , Silverman B , Schroeder B , Reitman D , et al A method for assessing the quality of a randomized control trial . Control Clin Trials . 1981 ; 2 (1 ):31 –49 . 7261638 16 Yu Hong , Liu Zhongyan . Relationship between Human Papillomavirus and Oral Lichen Planus . Chin J Dermato Venerol Integ Trad W Med . 2007 ; 6 (3 ):149 –50 . 17 Ma Jian , Yu Yi , Ma Rui , Ling Disheng . The detection of Human papillomavirus in the oral lichen planus patients . J Clin Stomatol . 2003 ; 19 (2 ):113 –4 . 18 Li Hui , Bai Jingchun , Cao Fenglin , Wang Huanfu . Correlative Study of Human Papilloma Virus and Oral Lichen Planus with Gene Detection . Heilongjiang Med J . 2000 ; 6 :9 –10 . 19 Qi Yanchun , Gao Limin , Hu Jianhong , Cao Fenglin . Correlative Study of Human Papilloma Virus and Oral Lichen Planus with Gene Detection . Herbin Med J . 1999 ; 19 (3 ):5 –6 . 20 lei Lei , Li Hui ; Sun Yu . Dection of HPV in Oral Lichen Planus . J Clin Stomatol . 1997 ; 01 :5 –7 . 21 Razavi SM , Ghalayani P , Salehi MR , Attarzadeh H , Shahmoradi M . Human papilloma virus as a possible factor in the pathogenesis of oral lichen planus . Dent Res J (Isfahan) . 2009 ; 6 (2 ):82 –6 .21528036 22 Arirachakaran P , Chansaengroj J , Lurchachaiwong W , Kanjanabud P , Thongprasom K , Poovorawan Y . Oral lichen planus in thai patients has a low prevalence of human papillomavirus . ISRN Dent . 2013 ; 2013 :362750 10.1155/2013/362750 23762572 23 Yildirim B , Senguven B , Demir C . Prevalence of herpes simplex, Epstein Barr and human papilloma viruses in oral lichen planus . Med Oral Patol Oral Cir Bucal . 2011 ; 16 (2 ):e170 –4 . 21196875 24 Campisi G , Giovannelli L , Arico P , Lama A , Di Liberto C , Ammatuna P , et al HPV DNA in clinically different variants of oral leukoplakia and lichen planus . Oral Surg Oral Med Oral Pathol Oral Radiol Endod . 2004 ; 98 (6 ):705 –11 . 15583544 25 Giovannelli L , Campisi G , Lama A , Giambalvo O , Osborn J , Margiotta V , et al Human papillomavirus DNA in oral mucosal lesions . J Infect Dis . 2002 ; 185 (6 ):833 –6 . 11920302 26 Fehér E , Gáll T , Murvai M , Kis A , Boda R , Sápy T , et al Investigation of the occurrence of torque tenovirus in malignant and potentially malignant disorders associated with human papillomavirus . J Med Virol . 2009 ; 81 (11 ):1975 –81 . 10.1002/jmv.21627 19774682 27 Szarka K , Tar I , Feher E , Gall T , Kis A , Toth ED , et al Progressive increase of human papillomavirus carriage rates in potentially malignant and malignant oral disorders with increasing malignant potential . Oral Microbiol Immunol . 2009 ; 24 (4 ):314 –8 . 10.1111/j.1399-302X.2009.00516.x 19572894 28 OFlatharta C , Flint SR , Toner M , Butler D , Mabruk MJ . Investigation into a possible association between oral lichen planus, the human herpesviruses, and the human papillomaviruses . Mol Diagn . 2003 ; 7 (2 ):73 –83 . 14580227 29 Sand L , Jalouli J , Larsson PA , Hirsch JM . Human papilloma viruses in oral lesions . Anticancer Res . 2000 ; 20 (2B ):1183 –8 . 10810419 30 Payeras MR , Cherubini K , Figueiredo MA , Salum FG . Oral lichen planus: Focus on etiopathogenesis . Arch Oral Biol . 2013 ; 58 (9 ):1057 –69 . 10.1016/j.archoralbio.2013.04.004 23660124 31 Marini A , Wagenmann M , Ting E , Hengge UR . Squamous cell cancer and human papillomavirus infection in oral lichen planus: case report and literature review . Dermatol Surg . 2007 ; 33 (6 ):756 –60 . 17550459 32 Gillison ML , Broutian T , Pickard RK , Tong ZY , Xiao W , Kahle L , et al Prevalence of oral HPV infection in the United States, 2009–2010 . JAMA . 2012 ; 307 (7 ):693 –703 . 10.1001/jama.2012.101 22282321 33 Brouwer AF , Eisenberg MC , Meza R . Age Effects and Temporal Trends in HPV-Related and HPV-Unrelated Oral Cancer in the United States: A Multistage Carcinogenesis Modeling Analysis . Plos One . 2016 ; 11 (3 ):e151098 . 34 Lodi G , Scully C , Carrozzo M , Griffiths M , Sugerman PB , Thongprasom K . Current controversies in oral lichen planus: report of an international consensus meeting. Part 2. Clinical management and malignant transformation . Oral Surg Oral Med Oral Pathol Oral Radiol Endod . 2005 ; 100 (2 ):164 –78 . 16037774 35 Zur HH . Papillomaviruses and cancer: from basic studies to clinical application . Nat Rev Cancer . 2002 ; 2 (5 ):342 –50 . 12044010 36 Syrjanen S . The role of human papillomavirus infection in head and neck cancers . Ann Oncol . 2010 ; 21 Suppl 7 :i243 –5 . 37 Dayyani F , Etzel CJ , Liu M , Ho CH , Lippman SM , Tsao AS . Meta-analysis of the impact of human papillomavirus (HPV) on cancer risk and overall survival in head and neck squamous cell carcinomas (HNSCC) . Head Neck Oncol . 2010 ; 2 :15 10.1186/1758-3284-2-15 20587061 38 Miller CS , Johnstone BM . Human papillomavirus as a risk factor for oral squamous cell carcinoma: a meta-analysis, 1982–1997 . Oral Surg Oral Med Oral Pathol Oral Radiol Endod . 2001 ; 91 (6 ):622 –35 . 11402272 39 Gillison ML , D'Souza G , Westra W , Sugar E , Xiao W , Begum S , et al Distinct risk factor profiles for human papillomavirus type 16-positive and human papillomavirus type 16-negative head and neck cancers . J Natl Cancer Inst . 2008 ; 100 (6 ):407 –20 . 10.1093/jnci/djn025 18334711 40 Sun Y , Zhang Y , Liu L , Song X , Li G . Genetic polymorphisms and HPV infection in oral squamous cell carcinomas . Curr Opin Virol . 2015 ;14 :1 –6 . 10.1016/j.coviro.2015.05.004 26057719 41 Syrjanen SM . Basic concepts and practical applications of recombinant DNA techniques in detection of human papillomavirus (HPV) infection. Review article . APMIS . 1990 ; 98 (2 ):95 –110 . 2154242 42 Termine N , Panzarella V , Falaschini S , Russo A , Matranga D , Lo ML , et al HPV in oral squamous cell carcinoma vs head and neck squamous cell carcinoma biopsies: a meta-analysis (1988–2007) . Ann Oncol . 2008 ; 19 (10 ):1681 –90 . 10.1093/annonc/mdn372 18558666

PMC005xxxxxx/PMC5003374.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757096910.1371/journal.pone.0161486PONE-D-16-27048Research ArticleBiology and Life SciencesCell BiologyCell ProcessesCell DeathApoptosisResearch and Analysis MethodsBiological CulturesCell CulturesCultured Tumor CellsGlioma CellsBiology and Life SciencesCell BiologyCellular Structures and OrganellesEndoplasmic ReticulumBiology and Life SciencesCell BiologyCell ProcessesSecretory PathwayEndoplasmic ReticulumBiology and Life SciencesBiochemistryEnzymologyEnzymesOxidoreductasesLuciferaseBiology and Life SciencesBiochemistryProteinsEnzymesOxidoreductasesLuciferaseResearch and Analysis MethodsBioassays and Physiological AnalysisCell AnalysisCell Viability TestingBiology and Life SciencesToxicologyCytotoxicityMedicine and Health SciencesPathology and Laboratory MedicineToxicologyCytotoxicityBiology and Life SciencesToxicologyCytotoxicity AssayMedicine and Health SciencesPathology and Laboratory MedicineToxicologyCytotoxicity AssayBiology and Life SciencesBiotechnologySmall MoleculesPhysical SciencesChemistryChemical CompoundsOrganic CompoundsSmall MoleculesPhysical SciencesChemistryOrganic ChemistryOrganic CompoundsSmall MoleculesA High-Throughput Screen Identifies 2,9-Diazaspiro[5.5]Undecanes as Inducers of the Endoplasmic Reticulum Stress Response with Cytotoxic Activity in 3D Glioma Cell Models Inducers of ERSR Active in 3D Glioma ModelsMartinez Natalia J. 1Rai Ganesha 1Yasgar Adam 1Lea Wendy A. 1Sun Hongmao 1Wang Yuhong 1Luci Diane K. 1Yang Shyh-Ming 1Nishihara Kana 12Takeda Shunichi 4Sagor Mohiuddin 2Earnshaw Irina 2Okada Tetsuya 3Mori Kazutoshi 3Wilson Kelli 14Riggins Gregory J. 4Xia Menghang 1Grimaldi Maurizio 5Jadhav Ajit 1Maloney David J. 1*Simeonov Anton 1*1 National Center for Advancing Translational Sciences, National Institutes of Health, Rockville, MD, 20850, United States of America2 Department of Radiation Genetics, Graduate School of Medicine, Kyoto University, Yoshidakonoe, Sakyo, Kyoto 606–8501, Japan3 Department of Biophysics, Graduate School of Science, Kyoto University, Kitashirakawa-Oiwake, Sakyo, Kyoto 606–8502, Japan4 Department of Neurosurgery, John Hopkins University, Baltimore, MD 21231, United States of America5 Laboratory of Neuropharmacology, Department of Biochemistry and Molecular Biology, Southern Research Institute, Birmingham, AL 35205, United States of AmericaNakano Hiroyasu EditorToho Daigaku, JAPANCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: AS MG DJM AJ. Data curation: AJ HS YW NJM AY. Formal analysis: AJ HS YW. Funding acquisition: AS AJ. Investigation: NJM AY WAL KN. Methodology: NJM AY WAL KN. Project administration: AS AJ DJM NJM. Resources: AS AJ DJM DKL SMY ST MS IE TO KM KW GJR. Software: AJ HS YW. Supervision: AS. Validation: NJM AY. Visualization: NJM. Writing – original draft: NJM AY GR. Writing – review & editing: NJM AY GR MX. * E-mail: asimeono@mail.nih.gov (AS); maloneyd@mail.nih.gov (DJM)29 8 2016 2016 11 8 e01614866 7 2016 6 8 2016 This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.The endoplasmic reticulum (ER) is involved in Ca2+ signaling and protein folding. ER Ca2+ depletion and accumulation of unfolded proteins activate the molecular chaperone GRP78 (glucose-regulated protein 78) which in turn triggers the ER stress response (ERSR) pathway aimed to restore ER homeostasis. Failure to adapt to stress, however, results in apoptosis. We and others have shown that malignant cells are more susceptible to ERSR-induced apoptosis than their normal counterparts, implicating the ERSR as a potential target for cancer therapeutics. Predicated on these findings, we developed an assay that uses a GRP78 biosensor to identify small molecule activators of ERSR in glioma cells. We performed a quantitative high-throughput screen (qHTS) against a collection of ~425,000 compounds and a comprehensive panel of orthogonal secondary assays was formulated for stringent compound validation. We identified novel activators of ERSR, including a compound with a 2,9-diazaspiro[5.5]undecane core, which depletes intracellular Ca2+ stores and induces apoptosis-mediated cell death in several cancer cell lines, including patient-derived and 3D cultures of glioma cells. This study demonstrates that our screening platform enables the identification and profiling of ERSR inducers with cytotoxic activity and advocates for characterization of these compound in in vivo models. http://dx.doi.org/10.13039/100000002National Institutes of HealthU54MH084681http://dx.doi.org/10.13039/100000002National Institutes of HealthDA031669NJM, GR, AY, WAL, HS, YW, DKL, SMY, MX, AJ, DJM and AS were supported by the intramural research program of the National Center for Advancing Translational Sciences and the Molecular Libraries Initiative of the National Institutes of Health Roadmap for Medical Research (U54MH084681 and DA031669). Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction The endoplasmic reticulum (ER) is a multifunctional organelle involved in the synthesis, folding, storage and trafficking of proteins [1]. Protein storage and folding are aided by ER-resident molecular chaperones (mainly the glucose regulated protein 78, GRP78; also referred to as BiP) and high levels of Ca2+ in the ER lumen [2]. The ER is indeed a major site for Ca2+ storage and participates in fast Ca2+ responses underlying many signaling pathways. Disruption of ER homeostasis by physiological and pathological stimuli results in an accumulation of unfolded proteins (a condition known as ER stress) that triggers a complex cascade of events, referred to as the Endoplasmic Reticulum Stress Response (ERSR). These events are aimed at restoring homeostasis and involve an initial attenuation of global protein synthesis and a transcriptional remodeling to mobilize a cohort of stress response genes. The hallmark of ERSR activation is the increase in GRP78 expression, the main sensor of unfolded proteins and a key regulator of the ERSR. In resting conditions, GRP78 levels are low and GRP78 binds to and represses ERSR-activating proteins such as activating transcription factor 6 (ATF6), inositol requiring protein (IRE) 1α, PKR-like endoplasmic reticulum kinase (PERK), and the ER-associated caspase 4/12 (human/mouse, respectively). In the presence of unfolded proteins, conformational changes in GRP78 and active cycling of its ATPase domain, trigger the release of the aforementioned binding partners [3–5]. When stressing conditions are long-lasting or abnormally intense, activation of ERSR will lead to apoptosis. Apoptosis is achieved via three different mechanisms, including transcriptional activation of C/EBP homologous protein (CHOP), activation of c-Jun NH2-terminal kinase (JNK), and activation of ER-associated caspase 4/12. These mechanisms culminate in activation of terminal effector caspases and subsequent cell death [6–9]. Tumor cells display an elevated basal level of ERSR, which allows them to meet the ER demands of rapid cell division under a hostile environment (e.g. hypoxia and low pH) [10]. In fact, the protective effect of mildly elevated levels of ERSR has been correlated to chemotherapeutic tolerance [11–13]. Attempts to block elevated basal levels of ERSR in cancer cells, mainly by inhibiting IRE1α, and subsequent splicing of its target XBP1, have been reported as potential anticancer therapeutics [14, 15]. However, further stimulation of ERSR beyond a critical point is accompanied by enhanced cell death, suggesting that potent ERSR-inducing agents may possess antineoplastic potential [13, 16–18]. Small molecules that induce ERSR, including agents affecting ER Ca2+ homeostasis (e.g. thapsigargin and nonsteroidal anti-inflammatory drugs), protein folding or maturation (e.g. tunicamycin and brefeldin A), and misfolded protein removal (e.g. bortezomib), reportedly cause cytotoxicity [19]; however, these molecules are poor candidates for oncology applications due to suboptimal potency and selectivity, and/or poor bioavailability at the site of tumor formation. Our previous work has not only confirmed the ERSR as a relevant target to induce gliotoxicity but has also identified significant differences in the deployment of ERSR activation that render malignant glioma cells more susceptible to ERSR augmentation compared to normal astrocytes [20]. To further explore this therapeutic advantage, we sought to identify small molecules that activate ERSR and induce gliotoxicity. We implemented a cell-based quantitative high-throughput screen (qHTS) using a luciferase reporter that monitors GRP78 levels in human malignant glioma (U87-MG) cells. For triaging of the resulting screening hits, we devised a panel of secondary assays that validate reporter activation and inform on cytotoxicity potential and mechanism of action (Fig 1). We screened more than 425,000 unique compounds and identified several novel activators of ERSR that induce apoptosis-mediated cell death in multiple glioma cell lines, including U87-MG and patient derived lines, grown both in two-dimensional (2D) monolayers as well as physiologically relevant three-dimensional (3D) spheroid cultures. Furthermore, we synthesized a library of analogs around a hit compound containing an 2,9-diazaspiro[5.5]undecane core (compound 8) and established a structure-activity relationship profile. We find that hit compounds also show dose-response cytotoxic effects in other non-glioma cancer cell lines indicating that these molecules have potentially broad anti-neoplastic activity. Finally, by pairwise combination screening, we show that compound 8 and its analogs synergize with hit compound 6 to reduce the viability of patient-derived glioma cell lines. 10.1371/journal.pone.0161486.g001Fig 1 Schematic workflow for compound screening and triage. qHTS indicates assays performed in 1,536- and 384-well format. AC50 obtained in qHTS assays are summarized in Fig 2. Results of remaining low-throughput assays are summarized in Fig 3. Results Development, miniaturization, and qHTS of a grp78-luciferase assay We have previously shown that ERSR activation as signaled by changes in the levels of GRP78 is a valuable predictor of cytotoxicity in glioma cells [20]. Therefore, we developed a cell-based assay that utilizes GRP78 as a biosensor to identify small molecule inducers of ERSR with associated cytotoxic properties. To this end, we engineered a stable U87-MG cell line to express a firefly luciferase reporter under the control of the GRP78 promoter (assay referred to herein as grp78-luciferase). The known ERSR activator thapsigargin elicited a concentration dependent increase on reporter activity with an EC50 of 11.4 nM (S1A Fig). Importantly, this increase in luciferase reporter expression is comparable to the increase in GRP78 native protein in U87-MG cells under the same treatment [21]. Since this assay could be affected by false negatives due to apoptosis-mediated cell death triggered by hit compounds, compound incubation time was optimized to 16 h, at which point luciferase induction achieves a maximal response of ~6-fold compared to vehicle-treated control. The ERSR activator tunicamycin also elicits an optimal reporter induction at 16 h (S1B Fig). qHTS provides concentration response curves (CRCs) of a chemical library directly from the primary screen and consequently has advantages over traditional screening paradigms in which each library member is tested at a single concentration [22, 23]. Therefore, we miniaturized the grp78-luciferase assay in 1,536-well format where a similar induction of ~6-fold and EC50 of 9.4 nM was observed with thapsigargin (S1C Fig and see S1 Supporting Information for additional details). To identify potential activators of ERSR, the grp78-luciferase assay was used in a robotic screen against a collection of 427,208 compounds contained in the NCGC chemical library, with 236 compounds exhibiting reporter activation (S1D Fig and S1 Supporting Information). Given that GRP78 activation leads to the deployment of the ERSR and potentially to cell death, we tested the activity of hit compounds in a cytotoxicity assay. To this end, we measured the viability of U87-MG cells after a 48 h compound exposure. The assay was run in 1,536-well format using the CellTiter-Glo (Promega) reagent, which quantifies cellular ATP levels as a proxy for viability. A total of 53 compounds were active in both the screening assay re-test and the CellTiter-Glo assays (S1E Fig). After we applied structural filters that eliminate electrophiles and other problematic compounds, a total of 8 compounds with favorable luciferase and cytotoxicity profiles (>100% and >50% efficacy, respectively) were chosen for further characterization (S2A and S2B Fig; Fig 2). To this end, we devised a panel of secondary assays designed to validate ERSR induction and cytotoxic activity (Fig 1). 10.1371/journal.pone.0161486.g002Fig 2 Chemical structure and AC50 (μM) of hit compounds in qHTS assays. For Ca2+ mobilization, values obtained in + EGTA conditions are displayed. Active compounds are considered those with a high quality CRC and efficacy. Efficacy cutoffs are as follow: grp78-luciferase and UPRE-luciferase, >40%; Caspase 3/7 and Ca2+ mobilization, >30%; all cell viability assays, >50%. Effect of compounds on the activation of the three ERSR branches We first sought to confirm the effect of the 8 hit compounds on ERSR deployment. To determine whether hit compounds increase endogenous GRP78 levels, U87-MG cells were incubated with indicated compounds at concentrations of 5, 10, or 20 μM (concentration points shown to elicit significant response based on grp78-reporter CRCs) and subjected to western blotting using an anti-GRP78 antibody. As shown in Fig 3A, 5 out 8 compounds (compounds 1, 2, 3, 5 and 8) upregulate endogenous GRP78 to levels ≥30% higher than the induction achieved by treatment with 20 μM of the positive control thapsigargin. In fact, treatment of U87-MG cells with 10 and 20 μM concentrations of compounds 5 and 8 upregulated GRP78 protein to comparable levels as those achieved with thapsigargin (Fig 3A and S3 Fig). 10.1371/journal.pone.0161486.g003Fig 3 Secondary assays identify bona fide ERSR inducers with cytotoxicity activity. (A) U87-MG cells treated with the indicated compounds were analyzed by immunoblot for GRP78 expression. GAPDH levels were used as housekeeping controls. Data are represented as GRP78/GAPDH ratio for each compound and normalized to the GRP78/GAPDH ratio for 20 μM thapsigargin. See S3 Fig for blot images. (B) U87-MG cells were treated with DMSO, 1 μM thapsigargin or 20 μM of indicated compound for 16 h. XBP-1 mRNA splicing was monitored by RT-PCR using XBP-1-specific primers, which amplify a spliced or unspliced fragment of 304 or 326 bp, respectively. (C) U87-MG cells were treated with DMSO, 1 μM thapsigargin or 20 μM of indicated compound for 24 h. CHOP and housekeeping GAPDH mRNA levels were monitored by qRT-PCR. Data is plotted relative to the DMSO treated sample set to 1. Error bars indicate the SD of three replicates. (D) U87-MG cells treated with the indicated compounds were tested for their ability to form colonies. Data are presented as the number of colonies, normalized to vehicle DMSO. Error bars indicate SE with n = 3. See S5 Fig for representative images. In addition, we utilized a complementary cell-based assay that reports on transcriptional activation of ERSR signaling. Specifically, we engineered a human lymphoblastoid TSCER2 cell line to express a firefly luciferase reporter under the control of 5 copies of the unfolded protein response element (UPRE). The UPRE was originally identified as an ATF6 DNA binding element through in vitro binding site selection and was later shown to bind to spliced XBP-1. Importantly, it was reported to be strongly induced during the ER stress response [24–26]. The assay (referred to as UPRE-luciferase) was performed in 1,536-well format with each compound tested in dose response (S1 Supporting Information). We found that the same compounds that increased endogenous GRP78 protein levels also upregulated the UPR reporter (Fig 2; S2C Fig). Together, these results indicate that compounds 1, 2, 3, 5 and 8 are able to inducethe ERSR response.We also investigated the effect of these compounds on two other components of the ERSR pathway. First, using an XBP1 mRNA splicing assay, which informs on activation of the IRE1α branch, we found that U87-MG cells treated with 20 μM of compounds 1, 3, 5, 6 and 8 for 16 h, have detectable levels of the XBP1 spliced transcript (Fig 3B). Second, we measured ATF4-mediated CHOP transcriptional induction to determine activation of the PERK branch [5]. We found that U87-MG cells treated with 20 μM of either compound 1 or 5 for 24 h display a 7–10 fold increase in CHOP transcript levels as measured by qRT-PCR (Fig 2C). Notably, compound 8 was the best-performing compound at increasing CHOP transcript levels (>60-fold compared to DMSO control). Effect of compounds on cytosolic Ca2+ levels ER stress can be triggered by various stimuli, including disturbances in Ca2+ homeostasis. To identify activators that trigger ERSR by depleting intracellular Ca2+ stores, we employed a Ca2+ mobilization assay, using the fluorescent reporter Fluo-8 [27]. In this assay, the ionophore A23187 and thapsigargin, which functions by blocking the sarco/endoplasmic reticulum ATPase (SERCA), lead to an increase of cytosolic Ca2+. On the other hand, tunicamycin, an inhibitor of the UDP-N-acetylglucosamine-dolichol-phosphate N-acetylglucosamine-1-phosphate transferase (GPT) that causes accumulation of unfolded glycoproteins in the ER, does not affect cytosolic Ca2+ concentrations (S4A Fig). The assay was run in a 1,536-well format and compounds that exhibited a high quality CRC and >30% increase in Fluo-8 fluorescence were considered active. Of the 8 hits, compounds 6 and 8 strongly mobilized intracellular Ca2+ levels and compound 2 did as well but to a lesser extent (S2D Fig). The elevated levels of cytosolic Ca2+ after compound treatment could in principle originate from the extracellular environment or intracellular stores like the ER or mitochondria. To distinguish between these two possibilities, we added 10 mM EGTA to the media to fully chelate extracellular Ca2+ before compound treatment (S4B Fig). Even in the presence of EGTA, compounds 2, 6 and 8 were able to increase cytosolic Ca2+, indicating that they function to deplete intracellular Ca2+ stores (Fig 2 and S2E Fig). Effect of compounds on cell death We sought to further investigate the cytotoxic effect of hit compounds. To verify apoptosis as the mechanism underlying the observed cytotoxicity, we tested compounds for their ability to activate key effector caspases using the Caspase-3/7 Glo detection system (Promega) (S1 Supporting Information). Compounds were tested as a dilution series and those that exhibited a high quality CRC and >30% increase in caspase 3/7 activation were considered active. Of the 8 hits, all the compounds showed significant caspase 3/7 activation except compound 1 (Fig 2; S2F Fig). To assess the potential therapeutic use of the newly-discovered activators, we next tested their cytotoxic effects against two patient-derived suspension cell lines. The cell lines JHH-136 and JHH-520 were generated from grade 4 glioblastomas at the Johns Hopkins Medical Institute [28]. Specifically, we performed cell viability assays using the CellTiter-Glo reagent (S1 Supporting Information). Of the 8 compounds, only compound 4 failed to elicit a significant decrease in cell viability (Fig 2 and S2G and S2H Fig). Next, we implemented a complementary clonogenic assay to examine the effects of compounds 1, 2, 3, 5, and 8 on cell survival and proliferation. Briefly, cells were plated at low density and proliferated to form colonies in the continuous presence of indicated compounds dozed at 10 and 20 μM. Importantly, the readout involves colony staining using crystal violet and counting as opposed to measuring global ATP levels as done in the CellTiter-Glo assay. As shown in Fig 3D and S5 Fig, at the highest concentration tested of 20 μM, compounds 1, 3, 5, 8 markedly inhibited the colony forming capacity of U87-MG cells relative to DMSO-treated control. While compounds 1 and 3 were less effective at 10 μM, compounds 5 and 8 reduced colony numbers by more than 50% at that concentration. Compound 2 failed to significantly reduce colony numbers at either concentration tested. Three-dimensional (3D) culture platforms are regarded in many ways as more physiologically relevant multicellular models than traditional monolayers. Cells maintained in 3D cultures have been reported to display altered sensitivities toward drugs [29, 30]. Hence, we tested the effect of the activators in a medium-throughput 3D spheroid culture assay. Briefly, cells were cultured in 384-well ultra-low attachment (ULA) plates for 3 days to allow the formation of aggregates. Of note, U87-MG cells formed a more compact spheroid compared to JHH-136 and JHH-520 cells (data not shown). Spheroids were treated with compounds for an additional 5 days and viability was measured using CellTiter-Glo (3D). Compounds 1, 2, 4 and 7 were inactive against all three cell lines. Compound 6 displayed activity in U87-MG and JHH-520 spheroids and compounds 3, and 5 only in U87-MG spheroids. Compound 8 was the only one to elicit toxicity in all three cell lines (Fig 2 and S2I and S2K Fig). Structure-Activity Relationship Studies of an ERSR inducer Our panel of secondary screens filtered out the majority of the compounds tested (Fig 2). Compounds 4, 6 and 7 could embody weak ERSR activators or false positives of the grp78-luciferase assay, since they failed to induce activation of either one (compound 6) or both (compound 4 and 7) of the ERSR branches. Interestingly, they are able to induce apoptosis-mediated cell death, perhaps by a mechanism different from the apoptotic arm of the ERSR. When cells were treated with these three compounds, cell death was evident in monolayer cultures, and in the case of compound 6, also in 3D cultures. On the other hand, compound 2 is likely a partial ERSR-inducing agent since it failed to induce CHOP mRNA as well as XBP-1 splicing. While it mobilizes intracellular Ca+2 and decreases viability of U87-MG cells in 2D CellTiter-Glo-based assays, compound 2 fails to reduce the colony forming capacity of U87-MG cells as well as the viability of spheroids of all three kinds. Based on these results, it is tempting to speculate that continuous exposure of U87-MG cells to compound 2 could induce ERSR but to levels that are still protective to cells. Alternatively, cells could develop resistance and overcome compound 2-induced ERSR activation. Compounds 1, 3, and 5 activate all three branches of the ERSR and induce cell death in 2D cultures; however, they possess a limited ability to induce significant death when cells are cultured in 3D structures, making them less attractive for follow-up studies. The remaining compound 8, on the other hand, met all the secondary screen criteria of a bona fide ERSR activator with intracellular mobilization Ca+2 potential and gliotoxic activity in all cell viability assays/formats tested (Fig 2). We then resynthesized compound 8 and confirmed its ability to activate GRP78 and induce apoptosis in U87-MG cells (data not shown) and focused our medicinal chemistry optimization efforts on compound 8 to investigate systematic structure activity relationships. Briefly, the 2,9-diazaspiro[5.5]undecane spirocyclic core (Fig 4A, yellow circle) is key for the ERSR activity as measured by activity in the grp78-luciferase assay. Attempts to replace the core with several acyclic, monocyclic, bicyclic and fused ring systems resulted in a significant loss of potency (S2 Table). Similarly, the diphenylmethyl group on the left side of the molecule is essential to induce grp78 reporter activation (Fig 4A, red circle). Any structural modifications to the diphenylmethyl group triggered a complete loss of potency (S3 Table). Several structural changes to the 3,5-dimethylisoxazole moiety on the right side of the molecule were tolerated (Fig 4A, blue circle; S4 Table). However, the amide functionality proved to be strongly preferred as sulfonamide or methylene groups decreased the potency (S5 Table). Replacement of the dimethylisoxazole moiety with other heterocycles improved the potency but aryl/substituted aryl groups decreased the potency. Overall this chemotype exhibited very tight SAR with even subtle changes to the core region or diphenylmethyl groups resulting in a loss of potency. In summary, we synthesized and tested over 150 novel analogs (representative analogs are shown in S2–S5 Tables) around compound 8 with only a few of those showing comparable potency to the original hit. 10.1371/journal.pone.0161486.g004Fig 4 Structure-Activity Relationship for compound 8. (A) Structure of compound 8. The 2,9-diazaspiro[5.5]undecane spirocyclic core, the diphenylmethyl group and the 3,5-dimethylisoxazole moiety are highlighted in yellow, red and blue, respectively. (B) U87-MG cells treated with the indicated analogs were analyzed by immunoblot for GRP78 expression. GAPDH levels were used as housekeeping controls. Data are represented as GRP78/GAPDH ratio for each compound and normalized to the GRP78/GAPDH ratio for 20 μM thapsigargin. See S3 Fig for blot images. (C) U87-MG cells were treated with DMSO, 1 μM thapsigargin or 20 μM of indicated analog for 16 h. XBP-1 mRNA splicing was monitored by RT-PCR using XBP-1-specific primers, which amplify a spliced or unspliced fragment of 304 or 326 bp, respectively. (D) U87-MG cells were treated with DMSO, 1 μM thapsigargin or 20 μM of indicated compound for 24 h. CHOP and housekeeping GAPDH mRNA levels were monitored by qRT-PCR. Data is plotted relative to the DMSO treated sample set to 1. Error bars indicate the SD of three replicates. (E) U87-MG cells treated with the indicated compounds were tested for their ability to form colonies. Data are represented as the number of colonies, normalized to vehicle DMSO. Error bars indicate SE with n = 3. Note that 8e is an inactive control. Analogs 8a and 8d were tested only at 20 μM. See S7 Fig for representative images. Selected analogs are bona fide ERSR inducers with cytotoxic activity in multiple cancer cell lines We chose a set of 4 active analogs (8a, 8b, 8c and 8d) and one inactive control molecule (8e) (Fig 5 and S6A Fig) for validation through our panel of secondary assays. We first look at the ability of these analogs to activate all three branches of the ERSR pathway. All active analogs increased endogenous GRP78 protein when tested at 20 μM. At lower doses however, only 8 and 8a significantly increased GRP78 levels (Fig 4B). All active analogs induced ERSR activation as determined by the UPRE reporter assay, XBP-1 splicing and CHOP qRT-PCR assays (S6B Fig, Fig 3C and 3D, respectively) with compounds 8 and 8d being the best performing molecules at increasing UPRE reporter levels and compounds 8 and 8a at increasing CHOP transcript levels. Compound 8 was the most efficacious at mobilizing intracellular Ca+2 (S6C and S6D Fig). Importantly, the inactive analog 8e did not induce ERSR nor mobilize intracellular Ca+2 stores. Second, we looked at the ability of these compounds to induce cell death. All active analogs induce apoptosis-mediated cell death in U87-MG, JHH-136 and JHH-520 cultures (S6E–S6H Fig). Colony forming assays indicated that at 20 μM, all active analogs were able to reduce the proliferating capacity of U87-MG cells (Fig 4E). When tested at 10 μM, the original hit was more efficacious than analogs 8c and 8d (analogs 8a and 8d were not tested at this concentration, Fig 4E). Notably, all analogs were able to decrease cell viability of 3D spheroids when incubated for 5 days (S6I–S6K Fig). U87-MG spheroids treated for 48 h with compound 8 had a higher population of dead cells, as measured by propidium iodide staining, compared to the inactive analog 8e (Fig 6). Overall, none of the analogs showed significant potency improvements over the original hit compound. Fig 5 summarizes the AC50 values of compound 8 analogs in all qHTS assays. The cytotoxic effects for these compounds are not specific for glioma cell lines since treatment of ovarian (HEY-A8 and IGROV-1) and colon (HT-29) with compound 8 and analogs showed similar reduction in viability as seen in glioma (U87-MG and LN-299) cell lines (Fig 7). 10.1371/journal.pone.0161486.g005Fig 5 Chemical structure and AC50 (μM) of compound 8 analogs in qHTS assays. Efficacy cutoffs are the same as in Fig 2. 10.1371/journal.pone.0161486.g006Fig 6 Compound 8 induces cell death in U87-MG spheroids. U87-MG spheroids were treated for 48 h with either compound 8 or inactive analog 8e at the indicated concentrations. Spheroids were stained with Hoechst (blue) and propidium iodide (red) to mark nuclei and dead cells, respectively. For each spheroid, one single Z’ plane is shown. 10.1371/journal.pone.0161486.g007Fig 7 Compound 8 analogs reduce the viability of cancer cell lines. CellTiter-Glo viability assay of glioma (U87-MG and LN-229), ovarian (IGROV-1 and HEY-A8) and colon (HT-29) cancer cell lines treated for 48 h with compound 8 analogs. Selected compounds synergize to reduce cell viability Compounds that trigger ERSR via different mechanisms/targets could potentially synergize to induce cell death. We selected three validated ERSR inducers with cytotoxicity effects (compounds 8, 8a and 8c), a validated ERSR inducer with limited cytotoxicity (compound 3), as well as a weak ERSR inducer (compound 6) for combinatorial screening in JHH-136 and JHH-520 cell lines. We utilized a combination screening platform previously described [31, 32]. Briefly, we performed a pairwise 6x6 dose response matrix, where compounds were tested in a range of 0–10 μM (final concentration), to assess antagonistic, synergistic, or additive effects on cell viability. The highest dose of each compound (10 μM) was chosen as it is lower than the AC50 of each compound in cell viability assays (Fig 2). Viability response matrices were characterized using the Bliss model and summarized using the DBSumNeg metric [32]. For compound 8 and both analogs, we observed a strong synergism with compound 6 but not compound 3, in both cell lines (Fig 8 and S8 Fig). Although compound 6 is a weak ERSR activator, it mobilizes intracellular Ca2+ to comparable levels as compound 8 does. It is tempting to speculate that a massive depletion of intracellular Ca2+ stores is causing the observed synergism. Importantly, no synergism was observed between compound 8 and its analogs, consistent with the idea that all of them work through the same mechanism/target (S9 Fig). No antagonism was observed for any of the combinations tested (Fig 8; S8 and S9 Figs). 10.1371/journal.pone.0161486.g008Fig 8 Compound 8 and 6 synergize to reduce the viability of patient-derived glioma cell lines. Combination (6 x 6) response profiles for compound 8, 6, and 3 in JHH-136 (A) and JHH-520 (B) cells. Each response profile is displayed as heatmaps of cell viability (percent response normalized to thapsigargin control; left panels) and DBSumNeg analysis (right panels). Discussion We implemented an assay that quantitatively monitors GRP78 levels in human malignant glioma cells upon treatment with small molecules. We devised a compound triage strategy that included multiple quantitative high throughput as well as fixed-dose low throughput assays to enable the confident identification of small molecules with potent ERSR-inducing and apoptotic-mediated cytotoxic activities. Notably, a hit (compound 8) with such properties was identified among our collection of over 425,000 unique chemical entities. Although further analyses are needed to fully understand the mechanism of action of this compound and its molecular target, it likely triggers ERSR by depleting intracellular Ca2+ stores. Our medicinal chemistry optimization efforts around compound 8 indicated that this chemotype exhibits very tight SAR and any significant changes to the molecule resulted in loss of potency. We describe here four top actives (analogs 8a, 8b, 8c and 8d), which along with the original hit compound, constitute a set of chemical tools worthy of in vivo ERSR-induced cytotoxicity studies. To our knowledge, this is the first cell-based qHTS aimed to identify ERSR activators, via GRP78 induction, that induce apoptosis-mediated cell death. The work by Kudo et al. utilized a similar screening strategy to identify small molecules that induce GRP78 expression to protect neurons from ERSR-mediated apoptosis [33]. The authors identified BIX (Bip inducer X), which was shown to activate ER stress response elements through the ATF6 pathway but did not induce XBP-1 splicing nor CHOP expression. BIX was not included in our compound collection. In the recent report by Bi et al. the authors described the generation of a HeLa reporter cell line expressing β-lactamase under the control of the grp78 promoter [34]. The authors performed a pilot screen against the NIH Chemical Genomics Center Pharmaceutical Collection (NPC) of 2,800 drugs and identified 6 compounds that induce reporter expression. Four of those six drugs (6-thioguanine, primaquine, amlodipine and roxindole) tested negative in our primary screen; AMI-193 induced luciferase activity in our primary screen with an efficacy below 40% and was not included in follow up studies; the AMI-193-related drug, spiperone, induced reporter activity with an efficacy of ~40% but did not induce significant cell death of glioma cell lines. A different strategy was recently implemented to screen Chinese hamster ovary (CHO) K1 cells expressing luciferase constructs that individually reported on CHOP activation or IRE1α activity (via XBP1 mRNA splicing). This screen identified sulfonamidebenzamides as selective CHOP activators among a collection of 331,676 MLSMR compounds. One of these sulfonamidebenzamide derivatives showed antiproliferative activity against multiple cancer cell lines [35]. These compounds tested negative in our grp78-luciferase assay, consistent with GRP78 functioning upstream of CHOP and XBP1. However, the different cell types interrogated in these three studies could account for differential activity of identified ERSR activators. The hits identified in this study provide excellent starting points not only for developing chemical probes but also to investigate the potential of combination therapy with other compounds, including ERSR stressors that function via alternative mechanisms. Indeed, we showed that compound 8 and analogs synergize with compound 6 to reduce the viability of patient-derived glioma cells. Other compounds of particular interest for combination studies are those that target autophagy. Autophagy and ERSR-mediated apoptosis are molecularly linked; thapsigargin and tunicamycin have both been shown to induce autophagy in several cell types [36–39]. Autophagy is currently considered a promising target for chemotherapy and targeting both pathways poses an attractive paradigm for oncology treatments [40–42]. Experimental Procedures Cell lines and culture conditions U87-MG, LN-229, and HT-29 cells were obtained from America Type Culture Collection, (ATCC #HTB-14, CRL-2611, and HTB-38, respectively). IGROV-1 and HEY-A8 were obtained from the NCI-60 panel of human cancer cell lines [43]. U87-MG, LN-229, HT-29, and HEY-A8 cells were cultured in Dulbecco's Modified Eagle Medium (DMEM; Life Technologies) supplemented with 10% fetal bovine serum (FBS; HyClone), 100 U/ml penicillin and 100 μg/ml streptomycin (referred to as 1% Pen/Strep; Life Technologies). IGROV-1 cells were cultured in RPMI-1640 (Life Technologies), supplemented with 2mM L-Glutamine (Life Technologies), 10% FBS and 1% Pen/Strep. Generation and characterization of JHH-136 and JHH-520 cell lines has been previously described [28]. Both cell lines were grown in suspension in Neurocult NS-A Stem Cell media (StemCell Technology), containing 20 ng/ml human EGF (Peprotech), 10 ng/ml human basic FGF (Peprotech) and 0.2% heparin (StemCell Technology). TSCER2 cells, derived from the human lymphoblast cell line TK6 [44], were cultured in RPMI 1640 medium (Life Technologies) supplemented with 5% FBS (Gemini Bio-Products), 1 mM sodium pyruvate (Life Technologies) and 1% Pen/Strep. All cultures were maintained in a 37°C incubator with 5% CO2 and under a humidified atmosphere. Generation of grp78-luciferase and UPRE-luciferase stable cell lines The pGL3-GRP78-(-132-+7)-Luc construct containing a firefly luciferase reporter gene under the control of the grp78 promoter was stably transfected into U87-MG cells. Transfection was performed using Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer instructions. A clone that robustly and consistently expressed luciferase following treatment with either thapsigargin or tunicamycin was isolated and used for assay optimization. To this end, cells were seeded at 250,000 cells/well in 24 well dishes in DMEM+10% FBS. Compounds were added as DMSO solutions at the indicated concentrations and incubated for 8, 16, 24 or 48 hr. Luciferase activity was measured by adding a volume of Bright Glo (Promega) reagent as per manufacturer’s instructions. The p5xUPRE-GL3 was stably transfected into TSCER2 cells [44] using a Gene Pulser apparatus (Bio-Rad, Hercules, CA) at 250 V and 950 μF. A clone that robustly and consistently expressed luciferase following treatment with either tunicamycin or 17-AAG was isolated and used for secondary screens. Colony Forming Assay U87-MG cells were plated onto 6-well plates at a density of 1,000 cells/well in DMEM, supplemented with 10%FBS and 1% Pen/Strep and incubated O.N. at 37°C, 5%CO2, 95% RH. Cells were treated with either DMSO vehicle or compound solution at the indicated concentration (keeping a final concentration of 0.2% DMSO). Twelve days after compound treatment, cells were fixed, stained and counted as described [45]. Western Blot Assay is based on a previous method with modifications [20]. U87-MG cells were seeded (~3 × 105 cells/well in 6-well plates) and treated with either vehicle DMSO, control thapsigargin or indicated compound at doses of 5, 10, or 20 μM for 16 h. Whole cell lysates were prepared using RIPA buffer (Cell Signaling) and protease inhibitor cocktail (Cell Signaling). Cell lysates were quantified using the Bio-Rad DC Protein Assay. 1 μg/μl of lysates were loaded on 4–12% gradient NuPAGE® Novex® Bis-Tris gels (Invitrogen) in MOPS SDS running buffer. Proteins were transferred to nitrocellulose membranes using the iBlot gel transfer system (Invitrogen) and blocked in TBST buffer (20 mM Tris pH 7.5, 150 mM NaCl, 0.05% Tween20, 5% BSA). Membranes were incubated with rabbit polyclonal anti-GRP78 primary antibody (# sc-13968, Santa Cruz Biotechnology) at a 1:1,000 dilution and mouse monoclonal anti-GAPDH antibody (# G8795, Sigma) at a 1:20,000 dilution. Either HRP- or Cy3/Cy5-conjugated secondary antibodies were used as follow: HRP-anti-rabbit IgG (# sc-2317, Santa Cruz Technology) or HRP-anti-mouse IgG (# sc-2031, Santa Cruz Biotechnology) were used at 1:2500 dilution and visualized with SuperSignal™ West Dura Chemiluminescent Substrate (Thermo Scientific) on a Bio-Rad Universal Hood II. ECL Plex-anti-rabbit IgG-Cy5 (# PA45012; GE Healthcare Life Sciences) or ECL Plex-anti-Mouse IgG-Cy3 (# PA43010, GE Healthcare Life Sciences) were used at 1:2,500 dilution and visualized using the Typhoon FLA 9500 (BPG1/532 nm/Cy3; LPR/635 nm/Cy5; GE). Protein quantification was performed using ImageQuant TL (GE) software. CHOP qRT-PCR and XBP-1 splicing assay U87-MG cells were seeded (~3 × 105 cells/well in 6-well plates) and treated with either DMSO, 1 μM thapsigargin or 20 μM of indicated compounds and incubated for either 24 (CHOP) or 16 h (XBP-1). RNA extraction and cDNA synthesis were performed using the TaqMan Gene Expression Cells-to-Ct Kit (TermoFisher Scientific) as per manufacturer’s instructions. Human CHOP and GAPDH transcripts were detected using the FAM-MGB Hs500358796_g1 and Human GAPD endogenous control (VIC-MGB, primer limited) probes, respectively (ThermoFisher Scientific) in a ViiA7 system (Applied Biosystems). XBP-1 splicing fragments were detected as described [34]. Spheroid Imaging U87-MG were cultured in DMEM, supplemented with 10% FBS and 1% P/S. Cells were plated onto 384-well, ultra-low attachment, spheroid microplates (Corning #3830) at a 750 cells/well/30 μl density using a Multidrop dispenser and spun down for 30 seconds at 1,000 rpm. Cells were incubated at 37°C, 5% CO2, under a humidified atmosphere for 3 days to allow spheroid formation (1 spheroid/well). Compounds were dissolved in growth media and 10 μl/well were added to a final concentration range of 156 nM to 50 μM (final assay contains 0.5% DMSO). To maintain the three dimensional structure, compounds were incubated for 48 h before imaging. Spheroids were stained with Hoechst 33342 (0.5 nM final) and PI (1 μg/ml final) for 2 h and imaged at 20X magnification on one Z’ plane using the InCell 6000 (GE) plate image reader. qHTS Assays Detailed descriptions of all qHTS assays, primary screening, data analysis and associated protocols are included in S1 Supporting Information. S1 Table summarizes qHTS assay performance. General Chemistry Methods Compound synthesis is described in S1 Supporting Information. Supporting Information S1 Fig Development and characterization of a qHTS grp78-luciferase assay. (TIF) Click here for additional data file. S2 Fig Activity plots for 8 hit compounds in secondary qHTS assays. (TIF) Click here for additional data file. S3 Fig Western blot to assess endogenous GRP78 expression. (TIF) Click here for additional data file. S4 Fig An intracellularCa2+ mobilization assay. (TIF) Click here for additional data file. S5 Fig Colony forming assay. (TIF) Click here for additional data file. S6 Fig Activity plots for top compound 8 analogs in secondary validation qHTS assays. (TIF) Click here for additional data file. S7 Fig Colony forming assay. (TIF) Click here for additional data file. S8 Fig Analogs 8a and 8c synergize with compound 6 but not with compound 3, to reduce the viability of patient-derived glioma cell lines. (TIF) Click here for additional data file. S9 Fig Compound combinations that exhibit minimal to no synergism. (TIF) Click here for additional data file. S1 Supporting Information This file contains experimental procedures, supporting figure legends and references. (DOCX) Click here for additional data file. S1 Table Performance summary of qHTS assays. (PDF) Click here for additional data file. S2 Table SAR around the core. (PDF) Click here for additional data file. S3 Table SAR around the diphenylmethyl region. (PDF) Click here for additional data file. S4 Table SAR around the isoxazole region. (PDF) Click here for additional data file. S5 Table SAR around the amide region. (PDF) Click here for additional data file. We thank Sam Michael, Carleen Klumpp-Thomas and Charles Bonney for assistance with robotic screens; Paul Shinn, Danielle VanLeer, and Crystal McKnight for management of compound libraries; William Leister, Heather Baker, Elizabeth Fernandez, Burchelle Blackman, and Christopher Leclair for analytical chemistry and purification support; Noel Southall for assistance with analysis of calcium mobilization datasets; Raj Guha for assistance with analysis of combinatorial screening datasets; Mark Henderson for critical comments, and Steven Titus for assistance with spheroid imaging. ==== Refs References 1 Estrada de Martin P , Novick P , Ferro-Novick S . The organization, structure, and inheritance of the ER in higher and lower eukaryotes . Biochemistry and cell biology = Biochimie et biologie cellulaire . 2005 ;83 (6 ):752 –61 . Epub 2005/12/08. 10.1139/o05-159 .16333327 2 Michalak M , Robert Parker JM , Opas M . Ca2+ signaling and calcium binding chaperones of the endoplasmic reticulum . Cell calcium . 2002 ;32 (5–6 ):269 –78 . Epub 2003/01/25. .12543089 3 Bertolotti A , Zhang Y , Hendershot LM , Harding HP , Ron D . Dynamic interaction of BiP and ER stress transducers in the unfolded-protein response . Nature cell biology . 2000 ;2 (6 ):326 –32 . Epub 2000/06/15. 10.1038/35014014 .10854322 4 Rao RV , Peel A , Logvinova A , del Rio G , Hermel E , Yokota T , et al Coupling endoplasmic reticulum stress to the cell death program: role of the ER chaperone GRP78 . FEBS letters . 2002 ;514 (2–3 ):122 –8 . Epub 2002/04/12. 11943137 5 Hetz C . The unfolded protein response: controlling cell fate decisions under ER stress and beyond . Nature reviews Molecular cell biology . 2012 ;13 (2 ):89 –102 . Epub 2012/01/19. 10.1038/nrm3270 .22251901 6 Yoshida I , Monji A , Tashiro K , Nakamura K , Inoue R , Kanba S . Depletion of intracellular Ca2+ store itself may be a major factor in thapsigargin-induced ER stress and apoptosis in PC12 cells . Neurochemistry international . 2006 ;48 (8 ):696 –702 . Epub 2006/02/17. 10.1016/j.neuint.2005.12.012 .16481070 7 Soboloff J , Berger SA . Sustained ER Ca2+ depletion suppresses protein synthesis and induces activation-enhanced cell death in mast cells . The Journal of biological chemistry . 2002 ;277 (16 ):13812 –20 . Epub 2002/02/12. 10.1074/jbc.M112129200 .11836247 8 Oyadomari S , Mori M . Roles of CHOP/GADD153 in endoplasmic reticulum stress . Cell Death Differ . 2004 ;11 (4 ):381 –9 . Epub 2003/12/20. 10.1038/sj.cdd.4401373 .14685163 9 Yoneda T , Imaizumi K , Oono K , Yui D , Gomi F , Katayama T , et al Activation of caspase-12, an endoplastic reticulum (ER) resident caspase, through tumor necrosis factor receptor-associated factor 2-dependent mechanism in response to the ER stress . The Journal of biological chemistry . 2001 ;276 (17 ):13935 –40 . Epub 2001/03/30. 10.1074/jbc.M010677200 .11278723 10 Warburg O . On the origin of cancer cells . Science . 1956 ;123 (3191 ):309 –14 . Epub 1956/02/24. .13298683 11 Li J , Lee AS . Stress induction of GRP78/BiP and its role in cancer . Current molecular medicine . 2006 ;6 (1 ):45 –54 . Epub 2006/02/14. .16472112 12 Dong D , Ni M , Li J , Xiong S , Ye W , Virrey JJ , et al Critical role of the stress chaperone GRP78/BiP in tumor proliferation, survival, and tumor angiogenesis in transgene-induced mammary tumor development . Cancer research . 2008 ;68 (2 ):498 –505 . Epub 2008/01/18. 10.1158/0008-5472.CAN-07-2950 .18199545 13 Pyrko P , Schonthal AH , Hofman FM , Chen TC , Lee AS . The unfolded protein response regulator GRP78/BiP as a novel target for increasing chemosensitivity in malignant gliomas . Cancer research . 2007 ;67 (20 ):9809 –16 . Epub 2007/10/19. 10.1158/0008-5472.CAN-07-0625 .17942911 14 Mimura N , Fulciniti M , Gorgun G , Tai YT , Cirstea D , Santo L , et al Blockade of XBP1 splicing by inhibition of IRE1alpha is a promising therapeutic option in multiple myeloma . Blood . 2012 ;119 (24 ):5772 –81 . Epub 2012/04/28. 10.1182/blood-2011-07-366633 22538852 15 Huang H , Liu H , Liu C , Fan L , Zhang X , Gao A , et al Disruption of the unfolded protein response (UPR) by lead compound selectively suppresses cancer cell growth . Cancer letters . 2015 ;360 (2 ):257 –68 . Epub 2015/02/28. 10.1016/j.canlet.2015.02.029 .25721085 16 Schonthal AH . Targeting endoplasmic reticulum stress for cancer therapy . Frontiers in bioscience . 2012 ;4 :412 –31 . Epub 2011/12/29. .22202068 17 Pyrko P , Kardosh A , Liu YT , Soriano N , Xiong W , Chow RH , et al Calcium-activated endoplasmic reticulum stress as a major component of tumor cell death induced by 2,5-dimethyl-celecoxib, a non-coxib analogue of celecoxib . Molecular cancer therapeutics . 2007 ;6 (4 ):1262 –75 . Epub 2007/04/14. 10.1158/1535-7163.MCT-06-0629 .17431104 18 Pyrko P , Kardosh A , Wang W , Xiong W , Schonthal AH , Chen TC . HIV-1 protease inhibitors nelfinavir and atazanavir induce malignant glioma death by triggering endoplasmic reticulum stress . Cancer research . 2007 ;67 (22 ):10920 –8 . Epub 2007/11/17. 10.1158/0008-5472.CAN-07-0796 .18006837 19 Johnson GG , White MC , Grimaldi M . Stressed to death: targeting endoplasmic reticulum stress response induced apoptosis in gliomas . Current pharmaceutical design . 2011 ;17 (3 ):284 –92 . Epub 2011/02/26. 21348829 20 Johnson GG , White MC , Wu JH , Vallejo M , Grimaldi M . The deadly connection between endoplasmic reticulum, Ca2+, protein synthesis, and the endoplasmic reticulum stress response in malignant glioma cells . Neuro-oncology . 2014 ;16 (8 ):1086 –99 . Epub 2014/02/27. 10.1093/neuonc/nou012 24569545 21 White MC , Johnson GG , Zhang W , Hobrath JV , Piazza GA , Grimaldi M . Sulindac sulfide inhibits sarcoendoplasmic reticulum Ca2+ ATPase, induces endoplasmic reticulum stress response, and exerts toxicity in glioma cells: relevant similarities to and important differences from celecoxib . Journal of neuroscience research . 2013 ;91 (3 ):393 –406 . Epub 2013/01/03. 10.1002/jnr.23169 23280445 22 Seethala R , Zhang L . Handbook of drug screening . 2nd ed. New York : Informa Healthcare ; 2009 xiii , 489 p. p. 23 Inglese J , Auld DS , Jadhav A , Johnson RL , Simeonov A , Yasgar A , et al Quantitative high-throughput screening: a titration-based approach that efficiently identifies biological activities in large chemical libraries . Proceedings of the National Academy of Sciences of the United States of America . 2006 ;103 (31 ):11473 –8 . Epub 2006/07/26. 10.1073/pnas.0604348103 16864780 24 Yamamoto K , Sato T , Matsui T , Sato M , Okada T , Yoshida H , et al Transcriptional induction of mammalian ER quality control proteins is mediated by single or combined action of ATF6alpha and XBP1 . Developmental cell . 2007 ;13 (3 ):365 –76 . Epub 2007/09/04. 10.1016/j.devcel.2007.07.018 .17765680 25 Wang Y , Shen J , Arenzana N , Tirasophon W , Kaufman RJ , Prywes R . Activation of ATF6 and an ATF6 DNA binding site by the endoplasmic reticulum stress response . The Journal of biological chemistry . 2000 ;275 (35 ):27013 –20 . Epub 2000/06/17. 10.1074/jbc.M003322200 .10856300 26 Yoshida H , Matsui T , Yamamoto A , Okada T , Mori K . XBP1 mRNA is induced by ATF6 and spliced by IRE1 in response to ER stress to produce a highly active transcription factor . Cell . 2001 ;107 (7 ):881 –91 . 10.1016/S0092-8674(01)00611-0 .11779464 27 Free RB , Chun LS , Moritz AE , Miller BN , Doyle TB , Conroy JL , et al Discovery and characterization of a G protein-biased agonist that inhibits beta-arrestin recruitment to the D2 dopamine receptor . Molecular pharmacology . 2014 ;86 (1 ):96 –105 . Epub 2014/04/24. 10.1124/mol.113.090563 24755247 28 Binder ZA , Wilson KM , Salmasi V , Orr BA , Eberhart CG , Siu IM , et al Establishment and Biological Characterization of a Panel of Glioblastoma Multiforme (GBM) and GBM Variant Oncosphere Cell Lines . PloS one . 2016 ;11 (3 ):e0150271 Epub 2016/03/31. 10.1371/journal.pone.0150271 27028405 29 Hongisto V , Jernstrom S , Fey V , Mpindi JP , Kleivi Sahlberg K , Kallioniemi O , et al High-throughput 3D screening reveals differences in drug sensitivities between culture models of JIMT1 breast cancer cells . PloS one . 2013 ;8 (10 ):e77232 Epub 2013/11/07. 10.1371/journal.pone.0077232 24194875 30 Kunz-Schughart LA , Freyer JP , Hofstaedter F , Ebner R . The use of 3-D cultures for high-throughput screening: the multicellular spheroid model . Journal of biomolecular screening . 2004 ;9 (4 ):273 –85 . Epub 2004/06/12. 10.1177/1087057104265040 .15191644 31 Griner LAM , Guha R , Shinn P , Young RM , Keller JM , Liu D , et al High-throughput combinatorial screening identifies drugs that cooperate with ibrutinib to kill activated B-cell-like diffuse large B-cell lymphoma cells . Proceedings of the National Academy of Sciences of the United States of America . 2014 ;111 (6 ):2349 –54 . 10.1073/pnas.1311846111 .24469833 32 Mott BT , Eastman RT , Guha R , Sherlach KS , Siriwardana A , Shinn P , et al High-throughput matrix screening identifies synergistic and antagonistic antimalarial drug combinations . Scientific reports . 2015 ;5 :13891 Epub 2015/09/26. 10.1038/srep13891 26403635 33 Kudo T , Kanemoto S , Hara H , Morimoto N , Morihara T , Kimura R , et al A molecular chaperone inducer protects neurons from ER stress . Cell Death Differ . 2008 ;15 (2 ):364 –75 . Epub 2007/12/01. 10.1038/sj.cdd.4402276 .18049481 34 Bi K , Nishihara K , Machleidt T , Hermanson S , Wang J , Sakamuru S , et al Identification of known drugs targeting the endoplasmic reticulum stress response . Analytical and bioanalytical chemistry . 2015 ;407 (18 ):5343 –51 . Epub 2015/05/01. 10.1007/s00216-015-8694-2 .25925857 35 Flaherty DP , Miller JR , Garshott DM , Hedrick M , Gosalia P , Li Y , et al Discovery of Sulfonamidebenzamides as Selective Apoptotic CHOP Pathway Activators of the Unfolded Protein Response . ACS medicinal chemistry letters . 2014 ;5 (12 ):1278 –83 . Epub 2014/12/23. 10.1021/ml5003234 25530830 36 Ciechomska IA , Gabrusiewicz K , Szczepankiewicz AA , Kaminska B . Endoplasmic reticulum stress triggers autophagy in malignant glioma cells undergoing cyclosporine a-induced cell death . Oncogene . 2013 ;32 (12 ):1518 –29 . Epub 2012/05/15. 10.1038/onc.2012.174 .22580614 37 Zhong J , Kong X , Zhang H , Yu C , Xu Y , Kang J , et al Inhibition of CLIC4 enhances autophagy and triggers mitochondrial and ER stress-induced apoptosis in human glioma U251 cells under starvation . PloS one . 2012 ;7 (6 ):e39378 Epub 2012/07/05. 10.1371/journal.pone.0039378 22761775 38 Kanzawa T , Germano IM , Komata T , Ito H , Kondo Y , Kondo S . Role of autophagy in temozolomide-induced cytotoxicity for malignant glioma cells . Cell Death Differ . 2004 ;11 (4 ):448 –57 . Epub 2004/01/10. 10.1038/sj.cdd.4401359 .14713959 39 Ogata M , Hino S , Saito A , Morikawa K , Kondo S , Kanemoto S , et al Autophagy is activated for cell survival after endoplasmic reticulum stress . Molecular and cellular biology . 2006 ;26 (24 ):9220 –31 . Epub 2006/10/13. 10.1128/MCB.01453-06 17030611 40 Yorimitsu T , Klionsky DJ . Endoplasmic reticulum stress: a new pathway to induce autophagy . Autophagy . 2007 ;3 (2 ):160 –2 . Epub 2007/01/06. .17204854 41 Ma Y , Brewer JW , Diehl JA , Hendershot LM . Two distinct stress signaling pathways converge upon the CHOP promoter during the mammalian unfolded protein response . Journal of molecular biology . 2002 ;318 (5 ):1351 –65 . Epub 2002/06/27. .12083523 42 Sun S , Lee D , Ho AS , Pu JK , Zhang XQ , Lee NP , et al Inhibition of prolyl 4-hydroxylase, beta polypeptide (P4HB) attenuates temozolomide resistance in malignant glioma via the endoplasmic reticulum stress response (ERSR) pathways . Neuro-oncology . 2013 ;15 (5 ):562 –77 . Epub 2013/02/28. 10.1093/neuonc/not005 23444257 43 Shoemaker RH . The NCI60 human tumour cell line anticancer drug screen . Nature reviews Cancer . 2006 ;6 (10 ):813 –23 . Epub 2006/09/23. 10.1038/nrc1951 .16990858 44 Honma M , Izumi M , Sakuraba M , Tadokoro S , Sakamoto H , Wang W , et al Deletion, rearrangement, and gene conversion; genetic consequences of chromosomal double-strand breaks in human cells . Environmental and molecular mutagenesis . 2003 ;42 (4 ):288 –98 . Epub 2003/12/16. 10.1002/em.10201 .14673874 45 Dexheimer TS , Rosenthal AS , Luci DK , Liang Q , Villamil MA , Chen J , et al Synthesis and structure-activity relationship studies of N-benzyl-2-phenylpyrimidin-4-amine derivatives as potent USP1/UAF1 deubiquitinase inhibitors with anticancer activity against nonsmall cell lung cancer . Journal of medicinal chemistry . 2014 ;57 (19 ):8099 –110 . Epub 2014/09/18. 10.1021/jm5010495 25229643

PMC005xxxxxx/PMC5003375.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 10.1371/journal.pone.0161736PONE-D-16-21376Research ArticleEngineering and TechnologyManufacturing ProcessesHeat TreatmentPhysical SciencesChemistryChemical ElementsTinPhysical SciencesMaterials ScienceMaterials by AttributeInsulatorsDielectricsPhysical SciencesMaterials ScienceMaterials by StructureSemiconductorsEngineering and TechnologyElectronicsPhysical SciencesChemistryChemical ElementsTitaniumEarth SciencesAtmospheric ScienceAtmospheric PhysicsAtmospheric LayersEarth SciencesGeophysicsAtmospheric PhysicsAtmospheric LayersPhysical SciencesPhysicsGeophysicsAtmospheric PhysicsAtmospheric LayersPhysical SciencesMaterials ScienceMaterial PropertiesCapacitanceEffect of Thermal Budget on the Electrical Characterization of Atomic Layer Deposited HfSiO/TiN Gate Stack MOSCAP Structure Electrical Characterization of Atomic Layer Deposited HfSiO/TiN Gate Stack MOSCAP StructureKhan Z. N. 12*Ahmed S. 1Ali M. 11 Advanced Electronics Labs (AEL), International Islamic University, Islamabad, Pakistan2 Comsats Institute of Information Technology, Islamabad, PakistanLisesivdin Sefer Bora EditorGazi Universitesi, TURKEYCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: SA ZNK. Data curation: Ali ZNK. Formal analysis: SA ZNK MA. Investigation: ZNK SA MA. Methodology: ZNK SA. Project administration: ZNK. Resources: ZNK MA. Software: ZNK. Supervision: SA. Validation: ZNK. Visualization: ZNK. Writing – original draft: ZNK SA MA. Writing – review & editing: ZNK SA MA. * E-mail: zeeshan.najam@comsats.edu.pk29 8 2016 2016 11 8 e016173627 5 2016 10 8 2016 © 2016 Khan et al2016Khan et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Metal Oxide Semiconductor (MOS) capacitors (MOSCAP) have been instrumental in making CMOS nano-electronics realized for back-to-back technology nodes. High-k gate stacks including the desirable metal gate processing and its integration into CMOS technology remain an active research area projecting the solution to address the requirements of technology roadmaps. Screening, selection and deposition of high-k gate dielectrics, post-deposition thermal processing, choice of metal gate structure and its post-metal deposition annealing are important parameters to optimize the process and possibly address the energy efficiency of CMOS electronics at nano scales. Atomic layer deposition technique is used throughout this work because of its known deposition kinetics resulting in excellent electrical properties and conformal structure of the device. The dynamics of annealing greatly influence the electrical properties of the gate stack and consequently the reliability of the process as well as manufacturable device. Again, the choice of the annealing technique (migration of thermal flux into the layer), time-temperature cycle and sequence are key parameters influencing the device’s output characteristics. This work presents a careful selection of annealing process parameters to provide sufficient thermal budget to Si MOSCAP with atomic layer deposited HfSiO high-k gate dielectric and TiN gate metal. The post-process annealing temperatures in the range of 600°C -1000°C with rapid dwell time provide a better trade-off between the desirable performance of Capacitance-Voltage hysteresis and the leakage current. The defect dynamics is thought to be responsible for the evolution of electrical characteristics in this Si MOSCAP structure specifically designed to tune the trade-off at low frequency for device application. No specific funding has been received for the project. Data AvailabilityAll relevant data are contained within the manuscript and its Supporting Information files.Data Availability All relevant data are contained within the manuscript and its Supporting Information files. ==== Body Introduction Titanium nitride (TiN) is a potential application specific material, which provides a favourable combination of physical and chemical properties for semiconductor manufacturing such as low resistance, comparatively high transmittance within the visible spectrum, hardness and chemical resistance [1]. Titanium nitride also possesses good quality mechanical, tribological, electrical, biomedical, and optical properties. The reliance of the mechanical and electrical properties of titanium nitride thin films deposited with the help of variety of techniques upon silicon substrates have been studied in past to exhibit their potential for CMOS industry driven applications [2]. In some previous studies [3–4], the TiN-Si hetrojunctions were fabricated by the deposition of TiN thin films onto the refined Si substrates by using the DC reactive magnetron sputtering. These nitrides have also shown to exhibit the property to tune the work function due to their thermal stability [5]. Currently, research on various electrode materials, starting from polycrystalline metal films to composite conducting oxides, is being carried out [5]. With TiN placed effectively as ultra-thin layer or part of the compositional stack in Si CMOS processing. The thermal stability and electrical properties of one such stack in metal oxide semiconductor (MOS) capacitors after post-metal annealing in H2 or N2+H2 ambient in a range of temperature 30 minutes are examined [6] and analyzed. C-V profiling for characterization of these nano-scale devices has been a standard analytical technique in manufacturing industry [7]. NMOS C-V profiling techniques (the conventional Capacitance-Voltage profiling and deep depletion (DD) C-V have also been in practice to profile the subject devices [8]. Similarly, optical characterization of titanium nitride is accomplished by means of ellipsometry, and electrical characterization is performed using C-V and I-V profiling to obtain the work function and resistivity of these thin sensitive films [6]. In this paper, we investigate the MOSCAP structure as part of the CMOS family in terms of its sensitive electrical characteristics including detailed I-V and C-V analysis. The study is targeted towards TiN ultra-thin film as the gate electrode in the MOSCAP structure. As a substitute to the conventional gate dielectric (SiO2), high-k HfSiO is used as dielectric material beneath the gate metal for the device structure as shown in Fig 1. Hafnium Silicate has proven to be a promising dielectric material with low diffusivity of impurities and channel mobility features. 10.1371/journal.pone.0161736.g001Fig 1 MOSCAP Structure used in this work (Device Cross Section). Experimental MOSCAP structure is fabricated by utilizing HfSiO layer as high-k dielectric material instead of conventional SiO2 knowing its competitiveness of EOT scaling. Atomic layer deposition (ALD) technique is used for the deposition of both the high-k dielectric HfSiO layer and TiN metal gate in NLD-4000 ALD system. The deposition chamber is preheated at 315°C and the substrate of Si (100) is placed in the reaction chamber. A high -k dielectric layer is deposited by injecting the Hf and Si liquid precursors one by one resulting into a proper reaction yielding a fine layer of dielectric material. Following this, TiN layer is deposited on top of the dielectric layer onto the same wafer using the same ALD technique. The chamber temperature is maintained at 400°C with TiCl4 and NH3 gases duly injected in the reaction chamber for TiN ultra-thin deposition. TiCl4 is injected with N2 which is used as a carrier gas (also acted as a purge gas) at the same time. High purity ammonia gas (~99%) is used as a reactant. The working pressure of the reaction chamber was maintained at 0.55 Torr (with ±5% of control accuracy). Detailed ALD process parameters to deposit these ultra- thin layers on Si (100) standard wafer are summarized in Table 1. 10.1371/journal.pone.0161736.t001Table 1 Atomic Layer Deposition Process Conditions for HfSiO and TiN Ultra-Thin Layers to fabricate the device structure. S.No Process Material Precursors Non Metal precursor Growth Rate per Cycle (at Source Temperature) Temperature (°C) Deposition Rate (at Source Temperature) Pulse sequence (seconds) 1 HfSiO TEMAH/TDMAS/O2 (for Hf, Si and O respectively) O2 0.43 Å/cycle 300 0.24 nm/min 1/2/2/2 (Precursor/Ar Purge/O2 Plasma/Ar Purge) 2 TiN TiCl4 NH3 0.3 Å/cycle 400 0.11 nm/min 2/2/2/2 (Precursor/N2/ NH3/N2) All the annealing schedules are performed on the Rapid Thermal Processor (RTP) during the processing of MOSCAP structures. Rapid Thermal Processor (RTP) is known to reduce the thermal redistribution of impurities at relatively high temperature. The exposure of the samples to the heating flow for relatively smaller times also tends to reduce the probabilities of oxidation of TiN films. Post deposition annealing of HfSiO is carried out in RTP for 470°C with no ramps of temperature in between and dwelled at 30 seconds. The ambience of nitrogen is chosen to purge the environment with a cooling effect. The atomic layer deposited TiN ultra-thin layer is also subjected to RTP with temperatures intentionally chosen to be in the process window from 600°C to 1000°C for 20 seconds in each case. Again, a nitrogen gas ambience is created to anneal the samples at nominal atmospheric pressure. The electrical characteristics of the MOSCAP devices are measured using a sophisticated Automatic System for Material Electro-Physical Characterization (ASMEC) tool in the Advanced Electronics Laboratory. Results and Discussion I-V Characteristics and Leakage Resistance Fig 2 shows the I-V profiles of the MOSCAP structure with HfSiO as dielectric material TiN and ultra-thin film as gate electrode, and subjected to relatively medium to slightly higher annealing temperatures. Fig 2, when closely examined, reveals that the leakage current is reduced with the increasing post-process annealing cycle until 900°C. A slight increase is noticed when the annealing temperature goes to 1000°C. This in turn shows an increase of leakage resistance with the increasing annealing temperatures. At relatively higher annealing temperatures (900–1000°C of process window); the change in the leakage resistance is almost negligible. The current profile, as depicted in Fig 2 is at maximum for the sample annealed at a temperature of 600°C. The obtained curves follow the pattern as in [9–10], which is a similar study for I-V curves for evaluating the metal gate electrode behavior. Our results are consistent with those obtained in [11–12] since the rise of temperature does not necessarily transform into enhanced conductivity; this phenomenon can be attributed to metal gate and dielectric interface defect states [13]. We also observe from our results that the rise of the annealing temperature also improves the linearity of the I-V correlation i.e. the linearity is much pronounced for the devices undergone the annealing temperatures of 800°C, 900°C and 1000°C in comparison to the ones with 600°C and 700°C. The possible presence of defects in the structure and randomness of the defect (vacancies in particular) interaction with the device matrix and their annihilation is thought to be responsible for the non-linearity and its gradual recovery in the signature I-V curves for various annealing conditions (Fig 2b–2f). Fig 2 also depicts that the MOSCAP structure having TiN thin layers deposited by ALD and undergone a rapid thermal process cycle at 600°C for 20 seconds does not maintain the leakage current against all bias values. For example, low voltages yield an appreciable leakage current. In other words; leakage resistance is maximum at low voltages for sample annealed at 600°C. As the voltage increases, the resistance shows a constant behavior. This provides a flexible process window to attain device output parameters at desirable bias condition. 10.1371/journal.pone.0161736.g002Fig 2 Leakage Resistance and I-V characteristics for different annealing temperatures. (a) Leakage Resistance at different annealing temperatures. (b) I-V characteristic curve for 600°C. (c) I-V characteristic curve for 700°C. (d) I-V characteristic curve for 800°C. (e) I-V characteristic curve for 900°C. (f) I-V characteristic curve for 1000°C. C-V Profiling An analysis is made for the electrical characteristics of MOSCAP with TiN gate electrode at various annealing temperatures. The Capacitance-Voltage profiling is done with the help of ASMEC equipment, which is used for the electrical characterization for parameters such as I-V, C-V, QDLTS etc. The C-V profiles obtained for MOSCAP devices undergoes different annealing cycles between 600°C and 1000°C are shown in Fig 3. The C-V curves shown in Fig 3 are in close approximation with the standard curve for MOSCAP found in the literature [7]. It is evident that the values of capacitances change with the variation of the annealing cycles subjected to the MOSCAP device structure. The devices, which have undergone the rapid thermal processing of 600°C and 700°C, reflect a positive threshold voltage as compared to the ones with negative voltage for higher annealing temperatures. The threshold voltage becomes more negative as soon as the temperature rises beyond 900°C. This signifies that the slope of the representing curves (depletion in physical terms) decreases with the increase of the annealing temperatures. It is also exhibited that the samples annealed at 800°C have the lowest value of capacitance i.e. 65 pF at a voltage of about 0.5 V. The lower value of capacitance translates into lower electrostatic losses and consequently lower leakages. As the C-V profiling executed for these MOSCAP devices is performed at an operating frequency of 1.2 kHz, this may provide an estimate of the possible presence of different kinds of defects and interstitial states present in the structure. The trends are somewhat similar to an earlier study carried out at low frequencies [12]. The increase in capacitance especially at low frequency is likely to be attributed to the presence of interstitial state at Si/SiO2 interface in conventional MOSCAP structures [13]. The dangling bonds, humps, stretch out oxygen vacancies during deposition are also reported to play a significant role in the shift of flat band voltage for such devices [14]. The post process annealing has known influence on the dynamic re-structuring of defects and is reported to improve the mobility and threshold slopes for equivalent devices [14], such as in our case. 10.1371/journal.pone.0161736.g003Fig 3 C-V Hysteresis for samples at different annealing Temperatures. (a) C-V profiles at different annealing ranging from 600 to 1000°C. (b) C-V profiling for 600°C. (c) C-V profiling for 700°C. (d) C-V profiling for 800°C. (e) C-V profiling for 900°C. (f) C-V profiling for 1000°C. Doping Profile and Built in Potential Fig 4 explains a correlation between the post-process annealing temperature with the effective doping profile (Nd-Na) and the built-in potential for MOSCAP with ultra-thin TiN metal gate structured with ALD. The doping profile Nd-Na is visible to be highest for samples annealed at 900°C. The built in voltage Vb is also the highest for the 900°C annealed samples owing to the higher annealing temperature drawing out relatively larger number of majority carriers from the lattice. 10.1371/journal.pone.0161736.g004Fig 4 Doping profile and Built in voltage as a function of Annealing Temperature. The built in voltage depends directly on the number of dopants in such a way that higher number of dopant atoms yield a concentrated doping profile and a wider depletion region. Consequently, the higher built in voltage is achieved. The higher annealing temperature, as expected, results in fewer interfaces state densities and hence the probability of having defects decreases in the interface state and consequently lesser energy is required to vacate any carrier implying energy efficiency [9]. Conclusion We fabricated specifically designed MOSCAP structures, for possible integration with CMOS technology, with atomic layer deposited HfSiO and TiN thin layers as high-k dielectric and metal gate, respectively. The devices were subjected to post deposition and post-metal annealing cycles with carefully chosen rapid thermal process parameters. Low frequency electrical characterization was performed in order to tune the trade-off between the C-V hysteresis and the leakage current at the gate to facilitate design an effective energy efficient CMOS nano-electronics. It is found that the profiles obtained after different annealing temperatures are shown to closely follow the standard C-V curve for MOSCAP at low frequencies. The largest value (voltage) of the knee of the curve signifying transition from accumulation to depletion is for a sample annealed at 600°C. The slope (depletion) of the electrical profile of the device structure is found to decrease with the increasing rapid thermal annealing temperature. The leakage resistance reflects a maximum value at low voltages with almost a constant behavior with increasing bias. The threshold voltages are proving to be more negative at relatively higher annealing temperatures. The optimized device output characteristic window provides a trade-off between the desirable electrical entities such as C-V hysteresis, leakage current, flat-band voltage and leakage resistance, with the control of process conditions during the annealing kinetics. This may prove to be a way forward to design energy efficient CMOS electronics at desirable bias. Supporting Information S1 Fig MOSCAP Structure used in this work (Device Cross Section). (TIF) Click here for additional data file. S2 Fig Leakage Resistance and I-V characteristics for different annealing temperatures. (TIF) Click here for additional data file. S3 Fig Leakage Resistance and I-V characteristics for 600°C–1000°C. (TIF) Click here for additional data file. S4 Fig I-V characteristic curve for 600°C. (TIF) Click here for additional data file. S5 Fig I-V characteristic curve for 700°C. (TIF) Click here for additional data file. S6 Fig I-V characteristic curve for 800°C. (TIF) Click here for additional data file. S7 Fig I-V characteristic curve for 900°C. (TIF) Click here for additional data file. S8 Fig I-V characteristic curve for 1000°C. (TIF) Click here for additional data file. S9 Fig C-V Hysteresis for samples at different annealing Temperatures. (TIF) Click here for additional data file. S10 Fig C-V profiles for samples at annealing temperatures (600°C–1000°C). (TIF) Click here for additional data file. S11 Fig C-V profiling for samples at 600°C. (TIF) Click here for additional data file. S12 Fig C-V profiling for samples at 700°C. (TIF) Click here for additional data file. S13 Fig C-V profiling for samples at 800°C. (TIF) Click here for additional data file. S14 Fig C-V profiling for samples at 900°C. (TIF) Click here for additional data file. S15 Fig C-V profiling for samples at 1000°C. (TIF) Click here for additional data file. S16 Fig Doping profile and Built in voltage as a function of Annealing Temperature. (TIF) Click here for additional data file. S1 Table Atomic Layer Deposition Process Conditions for HfSiO and TiN Ultra-Thin Layers to fabricate the device structure. (DOCX) Click here for additional data file. ==== Refs References 1 Gagnon G , Currie JF , Beique G , Brebner JL , Gujrathi SC , Ouellet L . Characterization of reactively evaporated TiN layers for diffusion barrier applications . Journal of applied physics . 1994 2 1 ; 75 (3 ):1565 –70 . 2 Khojier K , Savaloni H , Shokrai E , Dehghani Z , Dehnavi NZ . Influence of argon gas flow on mechanical and electrical properties of sputtered titanium nitride thin films . Journal of Theoretical and Applied Physics . 2013 12 1 ; 7 (1 ):1 –6 . 3 Solovan MN , Brus VV , Maryanchuk PD , Kovalyuk TT , Rappich J , Gluba M . Kinetic properties of TiN thin films prepared by reactive magnetron sputtering . Physics of the Solid State . 2013 11 1 ; 55 (11 ):2234 –8 . 4 Solovan M , Brus V , Maryanchuk P . Electrical and photoelectric properties of anisotype n-TiN/p-Si heterojunctions . Semiconductors . 2013 ; 47 (9 ):1174 –1179 . 5 Xueli M , Hong Y , Wenwu W , Huaxiang Y , Huilong Z , Chao Z ,et al The effects of process condition of top-TiN and TaN thickness on the effective work function of MOSCAP with high-k/metal gate stacks . Journal of Semiconductors . 2014 10 ; 35 (10 ):106002 . 6 Kai H , Xueli M , Jinjuan X , Hong Y , Wenwu W . Effect of low temperature annealing on the electrical properties of an MOS capacitor with a HfO2 dielectric and a TiN metal gate . Journal of Semiconductors . 2013 11 ; 34 (11 ):114007 . 7 Cristea MJ . Capacitance-voltage profiling techniques for characterization of semiconductor materials and device . Emerging Trends in Electrical, Electronics and Instrumentation Engineering: An international Journal (EEIEJ) 2014 8 :1 (3 ):29 –38 . 8 Bulucea C . Investigation of deep-depletion régime of mos structures using ramp-response method . Electron Lett . 1970 ; 6 (15 ):479 –481 . 9 Chen Z, Ong P, Samantaray CB. Large leakage current reduction of silicon oxide and high-k oxides using the phonon-energy-coupling enhancement effect. In Semiconductor Device Research Symposium, 2007 International 2007 Dec 12 (pp. 1–2). IEEE. 10 Chen Z . Mechanism for generation of the phonon-energy-coupling enhancement effect for ultrathin oxides on silicon . Applied Physics Letters . 2007 11 26 ; 91 (22 ):223513(1–3). 11 Wang Y. Leakage Current Reduction of MOS Capacitor Induced by Rapid Thermal Processing. Master’s Thesis, University of Kentucky. 2010. Available: http://uknowledge.uky.edu/cgi/viewcontent.cgi?article=1644&context=gradschool_theses. 12 Tataroğlu A , Altındal Ş . Analysis of electrical characteristics of Au/SiO/n-Si (MOS) capacitors using the high–low frequency capacitance and conductance methods . Microelectronic Engineering . 2008 11 30 ; 85 (11 ):2256 –60 . 13 Aydoğan Ş , Sağlam M , Türüt A . Effect of temperature on the capacitance–frequency and conductance–voltage characteristics of polyaniline/p-Si/Al MIS device at high frequencies . Microelectronics Reliability . 2012 7 31 ; 52 (7 ):1362 –6 . 14 Zhao C , Zhao CZ , Werner M , Taylor S , Chalker PR . Advanced CMOS gate stack: Present research progress . ISRN Nanotechnology . 2012 2 9 ; 2012 (2012 ):689023 .

PMC005xxxxxx/PMC5003376.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757151410.1371/journal.pone.0160998PONE-D-16-11233Research ArticleMedicine and Health SciencesHealth CareQuality of LifeMedicine and Health SciencesPublic and Occupational HealthGlobal HealthMedicine and Health SciencesOncologyMedicine and Health SciencesOncologyCancer TreatmentMedicine and Health SciencesClinical MedicineClinical TrialsRandomized Controlled TrialsMedicine and Health SciencesPharmacologyDrug Research and DevelopmentClinical TrialsRandomized Controlled TrialsResearch and Analysis MethodsClinical TrialsRandomized Controlled TrialsBiology and Life SciencesToxicologyToxicityMedicine and Health SciencesPathology and Laboratory MedicineToxicologyToxicityResearch and Analysis MethodsResearch AssessmentSystematic ReviewsBiology and Life SciencesNeuroscienceCognitive ScienceCognitionDecision Making‘Trial Exegesis’: Methods for Synthesizing Clinical and Patient Reported Outcome (PRO) Data in Trials to Inform Clinical Practice. A Systematic Review Methods to Define Clinical 'Trial Exegesis'http://orcid.org/0000-0002-2601-9258McNair Angus G. K. 12*Macefield Rhiannon C. 1Blencowe Natalie S. 12Brookes Sara T. 1Blazeby Jane M. 121 Centre for Surgical Research, School of Social and Community Medicine, University of Bristol, Canynge Hall, Bristol, United Kingdom2 Division of Surgery, Head and Neck, University Hospitals Bristol NHS Foundation Trust, Bristol, United KingdomHills Robert K EditorCardiff University, UNITED KINGDOMCompeting Interests: The authors have declared that no competing interests exist. Conceived and designed the experiments: AGKM RCM NSB STB JMB. Performed the experiments: AGKM RCM NSB STB JMB. Analyzed the data: AGKM RCM NSB STB JMB. Contributed reagents/materials/analysis tools: AGKM RCM NSB STB JMB. Wrote the paper: AGKM RCM NSB STB JMB. * E-mail: angus.mcnair@bristol.ac.uk29 8 2016 2016 11 8 e016099818 4 2016 28 7 2016 © 2016 McNair et al2016McNair et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Purpose The CONSORT extension for patient reported outcomes (PROs) aims to improve reporting, but guidance on the optimal integration with clinical data is lacking. This study examines in detail the reporting of PROs and clinical data from randomized controlled trials (RCTs) in gastro-intestinal cancer to inform design and reporting of combined PRO and clinical data from trials to improve the ‘take home’ message for clinicians to use in practice. Materials and Methods The case study was undertaken in gastro-intestinal cancer trials. Well-conducted RCTs reporting PROs with validated instruments were identified and categorized into those combining PRO and clinical data in a single paper, or those separating data into linked primary and supplemental papers. Qualitative methods were developed to examine reporting of the critical interpretation of the trial results (trial exegesis) in the papers in relation of the PRO and clinical outcomes and applied to each publication category. Results were used to inform recommendations for practice. Results From 1917 screened abstracts, 49 high quality RCTs were identified reported in 36 combined and 15 linked primary and supplemental papers. In-depth analysis of manuscript text identified three categories for understanding trial exegesis: where authors reported a “detailed”, “general”, or absent PRO rationale and integrated interpretation of clinical and PRO results. A total of 11 (30%) and 6 (16%) combined papers reported “detailed” PRO rationale and integrated interpretation of results although only 2 (14%) and 1 (7%) primary papers achieved the same standard respectively. Supplemental papers provide better information with 11 (73%) and 3 (20%) achieving “detailed” rationale and integrated interpretation of results. Supplemental papers, however, were published a median of 20 months after the primary RCT data in lower impact factor journals (median 16.8 versus 5.2). Conclusion It is recommended that single papers, with detailed PRO rationale and integrated PRO and clinical data are published to optimize trial exegesis. Further work to examine whether this improves the use of PRO data to inform practice is needed. http://dx.doi.org/10.13039/501100000265Medical Research CouncilMR/K025643/1Blazeby Jane M Support was provided by United Kingdom Medical Research Council Collaboration and innovation for Difficult and Complex randomised controlled Trials In Invasive procedures Hub grant code MR/K025643/1 director (to JMB). The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction The updated Consolidated Standards of Reporting Trials (CONSORT) extension for patient reported outcomes (PROs) aims to facilitate the use of PRO data in health policy and practice through the transparent reporting of PROs in randomized controlled trials (RCTs)[1]. It makes recommendations for reporting of PRO instrument validity, presentation and handling of missing data and reporting of PRO sample size calculations, data analyses and results within the main text and abstract. The statement endorses reporting the rationale/hypotheses for PRO assessment and it highlights the need for integrated reporting of the PROs with the clinical findings of the paper (extensions and elaborations 2a, 2b and P20/21 and 22). These latter recommendations are particularly essential for critical interpretation of the trial, so called exegesis or a “take-home message”, for clinicians to understand and use results in clinical practice. The CONSORT extension provides illustrations of how to report these issues. For example, elaboration 2a and extension P2b state that authors should “briefly establish the rationale for including PROs and why specific outcomes were selected”, and “report the rationale for the selection of specific patient-reported outcomes”. Furthermore, the guidelines state (in items P20/21 and elaboration 22) that “the clinical significance of PRO results is often not discussed in RCT reports but should be interpreted in relation to other important clinical outcomes such as survival”. Whilst this is helpful, the level of detail required for reporting the PRO data is unclear and this may have a detrimental impact on the overall clinically relevant trial conclusions. This problem is further compounded by the way that PRO data are published. Some trials publish results in a single paper combined with clinical findings, which may limit full explanation of PRO results in the context of finite manuscript word limits. Publication of clinical and PROs separately is therefore attractive, however, practicing oncologists may be less likely to read the supplemental paper and thus not use PRO data in decision-making. The PRO CONSORT statement does not provide guidance for PRO reporting within these different scenarios. Whether PRO data are published together with clinical outcomes or separately in two articles, there is a need for optimal reporting of clinical and PROs so that they can be used in clinical practice. The aim of this paper, therefore, is to explore current standards and make recommendations for reporting a combined PRO and clinical ‘take home’ message to use in reporting RCTs in oncology. Materials and Methods This study was conducted in two parts. Part 1: systematic identification of well-designed and conducted RCTs reporting PROs with validated instruments, categorisation of papers into combined PRO and clinical reports or linked primary and supplemental reports, and assessment of PRO reporting within each RCT (PRO CONSORT extension). Part 2: development of novel methods to examine the ‘take home/trial exegesis’ message of the RCT and application to the papers identified in Part 1. Part 1(a) Identification of well-designed and conducted RCTs reporting PROs with validated instruments Systematic review methodology was used to identify RCTs at a low risk of bias reporting PROs with validated instruments in radical treatments of gastro intestinal oncology. Full-text articles were obtained. Gastro-intestinal oncology trials of radical treatment were chosen because the research team were familiar with the clinical and PRO data in this area, and trials at a low risk of bias are examples of best practice. Electronic searches were performed in MEDLINE, Embase and Cochrane databases using the OVID SP gateway and Cochrane library. Search terms for esophageal, gastric and colorectal cancer were combined, as were terms for chemotherapy, radiotherapy, surgery or combined treatment. Results were restricted with the application of terms for “randomized clinical trials” and “patient reported outcomes”, and limited to articles published between January 2000 and October 2012 (see full search in S1 Appendix). The search output was imported into Reference Manager software and duplicate records removed. References for relevant studies before the year 2000 were obtained from a previous systematic review [2]. Titles and abstracts were screened by two researchers (AGKM and RM). Serial publications for the same trial (e.g. articles reporting short and long term PROs) were included. Excluded were phase II studies, RCTs of endoscopic and non-biomedical interventions, or trials limited to palliative treatment, screening or premalignant conditions. Only English-language publications were considered. Articles were assessed for risk of bias in the trials by three researchers (AGKM, RM, NB) using the Cochrane tool [3]. Studies classified with potential high or unascertainable risk of bias were excluded. Independent data extraction was conducted by at least two reviewers (AGKM, NB, RM, JMB) using a pre-designed and piloted form. Details of the trial were recorded including disease site, treatment intervention, primary and secondary outcomes, number of participants, and main trial results. The systematic review PRISMA checklist is presented in S2 Appendix. (b) Categorisation of papers into combined PRO and clinical reports or linked primary and supplemental reports Where included papers indicated the presence of previously published results from the same trial, these additional papers were sought and included in the analysis. These linked papers are hereafter considered in the order by which they were published, with those published first and second defined as “primary” and “supplemental” respectively. Thus, trials were categorized into those reporting PROs and clinical outcomes in a single, combined paper and those reporting results in linked primary and supplemental papers. The journal impact factor (Thomson Reuters, 2012) and the date of publication for each paper were recorded. Descriptive statistics compared journal impact factor for combined and linked primary and supplementary papers and median times between publications were summarized. (c) Assessment of PRO reporting using the new CONSORT extension The PRO CONSORT extension was applied to all trials to establish standards of PRO reporting, with the exception of item P6a which was an inclusion criterion (use of a validated PRO measure). Reporting of item 7a (PRO sample size) was recorded as present if a sample size calculation was completed in trials with patient reported primary outcomes, or if it were not applicable (for example, if there were no PROs as primary outcomes). Descriptive statistics are presented to consider PRO CONSORT standards within trials reporting in a single, combined publication or in linked primary and supplementary papers. Part 2 (a) Development of methods to present the ‘take home/trial exegesis’ message of PRO with clinical data in trials Methods to report the ‘take home’ messages of clinical and PROs in trials were developed through an in-depth analysis of items 2a and P2b (rationale/hypotheses for PRO measurement), and P20/21 and 22 (limitations and implications for clinical practice, and interpretation of PROs in relation to clinical outcomes) to identify good practice and produce methods to inform a PRO take home message from trials. All papers were read and re-read independently by at least two (AM, RM and JMB) researchers to become immersed in the data and relevant text was independently coded, copied verbatim into an electronic database and analysed for consistency between researchers. Discrepancies in coding were discussed within the study team (AM, RM and JMB). Methods for reporting trial ‘take home’ message were developed and applied iteratively to relevant text. Deviant examples were sought to challenge theories. Primary quotations are provided in accordance to methods of qualitative rigor. It was also noted whether primary papers indicated the future publication of a supplemental PRO paper (defined as “signposting”). (b) Application of the novel methods to included trials The methods described above were applied to the included trials and reported data examined by whether trials were published in combined or linked primary and supplemental reports. Results Part 1(a) Identification of well-designed and conducted RCTs reporting PROs with validated instruments OVID (MEDLINE and Embase) and Cochrane database search yields were 1815 and 939 records. After de-duplication, 1917 abstracts were screened, 1716 excluded, and 201 full text articles further assessed for eligibility, and these were supplemented with 13 studies from a previous systematic review [2]. Sixty-seven articles met the inclusion criteria describing trials at low risk of bias (Fig 1). The 66 included articles reported PROs from 49 RCTs, the majority of which were chemotherapy interventions (22/49, 44.9%) in colorectal disease (36/49, 73.5%, Table 1). 10.1371/journal.pone.0160998.g001Fig 1 PRISMA diagram of stages of the systematic review. 10.1371/journal.pone.0160998.t001Table 1 Characteristics of included trials grouped by whether PRO and clinical data were presented in combined PRO and clinical paper or separate primary and supplemental papers. Combined clinical and PRO* paper (n = 37) Separate clinical and PRO papers (n = 15) N (%) N (%) Trial type     • Chemotherapy 14 (37) (33)     • Radiotherapy 0 (14)     • Surgery 12 (32) (46)     • Other‡ 11 (30) 7 (46) Disease site     • Esophagogastric 22.4) 0     • Colorectal 27 (72) 12 (80) Sample size     • <100 8 (16) (7)     • 100–199 4 (10) (7)     • 200–499 13 (35) (33)     • 500–999 8 (21) (27)     • 1000–1999 3 (8) (27)     • 2000–3000 1 (3) 0 Primary outcome     • Survival 20 (54) (46)     • Response rate (8) 0     • Progression/ recurrence 6 (16) (13)     • PRO (16) 0     • Hospital stay (5) 0     • 30 day post-op morbidity 0 (7)     • Diarrhea (3) 0     • Pulmonary infection (3) 0     • Unclear 2 (5) 0 Journal impact factor Median (range) Median (range)     • Primary paper 16.8 (2.8 to 50) (5.1 to 50)     • Supplemental paper 5.2 (2.1 to 30) Time between primary and supplemental paper (months) n/a 20 (5 to 51) ‡other biochemical modulators e.g.monoclonal antibody, radioactive yttrium * PRO: Patient reported outcome (b) Categorisation of papers into combined PRO and clinical reports or linked primary and supplemental reports Some 36 (71%) single papers reported combined PRO and clinical results and the remaining 30 were linked primary and supplemental papers. The median journal impact factor for both combined and primary papers was the same (16.8), but supplemental PRO papers were reported in journals with a lower median impact factor (5.2). The mean time between publication of linked primary and supplementary papers was 20 months (range 5 to 51). (c) Assessment of PRO reporting using the new CONSORT extension Overall, the reporting of most papers did not meet the new PRO CONSORT standards (S1 Table). There were six (2 combined; 0 primary and 4 supplemental) papers reporting all PRO CONSORT items (Table 2). Primary papers reported the fewest items (median 3, range 2 to 7), typically lower than combined papers (median 6, range 2 to 12) and supplemental papers (median 10, range 4 to 12). The least frequently reported items related to “results” (13a, 15, 16, 17a and 18), and were reported in less than one third of papers. The least frequently reported items related to “results” (13a, 15, 16, 17a and 18), and were reported in less than one third of papers. 10.1371/journal.pone.0160998.t002Table 2 Analyses of reporting PRO CONSORT extension criteria, grouped by combined PRO and clinical papers, or separate primary and supplemental papers. CONSORT PRO item Combined clinical and PRO paper (n = 36) Separate clinical and PRO papers (n = 15 pairs) Total(n = 66) 1° 2° N (%) N (%) N (%) N (%) P1b The PRO should be identified in the abstract as a primary or secondary outcome. 28 (78) 4 (27) 15 (100) 47 (71) 2a/P2b* The relevant background and rationale for why PROs were assessed in the RCT should be briefly described/ The PROs hypothesis should be stated and relevant domains identified, if applicable. 22 (61) 5 (33) 15 (100) 42 (64) P6a** Evidence of PRO instrument validity and reliability should be provided or cited, if available. 36 (100) 15 (100) 15 (100) 66 (100) 7a† How PRO sample size was determined 33 (92) 15 (100) 15 (100) 63 (95) P12a Statistical approaches for dealing with missing data are explicitly stated 9 (25) 1 (7) 10 (66) 20 (30) 13a The number of PRO outcome data at baseline and at subsequent time points should be made transparent 6 (17) 1 (7) 9 (60) 16 (24) 15 A table showing baseline demographic and clinical characteristics for each group including PRO data 7 (19) 0 10 (67) 17 (26) 16 For each group, number of participants (denominator) included in each analysis and whether the analysis was by original assigned groups 8 (22) 0 8 (53) 16 (24) 17a For multidimensional PROs, results from each domain and time point specified for analysis. 9 (25) 0 10 (67) 19 (29) 18 Results of any other analyses performed, including subgroup analyses and adjusted analyses, distinguishing prespecified from exploratory including PRO analyses, where relevant 10 (28) 3 (20) 7 (46) 20 (30) P20/21‡ PRO–specific limitations and implications for generalizability and clinical practice 29 (81) 3 (20) 14 (93) 46 (70) 22‡ PRO data should be interpreted in relation to clinical outcomes including survival data, where relevant 29 (81) 3 (20) 14 (93) 46 (70) * For more detailed analysis of items 2a/P2b see Table 3 ** Study inclusion criteria † Only applicable to trials with PROs as primary outcome (n = 6, all combined papers). ‡ For more detailed analysis of items P20/21/22 see Table 3 Part 2 (a) Development of novel methods to examine the take home message of PRO with clinical data in trials The new method for understanding and improving combined reporting practice was developed based upon the in-depth analyses, emergent data and iterative discussions with AM and JMB. Sections of verbatim text were coded to items 2a and P2b were classified as providing a 1) detailed rationale/hypothesis—when authors included a specific PRO domain and/or hypothesized effect, 2) general rationale/hypothesis—when authors included non-specific rationale (i.e. “to examine quality of life”) or 3) no rationale provided. Verbatim quotes of PRO rationales are presented in Tables 3 and 4. An example of a detailed rationale included: “…we hypothesized a priori that [intervention] would result in a decrease in the magnitude and rate of decline in HRQL, particularly in physical function and overall well-being”[4], and an illustration of a general rationale included “…to compare…quality of life…between [intervention] and [control]”[5]. 10.1371/journal.pone.0160998.t003Table 3 Reported PRO rationale (items 2a and P2b) and authors’ interpretation of PRO in relation to clinical findings (items P20/21 and 22) in primary reports of trials with separate primary and supplemental papers. Extracted text was abridged where appropriate, as indicated by a series of periods (…), but otherwise presented verbatim. Author [citation] PRO rationale Level of detail* Interpretation of PRO in relation to clinical findings: Level of detail† Ajani [6] [7] 1° “To investigate whether adding [intervention] to [control] could improve patient outcomes (time-to-progression [TTP], overall survival [OS}, quality of life. . .” “Time to 5% definitive deterioration in global health status assed by QLQ-C30 was the primary quality of life parameter. “ Detailed “. . . [Intervention] resulted in significantly improved TTP (primary end point), OS, and overall response rate (secondary end points), with global health status (quality of life). . . preserved for a longer time.” Detailed 2° “. . .to investigate whether the better efficacy with [intervention] was counterbalanced by. . .the impact. . .on patient QOL” “The primary endpoint of the QOL assessment was. . .global health status” Detailed “.. significantly better preservation of QOL for patients treated with [intervention].. as a result of a significantly higher level of efficacy. . . . despite a higher incidence of some toxicities. . .” General Au[4] [8] 1° “. . .no trials have demonstrated an effect of [intervention] on. . . .quality of life. . .” “The secondary end points were. . .quality of life, assessed by mean changes in scores of physical function and global health status..” Detailed “[Intervention] improves overall survival and progression-free survival and preserves quality of life measures. . .” General 2° “…. we hypothesized a priori that [intervention] would result in a decrease in the magnitude and rate of decline in HRQL, particularly in physical function and overall well-being.” Detailed “Patients who received [intervention] experienced significantly less HRQL deterioration and a longer time before clinically significant deterioration occurred. These results are important, because…although [intervention]. . . results in improved OS, PFS, RR, and DCR…the magnitude of these benefits. . .was not large.” “. . .[intervention] offers clinically important survival and HRQL benefits. . .” General de Boer [9] [10] 1° No PRO rationale Absent No integration of PRO and clinical data Absent 2° “. . .to compare the quality of life of patients. . .who underwent [intervention] with patients.. who underwent [control]” General “..comparing. . .quality of life..is of great interest because a choice between the..two treatment options proves to be difficult..based on overall survival. However…no lasting differences in the quality of life of patients. . .were found.” General Braga[11] [12] 1° “To clarify the value of [intervention]. . .quality of life. . .should be considered” General “[intervention] resulted in earlier postoperative recovery, better cosmesis and improved quality of life. . .compared to [control].” General 2° “The primary endpoint was to compare the impact of [intervention] and [control] on 30-day postoperative morbidity.” “Recovery of social and physical activity was evaluated. . .by a specificd adaptation of the SF-36. . .” Detailed “. . .the [intervention] resulted in a reduction of both the overall morbidity rate and the length of hospital stay, and in a faster recovery of physical and social activity.” Detailed Chau [13] [14] 1° No PRO rationale Absent No integration of PRO and clinical data Absent 2° “. . .to assess QOL. . . .in patients receiving [intervention]” General “[Intervention] was associated with significantly better quality of life. . . Due to the shorter treatment duration, [intervention] had a faster time of QOL recovery. Quality adjusted survival was also in favour of the [intervention]… General Hallböök [15] [16] 1° No PRO rationale Absent No integration of PRO and clinical data Absent 2° “We hypothesized that such clear differences in clinical bowel function [with the intervention] would also be reflected in the score of a general quality of life instrument……” Detailed “The observed difference in clinical bowel function was not. . . reflected in an improved QOL score. . .” General Janson [17] [18] 1° No PRO rationale Absent No integration of PRO and clinical data Absent 2° “. . .with the hypothesis that [intervention] results in an improved HRQL when compared with [control]” Detailed “HRQL was better..after [intervention]. At present, several studies indicate that the oncologic results are at least equal after [intervention].” General Kabbinavar [19] [20] 1° No PRO rationale Absent No integration of PRO and clinical data Absent 2° “The primary HRQoL endpoint was the time to deterioration in HRQoL measured by the Colorectal Cancer Subscale score” Detailed “this prospective HRQoL analysis supports the clinical benefit of [intervention] in improving time to disease progression and prolonging overall survival, without compromising patients’ HRQoL”. General King [5] [21] 1° “The aim of this study was to compare. . . quality of life. . .in a prospective group of patients undergoing [intervention]” General “Patients undergoing [intervention] stay in hospital half as long. . .with no. . .deterioration in quality of life. . .” “. . .clinical improvements resulting from [intervention] did not cause significant deterioration in quality of life. . .” General 2° “. . .to compare recovery after [intervention] and [control]. . .using. . .self-report and observer data.” General “The earlier discharge in the [intervention] group did not result in any deterioration in quality of life outcomes compared with those in the [control] group” “Despite perioperative optimization of [control], short-term outcomes were better following [intervention]. There was no deterioration in quality of life or increased cost associated with the [intervention].” General Kopec [22] [23] 1° “A secondary aim was to compare quality of life. . .” General No integration of PRO and clinical data Absent 2° “We hypothesized that the [intervention] would be associated with higher HRQL and that it would be perceived as more convenient.” …. “The primary end point for this study was the FACT-C total score.” Detailed “The efficacy of the two regimens is similar, as demonstrated.. by the survival and disease-free survival analyses. . . This underscores the importance of patient-reported outcomes..” “Both regimens … do not differ in their impact on HRQL.” General Marijnen [24] [25] 1° No PRO rationale Absent No integration of PRO and clinical data Absent 2° “. . .we studies the effects of [intervention] on the HRQL and sexual functioning. . .” Detailed “The results of this study enable physicians and patients to weigh the beneficial effect of [intervention] on local recurrence against the price to be paid in terms of HRQL and sexual functioning.” Detailed Siena [26] [27] 1° No PRO rationale Absent No integration of PRO and clinical data Absent 2° “. . .exploratory analyses were conducted that assessed the association between [trial outcome variables] and HRQoL” General “. . . lack of disease progression was associated with … higher HRQoL for [intervention] patients only. . .Lack of disease progression was associated with better symptom control, HRQoL, and OS. General Stephens [28] [29] 1° No PRO rationale Absent No integration of PRO and clinical data Absent 2° “. . .the advantages of [intervention] to all patients needs to be balanced against any negative impact on patients’ quality of life”. “. . .the primary quality-of-life aims as “What is the longer-term (2-year) effect of the treatments on (1) sexual function and (2) bowel function?” Secondary outcome measures were “What is the effect of treatment on physical function and general health?” To address these questions, the sexual dysfunction and bowel function scales from the QLQ-CR38 and the physical function and general health scales from the MOS SF-36 were used.” Detailed “Therefore our results, together with those of the Dutch trial, provide convincing data on the impact of surgery and PRE on sexual and bowel function.” “The information presented in this article should allow clinicians to discuss with patients an estimate of the benefit of PRE in terms of reduction in LR risk balanced against the detrimental toxicity that is attributable to PRE.” Detailed Weeks [30] [31] 1° No PRO rationale Absent “The detailed quality of life component of this trial suggests that greater benefits in terms of the quality of life and recovery may be possible if fewer procedures are converted.” General 2° “The trial was also designed to test the hypothesis that [intervention] is associated with superior QOL outcomes” “the study protocol specified. . .the variability in pain distress item and the global ratings scale” Detailed “[Intervention].. results in statistically significant but clinically modest decreases in the duration of postoperative in-hospital analgesia and in length of stay. . . However, these differences do not translate into statistically significant improvements in symptoms or QOL. . .” General Wu[32] [33] 1° No PRO rationale Absent No integration of PRO and clinical data Absent 2° “We hypothesised that patients receiving [the intervention] would have more symptoms and greater fatigue than patients receiving [the control], with treatment arms difference most prominent at the 6-month assessment, and probably continuing up to 1 year after random assignment.” Detailed “Although the morbidity rate was higher in [intervention] patients than in [control] patients, our analysis indicates that [intervention] did not adversely influence QOL” General * Detailed rationale/hypothesis: specifying a PRO domain or hypothesized effect; general rationale/hypothesis: any other description; absent: no rationale. See methods for more details † Interpretation was considered “detailed” where authors discussed the direction of change (e.g. increased/decreased/no change) of a specific PRO domain (e.g. physical function) in relation to the direction of change of a specific clinical outcome (e.g. survival). All other discussions, where present, were considered “partial” interpretations. Where no appropriate text was identified, interpretation was considered “absent”. See methods for more details 10.1371/journal.pone.0160998.t004Table 4 Reported PRO rationale (items 2a and P2b) and authors’ interpretation of PRO in relation to clinical findings (items P20/21 and 22) in reports of trials with combined clinical and PRO papers. Extracted text was abridged where appropriate, as indicated by a series of periods (…), but otherwise presented verbatim. Author [citation] PRO rationale Level of detail* Interpretation of PRO in relation to clinical findings Level of detail† Biere [34] “We compared [intervention] with [control]. . . to assess the rate of pulmonary infection and quality of life associated with [intervention]” General “In this trial, [intervention] resulted in a lower incidence of pulmonary infections 2 weeks after surgery and during stay in hospital, a shorter hospital stay, and better short-term quality of life than did [control], with no compromise in the quality of the resected specimen.” “Additionally, [intervention] preserved quality of life better than [control] did. After 6 weeks, the SF 36 questionnaire and global health experience in the EORTC C30 module were better for patients in the [intervention] group than for those in the [control] group. In the oesophageal-specific OES 18 questionnaire, pain and talking were adversely affected in patients in the [control] group as compared with those in the [intervention] group.” Detailed Bramhall [35] No PRO rationale Absent No integration of PRO and clinical data Absent Carmichael [36] No PRO rationale Absent “The safety advantages of [intervention] surprisingly did not lead to demonstrable improvement in quality of life.” General Cunningham [37] No PRO rationale Absent No integration of PRO and clinical data Absent de Gramont [38] “. . .to compare the two treatments in terms of. . .QoL” General “The [intervention] seems beneficial. . ., demonstrating a prolonged progression free survival with acceptable tolerability and maintenance of QoL.” “Median QoL scores were similar for the two arms. . ., despite the increased incidence of [treatment]-related side effects …” General Doeksen [39] “The objective. . .was to compare functional and surgical results of [intervention] with [control] and their impact on quality of life.” “The primary end-point was the function. . . assessed at 12 months by the validated COlo-Rectal Functional Outcome (COREFO) questionnaire’s summary score.” Detailed “. . .a better functional outcome was found in patients with [intervention] than [control]. These functional differences did not influence health-related and overall quality of life.” General Douillard [40] “The QLQ-C30 questionnaire was analysed with the global health status/QoL scale (QL) as the primary endpoint. . .” Detailed “[Intervention] was well-tolerated and increased response rate, time to progression, and survival, with a later deterioration in quality of life.” † General Douillard [41] No PRO rationale Absent “It was surprising that there was no observed difference between the treatment arms in quality of life, despite the clear reduction in toxicity with [intervention].” General Fein [42] “. . .to identify optimal [treatment] in terms of quality of life” General “There were no differences in operative time, postoperative complications, and mortality.., there were no benefits of [intervention] in terms of quality of life, independent of the resection status. In the third, fourth, and fifth year after surgery quality of life was significantly improved for patients with [intervention]. General Fields [43] No PRO rationale Absent No integration of PRO and clinical data Absent Fuchs [44] “. . .to compare. . .effect on patient quality of life of these two [treatments]” General “This. . . trial provides comparative data on the efficacy, tolerability, and effect on patient quality of life between the [treatments].” General Furst [45] “. . .we tested [intervention] with [control] for. . . .quality of life. . .” General No integration of PRO and clinical data Absent Gray [46] “. . .to assess whether [intervention] could. . . .change quality of life” General “No decrease in quality of life was observed which is in accord with the lack of serious toxicity and treatment-related complications.” “[intervention] increases treatment effectiveness when measured by tumor response and time to disease progression and suggests an increase in survival for patients surviving more than 15 months. [Intervention] does not compromise quality of life or add significant toxicity.” General Guillou [47] No PRO rationale Absent “no differences were recorded between [control] and [intervention]. . . with respect to tumour and nodal status, short term endpoints, and quality of life.” General Hoksch [48] “. . .to evaluate the quality of life during the first postoperative year comparing [intervention] and [control]” General “In this study of global health status and quality of life, patients operated on with [control procedure] did not reach their preoperative values compared to the patients with the [intervention]. . .” “The clinical advantage manifested 6 months after operation. . . For that reason only patients with a good long-term prognosis might benefit from [intervention]. Detailed Jayne [49] No PRO rationale Absent “[Intervention]. . . . is as effective as [control] in terms of oncological outcomes and preservation of QoL” General Kang [50] No PRO rationale Absent “. . .[intervention] is feasible and does not increase short-term oncological risks, which are predicted by CRM positivity and macroscopic quality of TME specimens. . .The results of this trial also suggest that [intervention] results in a better quality of life for up to 3 months. . .” General Kataria [51] “To compare the quality of life (QOL) in patients undergoing [intervention] with [control]. . .” “The objective of this study is to assess the QOL following [intervention]. . .” General No integration of PRO and clinical data Absent Kemeny [52] “We hypothesized that patients in the [intervention] arm would have better physical and social functioning, fewer role limitations due to their emotional health, and better health perceptions than patients in the [control] arm.” Detailed “[Intervention] prolonged the median survival. . . was associated with a greater likelihood of objective tumor responses. . ., enhanced time to hepatic progression. . ., and improved physical functioning (QoL measurements).” Detailed Kohne [53] No PRO rationale Absent No integration of PRO and clinical data Absent Lal [54] “No studies have evaluated whether [intervention] is superior. . .in terms of. . .quality of life” Detailed “There were no improvements in failure-free survival from continuing [intervention]. . . . However,.. there was no deterioration in QoL. . .” General Maughan [55] “Several specific quality-of-life endpoints were predefined in the protocol: palliation of key symptoms, toxic effects, psychological effect, functional status, social functioning, and overall quality of life” Detailed “[A] and [B] regimens were similar in terms of survival, quality of life, and response rates. [C] showed similar response rates and overall survival to the [A] regimen and was easier to administer, but resulted in greater toxicity and inferior quality of life.”‡ “Since there was similar overall survival, quality of life became an important outcome measure.” General Punt [56] “The primary objective. . .was to examine the treatment effect on the mean global health status score. . .” Detailed No integration of PRO and clinical data Absent Punt [57] “The primary objective. . .was to examine the treatment effect on the mean global health status score. . .” Detailed “[Intervention]. . .results in a small but significant improvement in progression free survival without adding toxicity or worsening QoL. . . .” † General Rao [58] No PRO rationale Absent “Although there was a trend in favor of [intervention] for progression free survival, and more patients had stable disease, this did not translate in an improved QOL or survival advantage.” General Ross [59] “…we report results. . .comparing [intervention] with [control] using. . .QOL. . .as the study’s end points” General “The equivalent efficacy of [intervention] was demonstrated, but QOL was superior with [control].” General Sailer [60] “Randomised trials. . .have shown functional superiority of [intervention]. . .it was hypothesized that significant differences in bowel function should also be reflected in quality of life” “Sample size analysis was based on. . .global health status. . .” Detailed “. . .patients undergoing [intervention] may not only expect better functional results but also an improved quality of life. . .” General Saini [61] “. . .to assess QOL of patients undergoing [treatment]” General “this study has demonstrated that the [intervention] is associated with less acute toxicity and less impairment of QOL than [control]. Furthermore, this has been achieved without any obvious adverse effect on outcome” General Saltz [62] No PRO rationale Absent “the [intervention] was associated with higher rates of tumor regression, progression-free survival, and overall survival without compromising the quality of life.” General Sobrero [63] No PRO rationale Absent “.. Progression free survival was significantly longer in experimental. . ., while the overall survival was similar in both arms. . .; quality of life was similar as well.” † General Sobrero [64] No PRO rationale Absent “.. [intervention] reduced the risk of progression.., and improved median progression free survival. . ., and response rate.. The QOL assessments also support this benefit. Global health status as well as physical, emotional, and cognitive functioning were significantly better with [intervention].” Detailed Tebbutt [65] No PRO rationale Absent “the addition of [intervention]. . .has no effect on response rates compared with [control]. In addition, there was no significant effect on overall survival or quality of life,. . .” General Tol [66] No PRO rationale Absent “the [intervention] resulted in a significant decrease in progression free survival and a poorer quality of life” General Van Hooft [67] “We aimed to establish whether [intervention] has better health outcomes than does [control]” “The primary outcome was mean global health status. . . assessed with the QL2 subscale of the European Organisation for Research and Treatment of Cancer quality of life questionnaire.” “This measure was chosen because the outcome of the treatments, such as need for a stoma, incisional hernia, lengthy intensive care, and hospital stay, might affect patients’ quality of life” Detailed “. . .[intervention] or [control] did not have any distinct benefits for global health status, mortality, morbidity, other quality of life dimensions, and stoma rates.” Detailed Vlug [68] “. . .combining the [intervention] will result in the fastest postoperative recovery.” Detailed “Treatment groups had similar morbidity, reoperation and readmission rates, equal in-hospital mortality, comparable levels of quality of life. . .” General Zachariah [69] . . . .”if [intervention] was efficacious in reducing treatment-induced diarrhea, better QoL and bowel scores were expected for the [intervention] for all instruments.” Detailed “We found that [intervention] did not show a statistically significant reduction in the incidence or severity of diarrhea or change in patient-reported bowel function. . .” Detailed * Detailed rationale/hypothesis: specifying a PRO domain or hypothesized effect; general rationale/hypothesis: any other description; absent: no rationale. See methods for more details † Interpretation was considered “detailed” where authors discussed the direction of change (e.g. increased/decreased/no change) of a specific PRO domain (e.g. physical function) in relation to the direction of change of a specific clinical outcome (e.g. survival). All other discussions, where present, were considered “partial” interpretations. Where no appropriate text was identified, interpretation was considered “absent”. See methods for more details Likewise, sections of text coded to the items P20/21 and 22 were developed as 1) detailed interpretations—where authors discussed the hypothesized effect of a specific PRO in relation to the hypothesized effect of a clinical outcome, 2) general interpretations–when authors include non-specific interpretation of PROs in relation to clinical outcomes, or 3) no integrated PRO and clinical interpretation of results in the paper. An example of detailed interpretation of findings includes “… [Intervention] resulted in significantly improved TTP [time to progression] (primary end point), OS [overall survival], and overall response rate (secondary end points), with global health status (quality of life)… preserved for a longer time.”[6] Part 2, b) Application of the methods to included trials Of the 36 trials reporting combined papers, there were 11 (30%) papers that provided a PRO rationale/hypothesis, 10 (30%) providing general information and 15 (40%) not providing a PRO rationale (Table 5). The interpretation of PRO data in the context of clinical outcomes were detailed in six (16%), general in 24 (65%) and absent in 7 (17%) papers (Items P20/21 and 22, Table 3). There were seven papers that described both detailed PRO rationale/hypotheses and detailed interpretations of PROs in relation to clinical outcomes [5, 6, 13, 26, 30, 68, 70]. 10.1371/journal.pone.0160998.t005Table 5 Novel methods for assessing CONSORT PRO extension items 2a/P2b and P20/21/22, grouped by combined PRO and clinical papers, or linked primary and supplemental papers (n = 67). CONSORT PRO item Combined clinical and PRO paper (n = 36) Linked clinical and PRO papers (n = 15 pairs) 1° 2° N (%) N (%) N (%) Rationale/hypothesis (2a/P2b)*     • Detailed 11 (31) 2 (14) 73)     • General 10 (28) 3 (20) (27)     • Absent 15 (41) 10 (66) 0 Interpretation of findings (P20/21/22)†     • Detailed 6 (17) 1 (7) (20)     • General 2 (64) 4 (27) 80)     • Absent 7 (19) 10 (66) 0 * Detailed rationale/hypothesis: specifying a PRO domain or hypothesized effect; general rationale/hypothesis: any other description; absent: no rationale. See methods for more details † Interpretation was considered “detailed” where authors discussed the direction of change (e.g. increased/decreased/no change) of a specific PRO domain (e.g. physical function) in relation to the direction of change of a specific clinical outcome (e.g. survival). All other discussions, where present, were considered “partial” interpretations. Where no appropriate text was identified, interpretation was considered “absent”. See methods for more details Where PRO and clinical results were published separately (n = 15), most primary papers did not provide any PRO rationale (n = 10, 66%) or text interpreting PROs in relation to clinical findings (n = 10, 66%). One primary paper provided detailed descriptions of both these issues. In comparison, all supplemental papers described PRO rationales and most (14, 93%) contained detailed interpretation of PROs in relation to clinical findings, although only two had detailed descriptions of both of these. Of the 15 primary papers, 66% signposted the presence of the supplemental PRO report. Conclusions Reporting PRO rationale linked to clinical hypotheses, and clear reporting of PRO results interpreted appropriately in the context of the clinical outcomes are critical to ensure that oncologists gather a “take-home” message to communicate to patients which encompasses clinical and PROs. This review explored this issue in detail. Patient reported outcome reporting standards were at lowest levels in primary clinical papers (where clinical trial data was reported separately to the supplementary PROs). Whilst supplemental papers provided more detail there was a 5 to 51 month delay in publication in less well cited journals, thus diminishing their impact. Most (71%) trials did report combined results, demonstrating that it is possible to do so. New methods to examine reporting of trial “take home” messages recommend that detailed information about domain specific PRO rational and interpretation with clinical data are supplied within a main trial paper to arm clinicians with relevant outcomes to use in decision-making. Authors need to be allowed space to report these details alongside clinical outcomes to inform the take home message from papers to help clinicians in practice. Where this is not appropriate for scientific reasons, for example, if primary outcome data are available before secondary PROs, then this could be explicitly stated. Other systematic reviews have shown that PRO reporting standards are poor [71–77] which contributed to the need for the development of the PRO CONSORT extension. Similarly, other papers have recommended clear PRO hypotheses and integration with clinical findings, however, there is no empirical data presented on how to best achieve this [78]. Reviews also have examined how PROs in RCTs influence decision-making and confirm that PRO information is not used in practice [77, 79, 80]. Previous work, however, has not provided solutions to these problems or considered the conceptual reasons why PRO data are not used in practice. Theoretically, failure of PRO data to have an impact on clinical practice may stem from problems that start with the trial design and conduct, compounded by poor reporting and separate PRO and clinical publications. Further research is now needed to investigate whether improved reporting will have the desired effect of informing patient-centred care, clinical decision making and health policy decisions. For example, work could include a study directly exploring oncologists’ views of trial reports following the introduction of new reporting guidelines would be informative. Additionally in-depth research of clinical decision-making in multi-disciplinary teams or of oncology consultations may be undertaken to examine how PRO data are used. Research needs to be targeted into each of these areas in order to understand how improvements, such as the recently published SPIRIT statement [81] for improving RCT protocols or the CONSORT PRO extension, impact clinical practice. What is clear is that cancer patients want information about PROs and indeed rate such data of similar importance to survival information [82–84]. It is therefore critical that oncologists communicate PRO data in the context of a shared doctor-patient consultation and methods to do so are being established [85–87]. This may have occurred because authors split trial results and deliberately left the PRO methodology and findings to the supplemental report. This review included a systematic search for studies using PRISMA guidelines [88] and transparent methodology for an in-depth analyses if the textual data, but there were some limitations. Trials of radical treatments of gastrointestinal cancers were selected for analyses because the authors were familiar with the PRO and clinical data in this field. It is conceivable that including trials in other diseases or in the palliative setting may have identified different reporting standards. For example, in the palliative setting, authors may make greater reference to PROs and their integration with clinical outcomes because the main focus of treatment is not to cure disease. Further work is needed to examine this area in detail. In addition, studies with high or unascertainable risk of bias were excluded because lower standards of reporting are associated with bias and likely poor PRO reporting [89], and it is possible that important data were missed. This is considered to be unlikely however, because even the included “high-quality” trials demonstrated significant reporting weaknesses and inclusion of poor quality trials would probably not yield exemplar practice. In summary, this review presents and evidence based way of implementing the new CONSORT PRO extension items 2a, P2b, P20/21 and 22 based on current literature. It is recommended that RCTs report domain-specific PRO rationale with anticipated treatment effects, and integrate these findings with specific clinical outcomes in a single combined report. It is acknowledged that trials are structured around their primary (often clinical) endpoints, and it is appropriate to prioritise these data at the expense of other outcomes. It seems unnecessary, however, to relegate evaluation of patient experience to reports that may be less likely to influence practice. The adoption of better standards for PRO reporting could facilitate the use of PRO data by oncologists and patients for informed decision making. Supporting Information S1 Appendix OVID search strategy. (DOCX) Click here for additional data file. S2 Appendix PRISMA Checklist. (PDF) Click here for additional data file. S1 Table Reporting of CONSORT PRO extension by individual publication, including total number of appropriately reported items. (DOCX) Click here for additional data file. ==== Refs References 1 Calvert M , Brundage M , Jacobsen PB , Schunemann HJ , Efficace F . The CONSORT Patient-Reported Outcome (PRO) extension: implications for clinical trials and practice . Health and quality of life outcomes . 2013 ;11 :184 10.1186/1477-7525-11-184 24168680 2 Blazeby JM , Avery K , Sprangers M , Pikhart H , Fayers P , Donovan J . Health-related quality of life measurement in randomized clinical trials in surgical oncology . Journal of Clinical Oncology . 2006 ;24 (19 ):3178 –86 . 16809741 3 The Cochrane C. Cochrane handbook for systematic reviews of interventions2009. 4 Au HJ , Karapetis CS , O'Callaghan CJ , Tu D , Moore MJ , Zalcberg JR , et al Health-related quality of life in patients with advanced colorectal cancer treated with cetuximab: overall and KRAS-specific results of the NCIC CTG and AGITG CO.17 Trial . Journal of Clinical Oncology . 2009 ;27 (11 ):1822 –8 . 10.1200/JCO.2008.19.6048 19273701 5 King PM , Blazeby JM , Ewings P , Franks PJ , Longman RJ , Kendrick AH , et al Randomized clinical trial comparing laparoscopic and open surgery for colorectal cancer within an enhanced recovery programme . British Journal of Surgery . 2006 ;93 (3 ):300 –8 . 16363014 6 Ajani JA , Moiseyenko VM , Tjulandin S , Majlis A , Constenla M , Boni C , et al Quality of life with docetaxel plus cisplatin and fluorouracil compared with cisplatin and fluorouracil from a phase III trial for advanced gastric or gastroesophageal adenocarcinoma: the V-325 Study Group . Journal of Clinical Oncology . 2007 ;25 (22 ):3210 –6 . 17664468 7 Van Cutsem E , Moiseyenko VM , Tjulandin S , Majlis A , Constenla M , Boni C , et al Phase III study of docetaxel and cisplatin plus fluorouracil compared with cisplatin and fluorouracil as first-line therapy for advanced gastric cancer: a report of the V325 Study Group . Journal of Clinical Oncology . 2006 ;24 (31 ):4991 –7 . 17075117 8 Jonker DJ , O'Callaghan CJ , Karapetis CS , Zalcberg JR , Tu D , Au HJ , et al Cetuximab for the treatment of colorectal cancer . New England Journal of Medicine . 2007 ;357 (20 ):2040 –8 . 18003960 9 de Boer AGEM , van Lanschot JJB , van Sandick JW , Hulscher JBF , Stalmeier PFM , De Haes JCJM , et al Quality of life after transhiatal compared with extended transthoracic resection for adenocarcinoma of the esophagus . Journal of Clinical Oncology . 2004 ;22 (20 ):4202 –8 . 15483031 10 Hulscher JBF , van Sandick JW , De Boer AGEM , Wijnhoven BPL , Tijssen JGP , Fockens P , et al Extended transthoracic resection compared with limited transhiatal resection for adenocarcinoma of the esophagus . New England Journal of Medicine . 2002 ;347 (21 ):1662 –9 . 12444180 11 Braga M , Frasson M , Zuliani W , Vignali A , Pecorelli N , Di Carlo V . Randomized clinical trial of laparoscopic versus open left colonic resection . British Journal of Surgery . 2010 ;97 (8 ):1180 –6 . 10.1002/bjs.7094 20602506 12 Braga M , Vignali A , Gianotti L , Zuliani W , Radaelli G , Gruarin P , et al Laparoscopic versus open colorectal surgery—A randomized trial on short-term outcome . Annals of Surgery . 2002 ;236 (6 ):759 –66 . .12454514 13 Chau I , Norman AR , Cunningham D , Iveson T , Hill M , Hickish T , et al Longitudinal quality of life and quality adjusted survival in a randomised controlled trial comparing six months of bolus fluorouracil/leucovorin vs. twelve weeks of protracted venous infusion fluorouracil as adjuvant chemotherapy for colorectal cancer . European Journal of Cancer . 2005 ;41 (11 ):1551 –9 . 16026692 14 Chau I , Norman AR , Cunningham D , Tait D , Ross PJ , Iveson T , et al A randomised comparison between 6 months of bolus fluorouracil/leucovorin and 12 weeks of protracted venous infusion fluorouracil as adjuvant treatment in colorectal cancer . Annals of Oncology . 2005 ;16 (4 ):549 –57 . 15695501 15 Hallbook O , Hass U , Wanstrom A , Sjodahl R . Quality of life measurement after rectal excision for cancer. Comparison between straight and colonic J-pouch anastomosis . ScandJ Gastroenterol . 1997 ;32 (5 ):490 –3 .9175213 16 Hallbook O , Pahlman L , Krog M , Wexner SD , Sjodahl R . Randomized comparison of straight and colonic J pouch anastomosis after low anterior resection . Annals of Surgery . 1996 ;224 (1 ):58 –65 . 8678619 17 Janson M , Lindholm E , Anderberg B , Haglind E . Randomized trial of health-related quality of life after open and laparoscopic surgery for colon cancer . Surgical Endoscopy . 2007 ;21 (5 ):747 –53 . 17342556 18 Veldkamp R , Kuhry E , Hop WC , Jeekel J , Kazemier G , Bonjer HJ , et al Laparoscopic surgery versus open surgery for colon cancer: short-term outcomes of a randomised trial . The lancet oncology . 2005 ;6 :477 –84 . 15992696 19 Kabbinavar FF , Wallace JF , Holmgren E , Yi J , Cella D , Yost KJ , et al Health-related quality of life impact of bevacizumab when combined with irinotecan, 5-fluorouracil, and leucovorin or 5-fluorouracil and leucovorin for metastatic colorectal cancer . Oncologist . 2008 ;13 (9 ):1021 –9 . 10.1634/theoncologist.2008-0003 18776057 20 Hurwitz H , Fehrenbacher L , Novotny W , Cartwright T , Hainsworth J , Heim W , et al Bevacizumab plus irinotecan, fluorouracil, and leucovorin for metastatic colorectal cancer .[see comment]. New England Journal of Medicine . 2004 ;350 (23 ):2335 –42 . 15175435 21 King PM , Blazeby JM , Ewings P , Kennedy RH . Detailed evaluation of functional recovery following laparoscopic or open surgery for colorectal cancer within an enhanced recovery programme . International Journal of Colorectal Disease . 2008 ;23 (8 ):795 –800 . 10.1007/s00384-008-0478-0 18465136 22 Kopec JA , Yothers G , Ganz PA , Land SR , Cecchini RS , Wieand HS , et al Quality of life in operable colon cancer patients receiving oral compared with intravenous chemotherapy: results from National Surgical Adjuvant Breast and Bowel Project Trial C-06 . Journal of Clinical Oncology . 2007 ;25 (4 ):424 –30 . 17264338 23 Lembersky BC , Wieand HS , Petrelli NJ , O'Connell MJ , Colangelo LH , Smith RE , et al Oral uracil and tegafur plus leucovorin compared with intravenous fluorouracil and leucovorin in stage II and III carcinoma of the colon: results from National Surgical Adjuvant Breast and Bowel Project Protocol C-06 . Journal of Clinical Oncology . 2006 ;24 (13 ):2059 –64 . 16648506 24 Marijnen CA , van de Velde CJ , Putter H , van den Brink M , Maas CP , Martijn H , et al Impact of short-term preoperative radiotherapy on health-related quality of life and sexual functioning in primary rectal cancer: report of a multicenter randomized trial . Journal of Clinical Oncology . 2005 ;23 (9 ):1847 –58 . 15774778 25 Kapiteijn E , Marijnen CAM , Nagtegaal ID , Putter H , Steup WH , Wiggers T , et al Preoperative radiotherapy combined with total mesorectal excision for resectable rectal cancer . New England Journal of Medicine . 2001 ;345 (9 ):638 –46 . 11547717 26 Siena S , Peeters M , Van CE , Humblet Y , Conte P , Bajetta E , et al Association of progression-free survival with patient-reported outcomes and survival: results from a randomised phase 3 trial of panitumumab . British Journal of Cancer . 2007 ;97 (11 ):1469 –74 . 18040272 27 Van Cutsem E , Peeters M , Siena S , Humblet Y , Hendlisz A , Neyns B , et al Open-label phase III trial of panitumumab plus best supportive care compared with best supportive care alone in patients with chemotherapy-refractory metastatic colorectal cancer . Journal of Clinical Oncology . 2007 ;25 (13 ):1658 –64 . 17470858 28 Stephens RJ , Thompson LC , Quirke P , Steele R , Grieve R , Couture J , et al Impact of short-course preoperative radiotherapy for rectal cancer on patients' quality of life: data from the Medical Research Council CR07/National Cancer Institute of Canada Clinical Trials Group C016 randomized clinical trial . Journal of Clinical Oncology . 2010 ;28 (27 ):4233 –9 . 10.1200/JCO.2009.26.5264 20585099 29 Sebag-Montefiore D , Stephens RJ , Steele R , Monson J , Grieve R , Khanna S , et al Preoperative radiotherapy versus selective postoperative chemoradiotherapy in patients with rectal cancer (MRC CR07 and NCIC-CTG C016): a multicentre, randomised trial .[see comment]. Lancet . 2009 ;373 (9666 ):811 –20 . 10.1016/S0140-6736(09)60484-0 19269519 30 Weeks JC , Nelson H , Gelber S , Sargent D , Schroeder G . Short-term quality-of-life outcomes following laparoscopic-assisted colectomy vs open colectomy for colon cancer: a randomized trial . JAMA . 2002 ;287 (3 ):321 –8 . 11790211 31 Nelson H , Sargent D , Wieand HS , Fleshman J , Anvari M , Stryker SJ , et al A comparison of laparoscopically assisted and open colectomy for colon cancer . New England Journal of Medicine . 2004 ;350 (20 ):2050 –9 . 15141043 32 Wu CW , Chiou JM , Ko FS , Lo SS , Chen JH , Lui WY , et al Quality of life after curative gastrectomy for gastric cancer in a randomised controlled trial . British Journal of Cancer . 2008 ;98 (1 ):54 –9 . 10.1038/sj.bjc.6604097 18182977 33 Wu CW , Hsiung CA , Lo SS , Hsieh MC , Chen JH , Li AF , et al Nodal dissection for patients with gastric cancer: a randomised controlled trial . The lancet oncology . 2006 ;7 :309 –15 . 16574546 34 Biere SS , van Berge Henegouwen MI , Maas KW , Bonavina L , Rosman C , Garcia JR , et al Minimally invasive versus open oesophagectomy for patients with oesophageal cancer: a multicentre, open-label, randomised controlled trial . Lancet . 2012 ;379 (9829 ):1887 –92 . 10.1016/S0140-6736(12)60516-9 22552194 35 Bramhall SR , Hallissey MT , Whiting J , Scholefield J , Tierney G , Stuart RC , et al Marimastat as maintenance therapy for patients with advanced gastric cancer: a randomised trial . British Journal of Cancer . 2002 ;86 (12 ):1864 –70 . 12085177 36 Carmichael J , Popiela T , Radstone D , Falk S , Borner M , Oza A , et al Randomized comparative study of tegafur/uracil and oral leucovorin versus parenteral fluorouracil and leucovorin in patients with previously untreated metastatic colorectal cancer . Journal of Clinical Oncology . 2002 ;20 (17 ):3617 –27 . 12202662 37 Cunningham D , Starling N , Rao S , Iveson T , Nicolson M , Coxon F , et al Capecitabine and oxaliplatin for advanced esophagogastric cancer . New England Journal of Medicine . 2008 ;358 (1 ):36 –46 . 10.1056/NEJMoa073149 18172173 38 de Gramont A , Figer A , Seymour M , Homerin M , Hmissi A , Cassidy J , et al Leucovorin and fluorouracil with or without oxaliplatin as first-line treatment in advanced colorectal cancer . Journal of Clinical Oncology . 2000 ;18 (16 ):2938 –47 . 10944126 39 Doeksen A , Bakx R , Vincent A , van Tets WF , Sprangers MA , Gerhards MF , et al J-pouch vs side-to-end coloanal anastomosis after preoperative radiotherapy and total mesorectal excision for rectal cancer: a multicentre randomized trial . Colorectal Disease . 2012 ;14 (6 ):705 –13 . doi: 10.1111/j.1463-1318.2011.02725.x . 10.1111/j.1463-1318.2011.02725.x 21831100 40 Douillard JY , Cunningham D , Roth AD , Navarro M , James RD , Karasek P , et al Irinotecan combined with fluorouracil compared with fluorouracil alone as first-line treatment for metastatic colorectal cancer: a multicentre randomised trial . Lancet . 2000 ;355 (9209 ):1041 –7 . 10744089 41 Douillard JY , Hoff PM , Skillings JR , Eisenberg P , Davidson N , Harper P , et al Multicenter phase III study of uracil/tegafur and oral leucovorin versus fluorouracil and leucovorin in patients with previously untreated metastatic colorectal cancer . Journal of Clinical Oncology . 2002 ;20 (17 ):3605 –16 . 12202661 42 Fein M , Fuchs KH , Thalheimer A , Freys SM , Heimbucher J , Thiede A . Long-term benefits of Roux-en-Y pouch reconstruction after total gastrectomy: a randomized trial . Annals of Surgery . 2008 ;247 (5 ):759 –65 . 10.1097/SLA.0b013e318167748c 18438112 43 Fields AL , Keller A , Schwartzberg L , Bernard S , Kardinal C , Cohen A , et al Adjuvant therapy with the monoclonal antibody Edrecolomab plus fluorouracil-based therapy does not improve overall survival of patients with stage III colon cancer . Journal of Clinical Oncology . 2009 ;27 (12 ):1941 –7 . 10.1200/JCO.2008.18.5710 19273708 44 Fuchs CS , Moore MR , Harker G , Villa L , Rinaldi D , Hecht JR . Phase III comparison of two irinotecan dosing regimens in second-line therapy of metastatic colorectal cancer . Journal of Clinical Oncology . 2003 ;21 (5 ):807 –14 . 12610178 45 Furst A , Burghofer K , Hutzel L , Jauch KW . Neorectal reservoir is not the functional principle of the colonic J- pouch: the volume of a short colonic J-pouch does not differ from a straight coloanal anastomosis . DisColon Rectum . 2002 ;45 (5 ):660 –7 . 46 Gray B , Van HG , Hope M , Burton M , Moroz P , Anderson J , et al Randomised trial of SIR-Spheres plus chemotherapy vs. chemotherapy alone for treating patients with liver metastases from primary large bowel cancer . Annals of Oncology . 2001 ;12 (12 ):1711 –20 . 11843249 47 Guillou PJ , Quirke P , Thorpe H , Walker J , Jayne DG , Smith AM , et al Short-term endpoints of conventional versus laparoscopic-assisted surgery in patients with colorectal cancer (MRC CLASICC trial): multicentre, randomised controlled trial . Lancet . 2005 ;365 (9472 ):1718 –26 . 15894098 48 Hoksch B , Ablassmaier B , Zieren J , Muller JM . Quality of life after gastrectomy: Longmire's reconstruction alone compared with additional pouch reconstruction . World J Surg . 2002 ;26 (3 ):335 –41 . 11865371 49 Jayne DG , Guillou PJ , Thorpe H , Quirke P , Copeland J , Smith AM , et al Randomized trial of laparoscopic-assisted resection of colorectal carcinoma: 3-year results of the UK MRC CLASICC Trial Group . Journal of Clinical Oncology . 2007 ;25 (21 ):3061 –8 . 17634484 50 Kang SB , Park JW , Jeong SY , Nam BH , Choi HS , Kim DW , et al Open versus laparoscopic surgery for mid or low rectal cancer after neoadjuvant chemoradiotherapy (COREAN trial): short-term outcomes of an open-label randomised controlled trial . Lancet Oncology . 2010 ;11 (7 ):637 –45 . 10.1016/S1470-2045(10)70131-5 20610322 51 Kataria K , Verma GR , Malhotra A , Yadav R . Comparison of quality of life in patients undergoing transhiatal esophagectomy with or without chemotherapy . Saudi Journal of Gastroenterology . 2012 ;18 (3 ):195 –200 . doi: 10.4103/1319-3767.96454 . 10.4103/1319-3767.96454 22626799 52 Kemeny NE , Niedzwiecki D , Hollis DR , Lenz HJ , Warren RS , Naughton MJ , et al Hepatic arterial infusion versus systemic therapy for hepatic metastases from colorectal cancer: a randomized trial of efficacy, quality of life, and molecular markers (CALGB 9481) . Journal of Clinical Oncology . 2006 ;24 (9 ):1395 –403 . 16505413 53 Kohne CH , Wils J , Lorenz M , Schoffski P , Voigtmann R , Bokemeyer C , et al Randomized phase III study of high-dose fluorouracil given as a weekly 24-hour infusion with or without leucovorin versus bolus fluorouracil plus leucovorin in advanced colorectal cancer: European organization of Research and Treatment of Cancer Gastrointestinal Group Study 40952 . Journal of Clinical Oncology . 2003 ;21 (20 ):3721 –8 . 12963704 54 Lal R , Dickson J , Cunningham D , Chau I , Norman AR , Ross PJ , et al A randomized trial comparing defined-duration with continuous irinotecan until disease progression in fluoropyrimidine and thymidylate synthase inhibitor-resistant advanced colorectal cancer . Journal of Clinical Oncology . 2004 ;22 (15 ):3023 –31 . 15284252 55 Maughan TS , James RD , Kerr DJ , Ledermann JA , McArdle C , Seymour MT , et al Comparison of survival, palliation, and quality of life with three chemotherapy regimens in metastatic colorectal cancer: a multicentre randomised trial . Lancet . 2002 ;359 (9317 ):1555 –63 . 12047964 56 Punt CJ , Nagy A , Douillard JY , Figer A , Skovsgaard T , Monson J , et al Edrecolomab alone or in combination with fluorouracil and folinic acid in the adjuvant treatment of stage III colon cancer: a randomised study . Lancet . 2002 ;360 :671 –7 . 12241873 57 Punt CJ , Keizer HJ , Douma J , Skovsgaard T , Schuller J , Muller EW , et al Trimetrexate as biochemical modulator of 5-fluorouracil/leucovorin in advanced colorectal cancer: final results of a randomised European study . Annals of Oncology . 2002 ;13 (1 ):81 –6 . 11865814 58 Rao S , Cunningham D , de Gramont A , Scheithauer W , Smakal M , Humblet Y , et al Phase III double-blind placebo-controlled study of farnesyl transferase inhibitor R115777 in patients with refractory advanced colorectal cancer . Journal of Clinical Oncology . 2004 ;22 (19 ):3950 –7 . 15459217 59 Ross P , Nicolson M , Cunningham D , Valle J , Seymour M , Harper P , et al Prospective randomized trial comparing mitomycin, cisplatin, and protracted venous-infusion fluorouracil (PVI 5-FU) with epirubicin, cisplatin, and PVI 5-FU in advanced esophagogastric cancer . Journal of Clinical Oncology . 2002 ;20 (8 ):1996 –2004 . 11956258 60 Sailer M , Fuchs KH , Fein M , Thiede A . Randomized clinical trial comparing quality of life after straight and pouch coloanal reconstruction . BrJ Surg . 2002 ;89 (9 ):1108 –17 .12190675 61 Saini A , Norman AR , Cunningham D , Chau I , Hill M , Tait D , et al Twelve weeks of protracted venous infusion of fluorouracil (5-FU) is as effective as 6 months of bolus 5-FU and folinic acid as adjuvant treatment in colorectal cancer . British Journal of Cancer . 2003 ;88 (12 ):1859 –65 . 12799627 62 Saltz LB , Cox JV , Blanke C , Rosen LS , Fehrenbacher L , Moore MJ , et al Irinotecan plus fluorouracil and leucovorin for metastatic colorectal cancer . New England Journal of Medicine . 2000 ;343 :905 –14 . 11006366 63 Sobrero A , Zaniboni A , Frassineti GL , Aschele C , Guglielmi A , Giuliani R , et al Schedule specific biochemical modulation of 5-fluorouracil in advanced colorectal cancer: a randomized study. GISCAD, IOR and collaborating centers . Annals of Oncology . 2000 ;11 (11 ):1413 –20 . 11142481 64 Sobrero AF , Maurel J , Fehrenbacher L , Scheithauer W , Abubakr YA , Lutz MP , et al EPIC: phase III trial of cetuximab plus irinotecan after fluoropyrimidine and oxaliplatin failure in patients with metastatic colorectal cancer . Journal of Clinical Oncology . 2008 ;26 (14 ):2311 –9 . 10.1200/JCO.2007.13.1193 18390971 65 Tebbutt NC , Norman A , Cunningham D , Iveson T , Seymour M , Hickish T , et al A multicentre, randomised phase III trial comparing protracted venous infusion (PVI) 5-fluorouracil (5-FU) with PVI 5-FU plus mitomycin C in patients with inoperable oesophago-gastric cancer . Annals of Oncology . 2002 ;13 (10 ):1568 –75 . 12377644 66 Tol J , Koopman M , Cats A , Rodenburg CJ , Creemers GJ , Schrama JG , et al Chemotherapy, bevacizumab, and cetuximab in metastatic colorectal cancer . New England Journal of Medicine . 2009 ;360 (6 ):563 –72 . 10.1056/NEJMoa0808268 19196673 67 van Hooft JE , Bemelman WA , Oldenburg B , Marinelli AW , Holzik MF , Grubben MJ , et al Colonic stenting versus emergency surgery for acute left-sided malignant colonic obstruction: a multicentre randomised trial .[Erratum appears in Lancet Oncol. 2011 May;12(5):418]. Lancet Oncology . 2011 ;12 (4 ):344 –52 . 10.1016/S1470-2045(11)70035-3 21398178 68 Vlug MS , Wind J , Hollmann MW , Ubbink DT , Cense HA , Engel AF , et al Laparoscopy in combination with fast track multimodal management is the best perioperative strategy in patients undergoing colonic surgery: a randomized clinical trial (LAFA-study) . Annals of Surgery . 2011 ;254 (6 ):868 –75 . 10.1097/SLA.0b013e31821fd1ce 21597360 69 Zachariah B , Gwede CK , James J , Ajani J , Chin LJ , Donath D , et al Octreotide acetate in prevention of chemoradiation-induced diarrhea in anorectal cancer: randomized RTOG trial 0315 . Journal of the National Cancer Institute . 2010 ;102 (8 ):547 –56 . 10.1093/jnci/djq063 20339140 70 Kemeny NE , Niedzwiecki D , Hollis DR , Lenz HJ , Warren RS , Naughton MJ , et al Hepatic arterial infusion versus systemic therapy for hepatic metastases from colorectal cancer: a randomized trial of efficacy, quality of life, and molecular markers (CALGB 9481) .[see comment]. Journal of Clinical Oncology . 2006 ;24 (9 ):1395 –403 . 16505413 71 Efficace F , Bottomley A , Osoba D , Gotay C , Flechtner H , D'Haese S , et al Beyond the development of health-related quality-of-life (HRQOL) measures: A checklist for evaluating HRQOL outcomes in cancer clinical trials—Does HRQOL evaluation in prostate cancer research inform clinical decision making? Journal of Clinical Oncology . 2003 ;21 (18 ):3502 –11 . 12972527 72 Lee CW , Chi KN . The standard of reporting of health-related quality of life in clinical cancer trials . Journal of Clinical Epidemiology . 2000 ;53 (5 ):451 –8 . 10812316 73 Efficace F , Bottomley A , van Andel G . Health related quality of life in prostate carcinoma patients—A systematic review of randomized controlled trials . Cancer . 2003 ;97 (2 ):377 –88 . 12518362 74 Efficace F , Horneber M , Lejeune S , Van Dam F , Leering S , Rottmann M , et al Methodological quality of patient-reported outcome research was low in complementary and alternative medicine in oncology . Journal of Clinical Epidemiology . 2006 ;59 (12 ):1257 –65 . 17098568 75 Kong SX , Gandhi SK . Methodologic assessments of quality of life measures in clinical trials . Annals of Pharmacotherapy . 1997 ;31 (7–8 ):830 –6 . 9220039 76 Cocks K , King MT , Velikova G , Fayers PM , Brown JM . Quality, interpretation and presentation of European Organisation for Research and Treatment of Cancer quality of life questionnaire core 30 data in randomised controlled trials . European Journal of Cancer . 2008 ;44 (13 ):1793 –8 . 10.1016/j.ejca.2008.05.008 18599286 77 Efficace F , Kemmler G , Vignetti M , Mandelli F , Molica S , Holzner B . Health-related quality of life assessment and reported outcomes in leukaemia randomised controlled trials—A systematic review to evaluate the added value in supporting clinical decision making . European Journal of Cancer . 2008 ;44 (11 ):1497 –506 . 10.1016/j.ejca.2008.03.017 18555682 78 Lipscomb J , Reeve BB , Clauser SB , Abrams JS , Bruner DW , Burke LB , et al Patient-reported outcomes assessment in cancer trials: taking stock, moving forward . J Clin Oncol . 2007 ;25 (32 ):5133 –40 . .17991933 79 Lemieux J , Goodwin PJ , Bordeleau LJ , Lauzier S , Theberge V . Quality-of-life measurement in randomized clinical trials in breast cancer: an updated systematic review (2001–2009) . Journal of the National Cancer Institute . 2011 ;103 :178 –231 . 10.1093/jnci/djq508 21217081 80 Ganz PA . Assessing the Quality and Value of Quality-of-Life Measurement in Breast Cancer Clinical Trials . Journal of the National Cancer Institute . 2011 ;103 (3 ). 81 Chan A- W , Tetzlaff JM , Altman DG , Dickersin K , Moher D . SPIRIT 2013: new guidance for content of clinical trial protocols . Lancet . 2013 ;381 (9861 ):91 –2 . 10.1016/S0140-6736(12)62160-6 23305999 82 Davidson JR , Brundage MD , Feldman-Stewart D . Lung cancer treatment decisions: Patients' desires for participation and information . Psycho-Oncology . 1999 ;8 (6 ):511 –20 . 10607984 83 Brundage M , Feldman-Stewart D , Leis A , Bezjak A . The importance of quality of life information to a lung cancer (NSCLC) chemotherapy treatment decision—Results of a randomized evaluation . Psycho-Oncology . 2007 ;16 (9 ):S7 -S. 84 Feldman-Stewart D , Brundage MD , Hayter C , Groome P , Nickel JC , Downes H , et al What questions do patients with curable prostate cancer want answered? Medical Decision Making . 2000 ;20 (1 ):7 –19 . 10638532 85 McNair AGK , Brookes ST , Davis CR , Argyropoulos M , Blazeby JM . Communicating the Results of Randomized Clinical Trials: Do Patients Understand Multidimensional Patient-Reported Outcomes? Journal of Clinical Oncology . 2010 ;28 (5 ):738 –43 . 10.1200/JCO.2009.23.9111 20065187 86 Brundage M , Feldman-Stewart D , Leis A , Bezjak A , Degner L , Velji K , et al Communicating quality of life information to cancer patients: A study of six presentation formats . Journal of Clinical Oncology . 2005 ;23 (28 ):6949 –56 . 16192583 87 Brundage M , Bass B , Jolie R , Foley K . A knowledge translation challenge: clinical use of quality of life data from cancer clinical trials . Quality of Life Research . 2011 ;20 (7 ):979 –85 . 10.1007/s11136-011-9848-0 21279446 88 Liberati A , Altman DG , Tetzlaff J , Mulrow C , Gotzsche PC , Ioannidis JPA , et al The PRISMA Statement for Reporting Systematic Reviews and Meta-Analyses of Studies That Evaluate Health Care Interventions: Explanation and Elaboration . Plos Medicine . 2009 ;6 (7 ). 89 Moher D , Hopewell S , Schulz KF , Montori V , Gotzsche PC , Devereaux PJ , et al CONSORT 2010 Explanation and Elaboration: updated guidelines for reporting parallel group randomised trials . British Medical Journal . 2010 ;340 .

PMC005xxxxxx/PMC5003377.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757096510.1371/journal.pone.0161817PONE-D-15-49648Research ArticleBiology and life sciencesGeneticsEpigeneticsRNA interferenceBiology and life sciencesGeneticsGene expressionRNA interferenceBiology and life sciencesGeneticsGenetic interferenceRNA interferenceBiology and life sciencesBiochemistryNucleic acidsRNARNA interferenceBiology and Life SciencesGeneticsGene ExpressionResearch and Analysis MethodsModel OrganismsAnimal ModelsDrosophila MelanogasterBiology and Life SciencesOrganismsAnimalsInvertebratesArthropodaInsectsDrosophilaDrosophila MelanogasterBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesMolecular Biology Assays and Analysis TechniquesGene Expression and Vector TechniquesProtein ExpressionResearch and Analysis MethodsMolecular Biology TechniquesMolecular Biology Assays and Analysis TechniquesGene Expression and Vector TechniquesProtein ExpressionBiology and Life SciencesBiochemistryHormonesLipid HormonesProgesteroneBiology and Life SciencesBiochemistryProteinsCytoskeletal ProteinsTubulinsBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesMolecular Biology Assays and Analysis TechniquesGene Expression and Vector TechniquesHyperexpression TechniquesResearch and Analysis MethodsMolecular Biology TechniquesMolecular Biology Assays and Analysis TechniquesGene Expression and Vector TechniquesHyperexpression TechniquesBiology and life sciencesGeneticsGene expressionGene regulationSmall interfering RNAsBiology and life sciencesBiochemistryNucleic acidsRNANon-coding RNASmall interfering RNAsPractical Recommendations for the Use of the GeneSwitch Gal4 System to Knock-Down Genes in Drosophila melanogaster GeneSwitch Gal4 and DrosophilaScialo Filippo Sriram Ashwin Stefanatos Rhoda Sanz Alberto *Institute for Cell and Molecular Biosciences, Campus for Ageing and Vitality, University of Newcastle, Newcastle-upon-Tyne, NE4 5PL, United KingdomAntoniewski Christophe EditorCNRS UMR7622 & University Paris 6 Pierre-et-Marie-Curie, FRANCECompeting Interests: The authors have declared that no competing interests exist. Conceptualization: A. Sanz. Formal analysis: FS A. Sriram RS A. Sanz. Funding acquisition: A. Sanz. Investigation: FS A. Sriram RS A. Sanz. Methodology: A. Sanz. Project administration: A. Sanz. Supervision: A. Sanz. Validation: FS A. Sriram. Visualization: FS. Writing – original draft: A. Sanz. Writing – review & editing: FS RS A. Sanz. * E-mail: Alberto.Sanz@ncl.ac.uk29 8 2016 2016 11 8 e016181713 11 2015 13 8 2016 © 2016 Scialo et al2016Scialo et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Drosophila melanogaster is a popular research model organism thanks to its’ powerful genetic tools that allow spatial and temporal control of gene expression. The inducible GeneSwitch Gal4 system (GS) system is a modified version of the classic UAS/GAL4 system which allows inducible regulation of gene expression and eliminates background effects. It is widely acknowledged that the GS system is leaky, with low level expression of UAS transgenes in absence of the inducer RU-486 (the progesterone analog that activates the modified GAL4 protein). However, in the course of our experiments, we have observed that the extent of this leak depends on the nature of the transgene being expressed. In the absence of RU-486, when strong drivers are used to express protein coding transgenes, leaky expression is low or negligible, however expression of RNA interference (RNAi) transgenes results in complete depletion of protein levels. The majority of published studies, using the GS system and RNAi transgenes validate knock-down efficiency by comparing target gene mRNA levels between induced and non-induced groups. Here, we demonstrate that this approach is lacking and that both additional control groups and further validation is required at the protein level. Unfortunately, this experimental limitation of the GS system eliminates “the background advantage”, but does offer the possibility of performing more complex experiments (e.g. studying depletion and overexpression of different proteins in the same genetic background). The limitations and new possible applications of the GS system are discussed in detail. http://dx.doi.org/10.13039/501100000781European Research Council260632 - ComplexI&AgingSanz Alberto http://dx.doi.org/10.13039/501100000268Biotechnology and Biological Sciences Research CouncilBB/M023311/1Sanz Alberto This study is supported by the European Research Council (260632 - ComplexI&Aging) and the BBSRC (Biotechnology and Biological Sciences Research Council) (BB/M023311/1). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Drosophila melanogaster, popularly known as the fruit fly, is a powerful model organism to study genetic interactions, including those associated with human disease [1]. An extensive collection of genetic tools developed by the fly community over the last 50 years support the activity of researchers working with Drosophila. A non-exhaustive list of those tools includes [2–6]: (i) P-element mediated mutagenesis, (ii) site-specific recombination through the FLP-FRT system, (iii) site-specific transgene insertion using the PhiC31 integrase, (iv) silencing of specific genes using RNA interference (RNAi) and (v) genome editing using CRISPR-Cas9. The majority of the aforementioned techniques are in some way coupled to the GAL4-UAS (upstream activator sequence) binary expression system that allows spatial and temporal regulation of gene expression. The GAL4-UAS system, developed by Brand and Perrimon to study development [6] has two components: (i) the yeast transcription factor GAL4 that is placed under the control of either a ubiquitous or tissue-specific promoter, and (ii) a gene of interest cloned downstream of a UAS sequence. The Transgene is expressed upon binding of GAL4 to the UAS sequence (Fig 1A). 10.1371/journal.pone.0161817.g001Fig 1 Schematic representation of the GAL4/GeneSwitchGal4-UAS systems. (A) The GAL4-UAS system allows spatial control of gene expression. (B) The GeneSwitch (GS) system allows temporal control of gene expression thanks to a modified GAL4 protein that is active only when the synthetic progesterone analogue (mifespristone, RU-486) binds to the fused progesterone steroid receptor. In absence of RU-486, GAL4 activity is maintained at a minimum. Pictures of Drosophila male and female where obtained from: https://commons.wikimedia.org/wiki/File:Biology_Illustration_Animals_Insects_Drosophila_melanogaster.svg. Spatial control of gene expression is simple using GAL4-UAS and relies on the use of tissue-specific promoters. However, temporal control of gene expression requires co-expression of GAL4 with a temperature sensitive repressor GAL80ts [7]. GAL80ts, however, has two different caveats. Firstly, its use requires that rearing temperature is tightly controlled and secondly, full transgene expression can only be attained at high temperatures (≥29°C). Such conditions are not always suitable, in particular when studying genes whose miss-expression is lethal during development. In order to overcome these problems, the chemically inducible GeneSwitch-GAL4 (GS) system was developed[8]. GS uses a modified GAL4 protein fused to a progesterone steroid receptor, allowing the regulation of its GAL4 activity via the presence or absence of the synthetic progesterone analogue mifespristone (RU-486). In the presence of RU-486, the transactivating activity of GAL4 is enhanced leading to increased transgene expression. Conversely, in the absence of RU-486 GAL4 activity is maintained at a minimum (Fig 1B). Since the modified GAL4 protein is active in the absence of RU-486, appropriate controls must be used to quantify the extent of the leak in different experimental conditions. Indeed, expression of reporter proteins such as lacZ or GFP has been previously shown in non-induced conditions [9], however the differences between induced and non-induced conditions were clear and so considered acceptable. However, we have observed that the level of transgene expression in non-induced conditions is dependent on the nature of the transgene. For example, we observed negligible or minimal expression of protein coding transgenes in the absence of RU-486, using two different promoters: (i) daughterless-Gene-Switch (daGS) and (ii) tubulin-Gene-Switch (tubGS). On the other hand using the same drivers to express RNA interference (RNAi) transgenes to knock-down gene expression resulted in a significant decrease in target gene protein levels which was independent of the presence of RU-486. In fact, when using a strong driver such as tubGS no significant differences in gene expression were observed between induced and non-induced flies. Importantly, we show that validation of target knock-down by comparing mRNA levels of induced versus non-induced flies can be misleading and that inclusion of controls which do not carry the RNAi transgene as well as further validation at the protein level is advisable. Finally, we discuss based on the observations we report here, which experiments should and should not be performed using the GS system. Material and Methods Fly husbandry Virgin females carrying a daughterless (daGS)- or tubulin-GeneSwitch (tubGS) driver were crossed with (i) the following RNAi lines: 13131 (CG6020), 46799 (CG3683) 42162 (CG8905), 40466 (CG3731), 30892 (CG11015), 34664 (CG3612), (ii) UAS-lines carrying protein coding transgenes (NDI1, AOX, LacZ and ND-42-HA) or (iii) a wild type stock of Dahomey males [10]. Additionally, Dahomey virgin females were crossed with males carrying RNAi transgene against CG6020 as described above. Flies were collected following eclosion and transferred to new food for mating for 24 hours before being sorted for experiments. Mated 5 day old female flies maintained at 25°C were used for all experiments. Flies were maintained on standard media (1% agar, 1.5% sucrose, 3% glucose, 3.5% dried yeast, 1.5% maize, 1% wheat, 1% soya, 3% treacle, 0.5% propionic acid, 0.1% Nipagin) with a controlled 12hr:12hr light:dark cycle. All RNAi lines were obtained from the Vienna Drosophila Resource Center (VDRC) [5]. UAS-NDI1 and UAS-AOX have been previously described [11]. UAS-ND42-HA flies were a kind gift from the laboratory of Prof Hugo Bellen [12], UAS-LacZ was obtained from the Bloomington Drosophila Stock Center (BDSC) [13]. The daGS and tubGS drivers were a generous gift from the laboratories of Dr Veronique Monnier and Dr Scott Pletcher respectively. qPCR RNA extraction, cDNA synthesis and qPCR were performed as described in [14]. Primer sequences are available upon request. Western blots Sample preparation and western blotting were performed as described in [15]. The primary antibodies, employed together with the appropriate secondary antibodies, were as follows: anti-AOX described in[16], used at 1:100,000; anti-NDI1 described in [10], used at 10,000; anti-LacZ (Abcam, Oregon, USA), used at 1:1,000; anti- SOD2 (Abcam, Oregon, USA), used at 1:1,000; anti-NDUFA8 and anti-NDUFA9 (a gift from Prof Howy Jacobs (University of Helsinki)), used at 1:1,000 and 1:2,500 respectively; anti-ATP5A (Abcam, Oregon, USA), 1:500,000; anti-HA (Human influenza hemagglutinin), used at 1:1,100; anti-beta Tubulin (Abcam, Oregon, USA), used at 1:2,000 and anti-GAPDH (Everest Biotech, Oxfordshire, United Kingdom), 1:40,000. The secondary antibodies were as follows: HRP-conjugated horse anti-mouse IgG [H+L] (Vector Laboratories, Burlingame, USA), used at 1:10,000; HRP-conjugated horse anti-rabbit IgG [H+L] (Vector Laboratories, Burlingame, USA), 1:10,000; and HRP-conjugated horse anti-goat IgG [H+L] (Vector Laboratories, Burlingame, USA), 1:5,000. High resolution respirometry Oxygen consumption was measured using an O2K OROBOROS oxygraph (OROBOROS instruments, Innsbruck, Austria) as described in [17] with modifications. Briefly, whole fly homogenates were used for respirometry measurements. Briefly, 20–40 flies were homogenised in mitochondrial isolation buffer (250 mM sucrose, 5 mM Tris-HCl pH 7.4, 2 mM EGTA) and filtered before being immediately measured using an OROBOROS O2k oxygraph. Homogenates were incubated in assay buffer (120 mM KCl, 5 mM KH2PO4, 3 mM Hepes, 1 mM EGTA, 1 mM MgCl2, 0.2% bovine serum albumin, pH 7.2 at the same temperatures in which the flies were aged). State 4 respiration was measured by the addition of 5 mM pyruvate and 5 mM proline. State 3 was initiated with the addition of 1 mM ADP. CI-linked respiration was inhibited by 0.5 μM rotenone and 20 mM glycerol 3-phosphate was added to stimulate CIII-linked respiration. CIII-linked respiration was inhibited with the addition of 2.5 μM antimycin A. CIV respiration was initiated by the addition of 4 mM ascorbate and 2 mM TMPD. CIV respiration was inhibited with 0.5 mM KCN. Values were normalised to protein concentration as calculated by the Bradford method. Statistical analysis Data are shown as mean ± SEM. Data analysis was performed with Prism 6 (GraphPad) using either 1-way ANOVA with Newman-Keuls post-test or the unpaired Student’s t-test where appropriate. p values <0.05 were taken as statistically significant. A summary of the raw data for Figs 2–6 is shown as supplementary information in S1 Table. In all figures * = p<0.05 denotes significant difference from all other groups without * unless indicated otherwise by line art. n.d. = not detected. 10.1371/journal.pone.0161817.g002Fig 2 No significant leak is detected when expressing protein coding transgenes using GS. (A) qPCR expression data of NDI1 in induced vs. non-induced flies. Controls without the RNAi transgene are also included (n = 3). (B) Western blot analysis of NDI1 levels in induced vs non-induced flies. (C) Quantification of B (n = 2). (D) Western blot analysis of AOX expression driven by two (++) or one (+) copy of daughterless-GeneSwitch GAL4 (daGS) in flies carrying two (++) or one copy (+) of the AOX transgene in induced vs non-induced groups. (E) Quantification of D (n = 3). (F) Western blot analysis of NDI1 expression driven by two (++) or one (+) copy of daughterless-GeneSwitch GAL4 (daGS) in flies carrying two (++) or one copy (+) of the NDI1 gene in induced vs non-induced groups. (G) Quantification of F (n = 3). (H) Western blot analysis of AOX expression driven by two (++) or one (+) copy of tubulin-GeneSwitch GAL4 (tubGS) in flies carrying two (++) or one copy (+) of the AOX gene in induced vs non-induced groups. (I) Quantification of H (n = 3). (J) Western blot analysis of NDI1 expression driven by two (++) or one (+) copy of tubulin-GeneSwitch GAL4 (tubGS) in flies carrying two (++) or one copy (+) of the NDI1 gene in induced vs non-induced groups. (K) Quantification of J (n = 3). GAPDH or Tubulin is shown as loading control. +/- indicate presence/absence of the transgene or 500 μM RU-486 (inducer). 10.1371/journal.pone.0161817.g003Fig 3 Expression of protein coding transgenes correlates with the presence of RU-486 in the fly food. (A) Western blot analysis of ND-42-HA levels after feeding flies RU-486 for 5 days and after withdrawal of the drug from the food. (B) Quantification of A (n = 3). (C) Western blot analysis of LacZ levels driven by daGS or tubGS GAL4 in induced vs non-induced groups. (D) Quantification of C (n = 3). Tubulin is shown as loading control. daGS, daughterless-GeneSwitch GAL4; tubGS, tubulin-GeneSwitch GAL4; +/- indicate presence/absence of the transgene or 500 μM RU-486 (inducer). 10.1371/journal.pone.0161817.g004Fig 4 Use of qPCR to validate knock-down of a gene using GS requires additional controls. (A) qPCR expression data of ND-39 (CI subunit) in induced vs non-induced groups (n = 3). (B) qPCR expression data of UQCR-C1 (CIII subunit) in induced vs. non-induced groups (n = 3). (C) qPCR expression data of COX5B (CIV subunit) in induced vs. non-induced groups (n = 3). (D) qPCR expression data of Bellwether (CV subunit) in induced vs. non-induced groups (n = 3). daGS, daughterless-GeneSwitch GAL4; tubGS, tubulin-GeneSwitch GAL4; +/- indicate presence/absence of the transgene or 500 μM RU-486 (inducer). 10.1371/journal.pone.0161817.g005Fig 5 Expression of RNAi transgenes does not correlate with the presence of RU-486 in the fly food. (A) Western blot analysis of ND-39, NDI1 and ATP5A levels in control, induced and non-induced groups. (B) Quantification of A (n = 2). (C) Respirometry in tubGS flies with and without an ND-39 RNAi transgene (n = 3–8). (D) Western blot analysis of ND-39 and NDI1 levels in control, induced and non-induced groups. (E) Quantification of D (n = 2–3). (F) Western blot analysis of ND-39 levels in controls and experimental flies fed with RU-486 during development (1 μM), development and adulthood (1 μM and 500 μM respectively) or exclusively during adulthood (500 μM) (n = 2–3). All flies were 5 days old when proteins were extracted. Flies that were exclusively fed during development spent 5 days in food without RU-486. GAPDH is shown as loading control. daGS, daughterless-GeneSwitch GAL4; tubGS, tubulin-GeneSwitch GAL4; +/- indicate presence/absence of the transgene or 500 μM RU-486 (inducer). 10.1371/journal.pone.0161817.g006Fig 6 No difference in protein levels between induced and non-induced groups using a strong GS driver. (A) Western blot analysis of ND-19 levels in control (without the RNAi transgene), induced and non-induced groups. (B) Quantification of A (n = 3). (C) Western blot analysis of ND-39 levels in control (without the RNAi transgene), induced and non-induced groups. (D) Quantification of C (n = 3). (E) Western blot analysis of Sod2 levels in control (without the RNAi transgene), induced and non-induced groups. (D) Quantification of E (n = 3). Tubulin is shown as loading control. daGS, daughterless-GeneSwitch GAL4; tubGS, tubulin-GeneSwitch GAL4; +/- indicate presence/absence of the transgene or 500 μM RU-486 (inducer). Results and Discussion The nature of the transgene dictates the extent of non-induced expression when using the GS system GS has been used for the over-expression of endogenous and exogenous genes, with expression levels dependant on the activity of the promoter used to express the GAL4 transcription factor [8]. Accordingly, we used the daughterless (daGS) and the tubulin (tubGS) GS GAL4s to express the alternative NADH dehydrogenase internal 1 (NDI1) from Saccharomyces cerevisiae [11] at low and high levels respectively. We did not observe significant expression in non-induced flies at the mRNA (Fig 2A) or protein level (Fig 2B and 2C). Even in the presence of two copies of the NDI1 transgene, no NDI1 was detected in absence of RU-486 (Fig 2F, 2G, 2J and 2K). Similar results were observed when we expressed another alternative respiratory enzyme -the alternative oxidase: AOX [16]- using daGS (Fig 2D and 2E). However, when tubGS was used we did observe expression non-induced in flies carrying two copies of the AOX transgene (Fig 2H and 2I). Next, using a transgene for ND-42 protein tagged with the human influenza hemagglutinin epitope (ND-42-HA) we were able to confirm that levels of gene expression correlated with the presence of RU-486 in the fly food and that expression of ND-42-HA was reversible upon withdrawal of RU-486 (Fig 3A and 3B). In the past, a detectable leak in expression was reported when exogenous genes such as lacZ or green fluorescence protein (GFP) were expressed using GS[9]. The extent of the leak was dependent on the Gal4 driver utilized. Unfortunately as neither daGS nor tubGS were used in this article[9] our results cannot be compared. Nevertheless, when a strong GS GAL4 driver such as tubGS was used we found significant expression of LacZ in non-induced conditions, (Fig 3C and 3D) recapitulating previously reported results [9]. Importantly, we observed a strikingly different response when RNAi transgenes, designed to inhibit mRNA translation, were expressed in the same experimental conditions. Using an RNAi against complex I subunit: NADH dehydrogenase (ubiquinone) 39kDa subunit (ND-39) as an example, we observed a significant depletion in protein levels independent of the presence of RU-486. Firstly, using quantitative real time PCR (q-RT-PCR), an approach commonly used to validate the efficiency of the knock-down using GS, we measured mRNA levels in induced (flies fed with RU-486) and non-induced flies [18, 19]. We found a clear decrease in the levels of expression of the target gene using both daGS and tubGS (Fig 4A). However, we found that this approach could be misleading as when compared with control flies which did not carry the RNAi transgene non-induced flies displayed a strong depletion of the target gene (Fig 4A). We also observed for a further three target genes, with the level of knock down dependant on the driver and RNAi construct used, with a stronger driver (i.e. tubGS) showing lower levels of the target gene (Fig 4B–4D). Analysis of protein levels by western blot revealed depletion of ND-39 in the absence of RU-486 (Fig 5A and 5B) mirroring results at the mRNA level. As expected, the level of target gene knock down was dependent on the strength of the driver used. In the case of the strong driver tubGS, the level of depletion was 95% to 97% for non-induced versus induced, whereas with the weaker driver daGS only 45% depletion was observed in non-induced flies (Fig 5B). Significantly, NDI1 was only present in induced flies (Fig 5A and 5B), demonstrating that the extent of non-induced expression depends on the nature of the construct. We further confirmed our results using high resolution respirometry[14]. Complex I (CI)-linked respiration was strongly reduced in flies carrying the RNAi transgene in combination with a GS driver independently of the presence of RU-486 (Fig 5C). The fact that only CI-linked respiration was affected demonstrates that this is a specific consequence of the presence of an RNAi transgene against a CI subunit, and not due to an unspecific effect of inducing an RNAi response. No effect on the protein levels of ND-39 (Fig 5D and 5E) or on CI-linked respiration (data not shown) was observed in flies carrying only the RNAi transgene without a GS driver, indicating that the combination of a GS driver with an RNAi transgene results in the effects we observed. In order to understand if it is possible to rescue normal expression of the target protein after induction of the RNAi expression, as we did for the tagged ND-42 protein shown in Fig 3A, we fed flies with RU-486 during development, development and adulthood or only during adulthood. We hypothesised that if the knock-down was reversible, removal of the drug after development would return protein levels to normal. However, we found that ND-39 was depleted to a similar level independently of when RU-486 was administrated, i.e. during development, during adulthood or in both stages (Fig 5F). This indicates that it is not possible to regulate protein levels using RNAi constructs in combination with GS, and that knock-down is not reversible using this system. We found this to be the case for a further two RNAi lines tested in the same conditions ND-19 (CG3683) and Sod2 (CG8905) and for ND-39 in the absence of NDI1 (Fig 6A–6F). For all three lines the decrease at the protein level was over 80% in absence of RU-486, when tubGS was used as driver. However, when daGS was used, the depletion in the non-induced flies was between 10 and 61%. A similar phenomenon has been reported for two RNAi lines Prosβ5 (CG12323) and Prosα7 (CG1519) which in combination with the tubGS driver caused developmental lethality even in absence of RU-486[20, 21]. Although surprising to the best of our knowledge no one has previously reported that the leak in the GS system is dependent on the nature of the transgene, so it is possible that the level of non-induced expression varies depending on the background and diet. We hypothesize that non-induced expression observed when RNAi transgenes are expressed may be due to a systemic amplification mechanism such as that which has been described in Caenorhabditis elegans [22]. In fruit flies however there is no clear evidence for the existence of either secondary small interference RNA or of the RNA-dependent RNA-polymerase (RdRP) that drives this process in worms (reviewed in [23]) and therefore the phenomenon we observe here remains unclear at the mechanistic level. The fact that many other laboratories including ourselves have reported phenotypic differences between induced and non-induced groups could be due to a variety of reasons. Firstly, we have used whole flies for our analysis so it is possible that some tissues or cell types are more refractory to the phenomenon we observe, and that the differences reported are due the variations in protein levels in these tissues. Secondly, it is also possible that the presence of inducer during development accelerates protein depletion and specific changes during this period are responsible for the phenotypes described. Indeed, differences in diet composition during development have strong effects on the body composition of adults [24]. Thirdly, it is also possible that there is an interaction between RU-486 and knock-down of specific genes. Finally, we have observed that high concentrations of RU-486 induce the generation of mitochondrial ROS (Scialo and Sanz, unpublished observation) that in interaction with the knock-down can explain the phenotypes previously reported. Final remarks Dissecting the role of a gene requires manipulation of its function in different tissues and at different developmental stages which can be technically challenging. The GS system was generated to allow the induction of gene expression at specific time points during the fly life cycle and to solve issues related to genetic background. However, our studies indicate that the use of GS may present some technical challenges which prevent this when RNAi transgenes are expressed. Firstly, the selective leakiness of the system means that even when using weak GS Gal4 drivers (e.g. daughterless) the levels of the target protein are significantly reduced compared with control flies carrying only the RNAi transgene or the GS Gal4 driver. This necessitates additional controls negating the background advantage of using GS. However, the phenomenon described here does allow the study of the effects of overexpression of one or more genes in a background where other genes are being depleted. As an example, we demonstrate that it is possible to study the effect of ectopically expressed NDI1 in a background where ND-39 has been depleted. Similarly, other combinations of genes/proteins are possible, thus further increasing the experimental flexibility available when working with fruit flies. This would allow for genetic screens where overexpression of target genes is induced through the addition of RU-486 to the food, to rescue phenotypes resulting from RNAi depletion of target genes in the absence of RU-486. Using this approach it would be possible to test if knockdown of CI subunits can be rescued through for example the overexpression of mitophagy pathway components (e.g. Parkin, Pink1, Drp1, etc). Further combinations are possible, increasing the repertoire of tools available for manipulation of gene function in Drosophila melanogaster. Supporting Information S1 Table Summary of raw data of Figs 2–6. (XLSX) Click here for additional data file. A.S. is supported by the European Research Council (260632—ComplexI&Aging) and the BBSRC (BB/M023311/1). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. ==== Refs References 1 Clark IE , Dodson MW , Jiang C , Cao JH , Huh JR , Seol JH , et al Drosophila pink1 is required for mitochondrial function and interacts genetically with parkin . Nature . 2006 ;441 (7097 ):1162 –6 . Epub 2006/05/05. 10.1038/nature04779 .16672981 2 Matthews KA , Kaufman TC , Gelbart WM . Research resources for Drosophila: the expanding universe . Nat Rev Genet . 2005 ;6 (3 ):179 –93 . 10.1038/nrg1554 .15738962 3 Venken KJ , Bellen HJ . Emerging technologies for gene manipulation in Drosophila melanogaster . Nat Rev Genet . 2005 ;6 (3 ):167 –78 . 10.1038/nrg1553 .15738961 4 Gratz SJ , Cummings AM , Nguyen JN , Hamm DC , Donohue LK , Harrison MM , et al Genome engineering of Drosophila with the CRISPR RNA-guided Cas9 nuclease . Genetics . 2013 ;194 (4 ):1029 –35 . 10.1534/genetics.113.152710 23709638 5 Dietzl G , Chen D , Schnorrer F , Su KC , Barinova Y , Fellner M , et al A genome-wide transgenic RNAi library for conditional gene inactivation in Drosophila . Nature . 2007 ;448 (7150 ):151 –6 . Epub 2007/07/13. nature05954 [pii] 10.1038/nature05954 .17625558 6 Brand AH , Perrimon N . Targeted gene expression as a means of altering cell fates and generating dominant phenotypes . Development . 1993 ;118 (2 ):401 –15 . .8223268 7 Suster ML , Seugnet L , Bate M , Sokolowski MB . Refining GAL4-driven transgene expression in Drosophila with a GAL80 enhancer-trap . Genesis . 2004 ;39 (4 ):240 –5 . 10.1002/gene.20051 .15286996 8 Osterwalder T , Yoon KS , White BH , Keshishian H . A conditional tissue-specific transgene expression system using inducible GAL4 . Proc Natl Acad Sci U S A . 2001 ;98 (22 ):12596 –601 . 10.1073/pnas.221303298 11675495 9 Poirier L , Shane A , Zheng J , Seroude L . Characterization of the Drosophila gene-switch system in aging studies: a cautionary tale . Aging Cell . 2008 ;7 (5 ):758 –70 . 10.1111/j.1474-9726.2008.00421.x .18691185 10 Sanz A , Fernandez-Ayala DJ , Stefanatos RK , Jacobs HT . Mitochondrial ROS production correlates with, but does not directly regulate lifespan in Drosophila . Aging (Albany NY) . 2010 ;2 (4 ):200 –23 . Epub 2010/05/11. v2/n4/full/100137.html [pii]. 20453260 11 Sanz A , Soikkeli M , Portero-Otin M , Wilson A , Kemppainen E , McIlroy G , et al Expression of the yeast NADH dehydrogenase Ndi1 in Drosophila confers increased lifespan independently of dietary restriction . Proc Natl Acad Sci U S A . 2010 ;107 (20 ):9105 –10 . Epub 2010/05/04. 0911539107 [pii] 10.1073/pnas.0911539107 20435911 12 Zhang K , Li Z , Jaiswal M , Bayat V , Xiong B , Sandoval H , et al The C8ORF38 homologue Sicily is a cytosolic chaperone for a mitochondrial complex I subunit . J Cell Biol . 2013 ;200 (6 ):807 –20 . 10.1083/jcb.201208033 23509070 13 Park J , Kim Y , Choi S , Koh H , Lee SH , Kim JM , et al Drosophila Porin/VDAC affects mitochondrial morphology . PLoS One . 2010 ;5 (10 ):e13151 10.1371/journal.pone.0013151 20949033 14 Stefanatos R , Sriram A , Kiviranta E , Mohan A , Ayala V , Jacobs HT , et al dj-1beta regulates oxidative stress, insulin-like signaling and development in Drosophila melanogaster . Cell Cycle . 2012 ;11 (20 ):3876 –86 . Epub 2012/09/18. 10.4161/cc.22073 22073 [pii]. 22983063 15 Scialo F , Sriram A , Naudi A , Ayala V , Jove M , Pamplona R , et al Target of rapamycin activation predicts lifespan in fruit flies . Cell Cycle . 2015 ;14 (18 ):2949 –58 . 10.1080/15384101.2015.1071745 .26259964 16 Fernandez-Ayala DJ , Sanz A , Vartiainen S , Kemppainen KK , Babusiak M , Mustalahti E , et al Expression of the Ciona intestinalis alternative oxidase (AOX) in Drosophila complements defects in mitochondrial oxidative phosphorylation . Cell Metab . 2009 ;9 (5 ):449 –60 . Epub 2009/05/07. S1550-4131(09)00064-3 [pii] 10.1016/j.cmet.2009.03.004 .19416715 17 Kemppainen KK , Rinne J , Sriram A , Lakanmaa M , Zeb A , Tuomela T , et al Expression of alternative oxidase in Drosophila ameliorates diverse phenotypes due to cytochrome oxidase deficiency . Hum Mol Genet . 2014 ;23 (8 ):2078 –93 . 10.1093/hmg/ddt601 24293544 18 Copeland JM , Cho J , Lo T Jr, Hur JH , Bahadorani S , Arabyan T , et al Extension of Drosophila life span by RNAi of the mitochondrial respiratory chain . Curr Biol . 2009 ;19 (19 ):1591 –8 . Epub 2009/09/15. S0960-9822(09)01586-3 [pii] 10.1016/j.cub.2009.08.016 .19747824 19 Sun X , Wheeler CT , Yolitz J , Laslo M , Alberico T , Sun Y , et al A mitochondrial ATP synthase subunit interacts with TOR signaling to modulate protein homeostasis and lifespan in Drosophila . Cell reports . 2014 ;8 (6 ):1781 –92 . 10.1016/j.celrep.2014.08.022 25220459 20 Tsakiri EN , Sykiotis GP , Papassideri IS , Terpos E , Dimopoulos MA , Gorgoulis VG , et al Proteasome dysfunction in Drosophila signals to an Nrf2-dependent regulatory circuit aiming to restore proteostasis and prevent premature aging . Aging Cell . 2013 ;12 (5 ):802 –13 . 10.1111/acel.12111 23738891 21 Ford D , Hoe N , Landis GN , Tozer K , Luu A , Bhole D , et al Alteration of Drosophila life span using conditional, tissue-specific expression of transgenes triggered by doxycyline or RU486/Mifepristone . Exp Gerontol . 2007 ;42 (6 ):483 –97 . 10.1016/j.exger.2007.01.004 17349761 22 Pak J , Maniar JM , Mello CC , Fire A . Protection from feed-forward amplification in an amplified RNAi mechanism . Cell . 2012 ;151 (4 ):885 –99 . 10.1016/j.cell.2012.10.022 23141544 23 Karlikow M , Goic B , Saleh MC . RNAi and antiviral defense in Drosophila: setting up a systemic immune response . Dev Comp Immunol . 2014 ;42 (1 ):85 –92 . 10.1016/j.dci.2013.05.004 .23684730 24 Rovenko BM , Kubrak OI , Gospodaryov DV , Perkhulyn NV , Yurkevych IS , Sanz A , et al High sucrose consumption promotes obesity whereas its low consumption induces oxidative stress in Drosophila melanogaster . J Insect Physiol . 2015 ;79 :42 –54 . 10.1016/j.jinsphys.2015.05.007 .26050918

PMC005xxxxxx/PMC5003378.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 27571414PONE-D-16-0251610.1371/journal.pone.0160834Research ArticleEngineering and TechnologySignal ProcessingBiology and Life SciencesCell BiologySignal TransductionCell SignalingMembrane Receptor SignalingResearch and Analysis MethodsMathematical and Statistical TechniquesStatistical MethodsMultivariate AnalysisPrincipal Component AnalysisPhysical SciencesMathematicsStatistics (Mathematics)Statistical MethodsMultivariate AnalysisPrincipal Component AnalysisBiology and Life SciencesImmunologyImmune ResponseInflammationMedicine and Health SciencesImmunologyImmune ResponseInflammationMedicine and Health SciencesDiagnostic MedicineSigns and SymptomsInflammationMedicine and Health SciencesPathology and Laboratory MedicineSigns and SymptomsInflammationResearch and Analysis MethodsSimulation and ModelingPhysical SciencesMathematicsStatistics (Mathematics)Statistical DataBiology and Life SciencesCell BiologyCellular Structures and OrganellesCytoplasmBiology and Life SciencesCell BiologyCellular Structures and OrganellesCell MembranesStatistical Techniques Complement UML When Developing Domain Models of Complex Dynamical Biosystems Developing Domain Models Using Statistics and UMLhttp://orcid.org/0000-0001-6333-9448Williams Richard A. 12¤*Timmis Jon 23Qwarnstrom Eva E. 45 1 Department of Computer Science, University of York, York, United Kingdom 2 York Computational Immunology Laboratory, University of York, York, United Kingdom 3 Department of Electronics, University of York, York, United Kingdom 4 Department of Infection, Immunity and Cardiovascular Disease, Medical School, University of Sheffield, Sheffield, United Kingdom 5 Affiliated, Department of Pathology, School of Medicine, University of Washington, Seattle, Washington, United States of America Castiglione Filippo Editor Consiglio Nazionale delle Ricerche, ITALY Competing Interests: The authors of this manuscript have read the journal’s policy and have the following competing interests: JT is Director of SimOmics Ltd. This does not alter the authors’ adherence to PLOS ONE policies on sharing data and materials. Conceived and designed the experiments: RAW JT EEQ. Performed the experiments: RAW. Analyzed the data: RAW JT EEQ. Wrote the paper: RAW JT EEQ. ¤ Current address: Department of Management Science, Lancaster University, Lancaster, United Kingdom * E-mail: r.williams4@lancaster.ac.uk2016 29 8 2016 11 8 e016083419 1 2016 26 7 2016 © 2016 Williams et al2016Williams et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Computational modelling and simulation is increasingly being used to complement traditional wet-lab techniques when investigating the mechanistic behaviours of complex biological systems. In order to ensure computational models are fit for purpose, it is essential that the abstracted view of biology captured in the computational model, is clearly and unambiguously defined within a conceptual model of the biological domain (a domain model), that acts to accurately represent the biological system and to document the functional requirements for the resultant computational model. We present a domain model of the IL-1 stimulated NF-κB signalling pathway, which unambiguously defines the spatial, temporal and stochastic requirements for our future computational model. Through the development of this model, we observe that, in isolation, UML is not sufficient for the purpose of creating a domain model, and that a number of descriptive and multivariate statistical techniques provide complementary perspectives, in particular when modelling the heterogeneity of dynamics at the single-cell level. We believe this approach of using UML to define the structure and interactions within a complex system, along with statistics to define the stochastic and dynamic nature of complex systems, is crucial for ensuring that conceptual models of complex dynamical biosystems, which are developed using UML, are fit for purpose, and unambiguously define the functional requirements for the resultant computational model. http://dx.doi.org/10.13039/501100000287Royal Academy of EngineeringTimmis Jon http://dx.doi.org/10.13039/501100000266Engineering and Physical Sciences Research CouncilEP/K040820/1Timmis Jon http://dx.doi.org/10.13039/100004440Wellcome Trust097829Timmis Jon http://dx.doi.org/10.13039/501100000268Biotechnology and Biological Sciences Research CouncilBB/J009687/1Qwarnstrom Eva E. http://dx.doi.org/10.13039/501100000274British Heart FoundationPG/11/103/29219Qwarnstrom Eva E. RAW would like to acknowledge the White Rose Universities Consortium for an Immune Modelling Network Doctoral Studentship. JT is part-funded by The Royal Society, The Royal Academy of Engineering, and the Engineering and Physical Sciences Research Council (Grant No. EP/K040820/1). The York Computational Immunology Laboratory is part-funded by the Wellcome Trust (Ref: 097829) through the Centre for Chronic Diseases and Disorders (C2D2) at the University of York. EEQ (The Cell Biology Laboratory in the Department of Infection, Immunity and Cardiovascular Disease, at the University of Sheffield) is part-funded by the Biotechnology and Biological Sciences Research Council (Grant No. BB/J009687/1) and the British Heart Foundation (Grant No. PG/11/103/29219). Data AvailabilityThis research did not produce new empirical data, but instead used published biological data to develop a domain model of the IL-1 stimulated NF-kB signalling pathway, and to investigate the validity of using UML and statistical techniques when developing such models. The previously published articles on which this modelling research is based are referenced throughout the paper, and the subset of data used within our data analysis has been reproduced in the supporting information files.Data Availability This research did not produce new empirical data, but instead used published biological data to develop a domain model of the IL-1 stimulated NF-kB signalling pathway, and to investigate the validity of using UML and statistical techniques when developing such models. The previously published articles on which this modelling research is based are referenced throughout the paper, and the subset of data used within our data analysis has been reproduced in the supporting information files. ==== Body Introduction Over the past twenty years, a systems approach to research has become more widespread within biology. Researchers in the biological sciences are now increasingly using computer models and simulations to better understand intercellular and intracellular processes of living organisms. The merits of a systems biology approach have been discussed in depth by Kitano [1–3], with his project lifecycle diagram (Fig 1 from [2]) having the potential to become the classic diagrammatic representation of how systems biology is underpinned by a hypothesis-driven research cycle. More recently, the Complex Systems Modelling and Simulation (CoSMoS) process has been developed [4–6]. This process provides a framework of leading practice for developing and using simulations to explore complex systems, and is comparable to project lifecycle methodologies used in industry. Like these traditional methodologies, the CoSMoS process is organised around phases, which contain a set of products (deliverables), and associated activities. The CoSMoS process has three phases: Discovery phase, which establishes the scientific basis of the project, identifies and models the domain of interest, and formulates scientific questions; the Development phase, which produces the actual simulator; and the Exploration phase, which uses the simulator for in silico experimentation; the results of which are used to explore the scientific questions defined previously. Along with these phases, there are key products associated with CoSMoS projects: Domain Model, Platform Model, Simulation Platform, and the Results Model (see Fig 1). 10.1371/journal.pone.0160834.g001Fig 1 The CoSMoS Process. The CoSMoS process advocates an iterative lifecycle, consisting of three separate phases (discovery, development and exploration), and creation of four key project artefacts (domain model, platform model, simulation platform, and results model). The discovery phase focuses on formulation of the problems to be investigated through use of the computational model, resulting in creation of a functional specification of the required biological behaviour to be simulated (domain model). The development phase focuses on transforming the domain model into a technical specification (platform model) specific to the programming language(s) and computer architectures to be used, and actual development of the computational model (simulation platform), including calibration, validation and verification. The exploration phase focuses on the in silico experimentation to investigate the biological problems of interest, and the generation of predictions (documented in the results model), which facilitate the generation of novel hypotheses for subsequent testing in the biological arena. The domain model is an abstract representation of the actual system of interest (the Domain), which documents our understanding of the domain into explicit statements, that may relate to assumptions, constraints, definitions, and indeed relationships or interactions between components of the domain (discussed by Polack et al [7, 8]). One approach to semi-formally document the required functionality of a system, which has become the de facto standard for modelling software systems by the software engineering community [9], uses the Unified Modelling Language (UML). The UML specification (version 2.4) [10] defines 14 separate diagramming notations, split across three main groups: structure diagrams, which show the static structure of components within a system; behaviour diagrams, which show the dynamic behaviour(s) of components within a system; and implementation diagrams, which show the hardware and software infrastructures within a system (see Fig 2 for the taxonomy of UML diagramming notations). 10.1371/journal.pone.0160834.g002Fig 2 Taxonomy of UML diagramming notations. Structure diagrams show the static structure of the components within a system, and comprise: Class, Composite Structure, Package, Profile and Object diagrams. Behaviour diagrams show the dynamic behaviour of the components within a system, and comprise: Activity, Sequence, Communication, State Machine, Use Case, Interaction Overview, and Timing diagrams. Finally, implementation diagrams comprise Component and Deployment diagrams (after [10]). With respect to biology, UML has previously been used by Bersini et al [11] to diagrammatically model the content and functions associated with biological systems, and [12] to model Rosen’s Metabolism-Replacement system that is used within relational biology. With specific reference to domain modelling, UML has also been used by Read et al [13], to define the domain model of the intercellular interactions within an autoimmune disease (Experimental Autoimmune Encephalomyelitis, an animal form of multiple sclerosis), and Alden et al [4], to define the domain model of cell interactions within tissue formation in the immune system (specifically the lymphoid organ). In addition to UML, the Systems Biology Graphical Notation (SBGN) [14] has also been used as a diagrammatic notation to model biological systems, such as the Toll-like receptor network [15] and the mTOR signalling network [16]. The SBGN was developed by an international collaboration of biochemists, computational biologists and computer scientists, with the overriding objective to allow scientists to diagrammatically represent networks of biochemical interactions using standardised terminology and notation. Whereas UML contains 14 different notations, the SBGN currently only contains 3, being: the process diagram, the entity-relationship diagram, and the activity diagram. Taking these 3 notations in turn, firstly, the process diagrams are used for modelling the interactions that take place between biomolecules and the various state-transitions that occur as part of the biochemical reaction. They are able to convey the temporal aspects of molecular events occurring in biochemical reactions, and are analogous to UML sequence and communication diagrams. The main drawback with process diagrams appears to be that a given component must appear multiple times on the same diagram if it exists under several states, whereas in UML you can have one entity with several activities coming off, that through the use of guard conditions, can specify which activity occurs under specific circumstances. Indeed, the requirement for SBGN process diagrams to diagrammatically define all states that a component can take, can become problematic. For example, a biological component that acts as a hub in a network will have a large number of connections and therefore possible network states. These all have to be defined separately in process diagrams, which leads to the issue of combinatorial explosion identified by [17]. Secondly, the entity-relationship diagrams are based on Kohn’s molecular interaction maps and are used for modelling the relationships between biomolecules. These focus on the influences that entities have on each other, but not the state transformations that occur following interactions; they are akin to UML class diagrams and activity diagrams. Unlike the process diagrams, an entity appears only once, which is closer to the approach of UML. An enhancement over UML with respect to modelling biology is that these diagrams have specific notations for low-level biochemical reactions such as phosphorylation, which can be displayed on specific amino acid residues of protein entities. Finally, the activity flow diagrams are used for modelling the activities of biomolecules at a high-level of abstraction. They can be used to convey component-level interactions (e.g. protein-protein), without the need to show the detail of specific chemical reactions at the level of individual amino acids (e.g. phosphorylation events). As the activity diagram ignores the specific biochemical processes that entities are involved in, and their associated state transitions, they are quite compact in nature, and can be thought of as the typical network diagram found in traditional biochemical textbooks. As per UML, these 3 notations complement each other and are used to diagrammatically model different aspects/views of the biological system. With respect to the conceptual modelling of complex biological systems, we agree with Grizzi [18] who advise that complex systems can be viewed from many perspectives, and therefore can be described in many ways, each of which will be only partially true. We believe that SBGN and UML are both suitable for developing domain models of complex biological systems, however we do not believe that SBGN will be suitable for the platform model (technical specification) that will be developed in the next phase of our CoSMoS project. As such, we have selected UML as the technique of choice, so that we are able to utilise a single diagrammatic notation throughout the full lifecycle of our CoSMoS project. The Domain The NF-κB signalling pathway is one of the key signalling pathways involved in the control and regulation of the immune system [19]. Activation of the NF-κB transcription factor and signalling pathway is a tightly regulated event, involving activation of a number of signalling components [20]. NF-κB is normally sequestered in the cytosol of non-stimulated cells and consequently must be translocated into the nucleus to function as a transcriptional activator of target genes. NF-κB is activated by a wide variety of different extracellular stimuli, including proinflammatory signalling molecules, bacteria, viruses, and physical and chemical stresses [21]. As previously advocated by Kitano [2], we also believe that computational modelling and simulation can complement wet-lab experimental approaches. With specific reference to the NF-κB intracellular signalling pathway, there is potential to facilitate a more comprehensive understanding of the underlying mechanistic behaviours of the system, which could then be harnessed for identifying targets for therapeutic interventions to resolve system dysregulation [1]. Existing Ordinary Differential Equation (ODE) based models [22–24] have been useful in increasing our understanding at the cell population-level, however we believe that the field will gain further benefits from computational models at the single-cell level that contain increased scope and granularity of components over and above these mathematical models, and will also allow us to investigate the mechanistic underpinning (i.e. not just the dynamics) of the system (see [25] and our recent review [26]). Our long term objective is to build on previous work [27, 28] that used agent-based modelling, and develop a detailed model of the IL-1 stimulated NF-κB signalling pathway, for the purpose of performing in silico experimentation as a basis of hypothesis generation for the biological domain. In order to ensure that the computational model appropriately reflects the biological case-study, good software engineering practices through a principled approach to design and development, such as the CoSMoS process [5] should be adopted. This advocates the creation of a domain model, which captures the essential processes and entities of the real-world system under study; in particular, the emergent behaviour, at an appropriate level of abstraction. Having a separate domain model (akin to a functional specification) from a platform model (which details how the simulation is designed, and is akin to a technical specification), allows for the concentration on biological fact and for us to scope the system to be modelled, and therefore not be biased by implementation specific details at this early stage of the research project. There exist substantial quantities of literature on the NF-κB signalling pathway, with various aspects of the pathway being independently studied by a wide variety of labs. Furthermore, it is generally understood that representing every aspect of a real-world system in models and simulations is computationally intractable, and therefore requires focus on a subset of the properties and behaviours for subsequent model-driven investigations. One of the primary purposes of the domain model is to capture this subset of real-world system properties, and therefore provide a definition of the abstraction level taken for the modelling project. This paper presents our domain model of the IL-1 stimulated NF-κB signalling pathway, based on the previously published work of Carlotti et al [29, 30] and Yang et al [31, 32], who used fluorescent protein constructs and confocal fluorescence microscopy to investigate pathway dynamics in living cells. Data from the observations presented in the publications by Yang et al were selected for analysis to test the effectiveness in using the Unified Modelling Language in developing a domain model of the widely distributed data. Materials and Methods The domain model utilises the empirical findings of Carlotti et al [29, 30] and Yang et al [31, 32], with the statistical analysis being specifically based on a subset of data from Yang et al [32]. This subset contained measurements from single-cell analysis performed on 88 cells: 52 were transfected with IκBα Enhanced Green Fluorescent Protein (EGFP) and stimulated with IL-1; and 36 were transfected with IκBα-EGFP, but not stimulated with extracellular ligand, thus representing a control group. Single-cell analysis on live cells, include continuous monitoring of the same set of cells over time. All measurements within the data related to cytoplasmic fluorescence and were taken over a period of one hour, at intervals corresponding to 0, 10, 30 and 60 min. The subset of data used within the statistical analysis of this manuscript can be found in S1 and S2 Tables of the supplementary information. The data sets were divided into 3 groups based on transfection levels of the exogenous protein as in the original analysis by Yang et al: 0-1.5 fluorescent units (corresponding to up to 4 fold levels of the endogenous protein); 1.5-3.0 fluorescent units (4-8 fold levels of the endogenous protein); and above 3.0 fluorescent units (above 8 fold levels of the endogenous protein) [32]. The domain model was developed in an iterative manner by the modeller (RAW), senior software engineer (JT) and domain expert (EEQ), using the deep-curation approach [33]. We have chosen to follow the approach of Read et al [13] in using UML as the basis to semi-formally define the domain model of our complex dynamical biosystem. Along with a number of UML diagrammatic notations (UML v2.4, [10]), a number of less formal cartoon diagrams were also used to ensure the biological meaning could be conveyed efficiently. Furthermore, a number of statistical techniques were used to complement UML when modelling the temporal dynamics and stochastic characteristics of the system. Initial focus is the emergent system-wide behaviours of the pathway, before increasing the level of detail to the interactions between system components, and then the dynamics of individual components. The domain model is presented in a top-down manner, comprising three levels of abstraction, as defined below: A system-level overview of the domain model. This highly abstract level provides an outline of the biology of the IL-1 stimulated NF-κB signalling pathway. Particular focus is made to the behaviours of the system following induction by extracellular signal, and how these are believed to correspond to phenomena observed in the real-world domain. This abstraction level of the domain model does not make use of UML, but instead utilises less formal cartoon diagrams to convey system-wide properties, along with a number of statistical approaches to convey temporal dynamics. In particular, we have used the R data analysis and graphics software [34] to perform Chi-squared goodness of fit tests to ascertain the statistical distribution of the wet-lab data, along with hierarchical clustering and Principal Component Analysis to investigate any underlying groupings with the dataset of Yang et al [32] (see supplementary information for detailed descriptions). Modelling component-level interactions of the domain model. This medium level abstraction, decomposes the IL-1 stimulated NF-κB signalling pathway into its constituent molecular components. This level models an abstracted view of the key molecular interactions between the components, that together give rise to the emergent behaviours of the system. A cartoon diagram, along with UML communication and activity diagrams have been used in modelling these component-level interactions. Modelling individual component dynamics. This level of abstraction provides the greatest detail within the domain model, through modelling the dynamics of individual components within the system. A set of linked UML state machine diagrams have been used to develop this level of the model. Validation of the Domain Model The iterative approach to developing the domain model within a CoSMoS project provides an ability to use a number of validation techniques during model formulation. Balci [35] has extensively reviewed the verification and validation techniques that are suitable for computational model development and simulation-based experiments. We have used a subset of these techniques to validate our domain model, comprising: audits by the senior software engineer to ensure that the modelling adheres to established practices; desk checking by the modeller to ensure that individual diagrammatic and statistical models are correct, complete, consistent and unambiguous; face validation by the domain expert to compare the complete domain model against her detailed understanding and judgment of the real-world biological system; and structured walkthroughs by the whole group (modeller, senior software engineer and domain expert) to detect and document faults. Results As recently argued by Read et al [36], the domain model defines our understanding of how system-level behaviours emerge from the cumulative actions of lower-level components, such as intracellular signalling molecules. The domain model presented in this paper represents the subset of signalling components that give rise to system-level dynamics. It was developed in a top-down manner, through close alignment to the process advocated by Read et al, and comprises three levels of abstraction, at the system-level, component-level, and individual components. Modelling System-Level Properties This highly abstract level provides an outline of the biology of the IL-1 stimulated NF-κB signalling pathway. Particular focus is made to the behaviours of the system following induction by extracellular signal, and how these are believed to correspond to phenomena observed in the real-world domain. Modelling Expected Behaviours Following the approach of Andrews et al [5] and the example of Read et al [13], we have chosen to commence the domain modelling process with a cartoon-like diagram, termed an expected behaviours diagram (Fig 3). This diagram depicts the observable phenomena of the IL-1 stimulated NF-κB signalling pathway, along with the known interactions between system components that generate system-wide behaviours. The diagram also provides us with an opportunity to define a number of hypotheses on how these known component interactions may yield the observable phenomena. The expected behaviours diagram therefore provides a diagrammatic view of the relationship between the real-world domain and the domain model [13]. 10.1371/journal.pone.0160834.g003Fig 3 Expected behaviours diagram. Expected behaviours diagram depicting the observable phenomena of the IL-1 stimulated NF-κB signalling pathway; the behaviours that are hypothesised to be responsible for these phenomena; and at an abstracted level the components of the complex system that are believed to be responsible for the development of these emergent behaviours. At the highest level of the system, activation of the NF-κB pathway initiates a transitory inflammatory response (the system dynamics automatically cease the response). It is hypothesised (expected) that these phenomena occur through interaction of three functional modules that relate to activation of cell membrane receptors, amplification of the signalling cascade, and upregulation of transcription. Developed from the reviews of [37–39]. The top section of Fig 3 defines the observable phenomena of the signalling pathway, in that the pathway results in an inflammatory response against extracellular stimuli, and that after a period of time, this inflammatory response ceases. The dotted horizontal line demarcates these observable phenomena from hypotheses that are believed to be responsible for their emergence. These hypotheses consist of expected behaviours (using ‘<<expected>>’ tags) that emerge through the interactions of the underlying system components. The known interactions between system components are represented through a set of solid, directed lines, whilst the expected behaviours are linked to these system components through a set of dashed lines. Wet-lab experimental research into NF-κB since its discovery in 1986 [40], has identified that a large number of inflammatory signals (extracellular stimuli) activate cell membrane receptors to initiate its signalling pathway. Signal transduction through the intracellular network, via activation of various intermediate signalling components, amplifies the signalling cascade so that a short, transitory burst of stimuli, induces the transcription of target genes and the corresponding translation of the mRNA into proteins. One of the early genes activated by NF-κB is its inhibitor IκBα, which induces negative feedback to dampen the inflammatory response [41]. Modelling Physical Containment The spatial relationships of the components detailed within the expected behaviours diagram can be seen in the cartoon containment diagram (see Fig 4), which provides an abstract representation of a Eukaryotic cell. For the purposes of modelling the IL-1 stimulated NF-κB signalling pathway, the cellular structure can be abstracted away to contain just three cellular structures: the membrane, which for our purposes contains the cell membrane receptor and co-receptor proteins; the cytoplasm, which contains the cytosol (intracellular fluid) that further contains: adaptor proteins, intermediate signalling components, NF-κB, IκBα, and the mRNA generated from gene transcription; and the nucleus, which contains DNA, its nuclear membrane, which houses the nuclear membrane transporter proteins involved in translocation (movement of proteins between cytoplasm and nucleus), and the NF-κB and IκBα that have been translocated from the cytoplasm. 10.1371/journal.pone.0160834.g004Fig 4 Cartoon-like containment diagram. Cartoon-like containment diagram showing the physical containment of the components involved in the IL-1 stimulated NF-κB signalling pathway and the physical environment in which they are situated within a Eukaryotic cell. We believe that this is a much more intuitive way of representing physical containment than the corresponding UML class containment diagram (not shown). Developed from [42]. Modelling Dynamics The single-cell analysis work of Carlotti et al and Yang et al generated time-series fluorescence data relating to the dynamics of NF-κB translocation and IκBα degradation, which demonstrated that genetically identical cells in a standard environment display significant differences in their response to perturbations [29–32] (see S1 and S2 Tables). At a molecular level, this was demonstrated to correlate with the level of protein expression within individual cells, and to underlie the complexity and variations seen within biological populations [29–32, 43]. Unfortunately, UML does not currently have a mechanism for depicting this variation between individuals within a population. As such, we have found UML deficient in conveying the dynamics of IκBα degradation (along with the associated NF-κB release and subsequent activation), and also deficient in modelling the quantitative aspects of the signalling pathway. We have therefore used a number of statistical techniques to complement the UML and cartoon diagrams, in order to develop a more comprehensive domain model of the signalling pathway. The general aspect of variation was later discussed by Tijskens et al [44] and Elowitz et al [45], when they conjecture that a degree of variance is inherent to all aspects of biology due to the underlying stochastic physiological events of individual cells. We feel that this should be explored further within the domain model. We have used two-tailed Chi-squared (χ2) goodness of fit tests to ascertain that the single-cell analysis (IκBα degradation) fluorescence data approximates to a Negative Binomial distribution, which we believe follows the usual patterns in biology of variation due to stochasticity [46]. Figs 5 and 6 illustrate how the control and IL-1 stimulated single-cell data (at time 0 min) approximate to negative binomial distribution for the population of cells (see Supplementary Information for calculations). 10.1371/journal.pone.0160834.g005Fig 5 Histogram of control observations. Histrogram of control observations from the dataset of [31], that have been binned (grouped) using an integer interval of initial (time 0 min) fluorescence. The superimposed line represents a Negative Binomial distribution, using the median calculated from the raw data. The median average has been calculated as 1.947153. 10.1371/journal.pone.0160834.g006Fig 6 Histogram of IL-1 stimulated observations. Histogram of IL-1 stimulated observations from the dataset of [31], that have been binned (grouped) using an integer interval of initial (time 0 min) fluorescence. The superimposed line represents a Negative Binomial distribution, using the median calculated from the raw data. The median average has been calculated as 1.729876. Due to the stochastic nature of the process and the cell-to-cell variation, data on IκBα degradation by the cytokine IL-1 were expressed relative to unstimulated levels at time 0, using each cell as its own control. This is consistent with Bliss and Fisher [46], who advise that an adequate fit of data to the negative binomial distribution provides a justification for transformation of the data to stabilize the variance, as a preparatory step for further statistical analysis by other techniques. Fig 7 is a graph of the control (unstimulated) and IL-1 stimulated data using a subset of data that had initial fluorescence up to and including 1.5 arbitrary fluorescence units using median average and variance bars for interquartile ranges (25th to 75th percentiles). This accurately reproduces the findings of Yang et al, who demonstrated a pronounced reduction in IκBα degradation levels at fluorescence levels higher than 1-1.5 [31, 32]. It can be seen that good separation is gained at 30 min onwards, with a little overlap still apparent at 10 min. The rates of degradation are 0.366 fluorescence units per hour for control and 0.864 fluorescence units per hour for IL-1 stimulated. 10.1371/journal.pone.0160834.g007Fig 7 Graph of median average fluorescence. Graph of median average fluorescence for control (No IL-1) and IL-1 stimulated observations from [31]. The data has been transformed so that each cell has become its own control. The error bars illustrate the spread of observations between the 25th and 75th percentiles. Hierarchical cluster analysis was used to find similarities in the single-cell observations and to assist us in understanding the significance of the characteristics of the groups [47, 48]. This was performed using seven different clustering algorithms (Ward, single, complete, average, McQuitty, median and centroid), which are all part of the hclust function within the standard R library. The resulting dendrograms for each method (not shown) were consistent in that no clear clustering was evident between unstimulated and stimulated cells (see Supplementary Information). Further investigation used Principal Component Analysis (PCA), a powerful approach for ascertaining natural groupings and a common multivariate technique for exploration and reduction of high-dimensional data. It identifies underlying patterns within the data by producing linear combinations of the underlying orthogonal variables within the dataset [49]. Therefore it can be used to reduce the dimensionality of data for detecting underlying structures [50]. The variances associated with the four principal components within the data (PC1-4), indicated that only a single principal component (PC1) is required to explain the variation. In addition, it was found that separation of observations using PC1 was dominated by the fluorescence measurements at time 0, 10 and 30 min (see Supplementary Information). Subsequent analysis was performed on the four principal components, with the individual observations being coded depending on the relevant category. Initial comparisons representing control and IL-1 stimulated observations did not yield separation of observations. Additional granularity of coding the individual observations allowed us to compare stimulation status (control and IL-1 stimulated) against ranges of the cytoplasmic fluorescence data expressed relative to levels at time0. The best separation occurred using initial fluorescence ranges of 0-1.5 fluorescence units, consistent with the biological analysis by Carlotti et al [29] and Yang et al [31, 32]. Complete separation does not occur for any combinations, however separation emerges between control and stimulated conditions for cells with initial cytoplasmic fluorescence up to 1.5 fluorescence units. Fig 8 represents the plot of PC1 versus PC2, which separates control and IL-1 stimulated observations grouped by their initial cytoplasmic fluorescence. There is limited separation between control and IL-1 stimulated cells with initial fluorescence levels of 1.5-3.0 and no appreciable difference at initial fluorescence > 3.0 units using PCA. This accurately reproduces the findings of Yang et al [31, 32], which showed reduced activity at concentrations above 1.5 fluorescent units. Similarly, Carlotti et al [29] demonstrated that lag time and nuclear translocation rate of the transcription factor are markedly decreased at higher concentrations and that, the subsequent step involving nuclear translocation of NF-κB, was completely blocked at fluorescence units > 3.0. This likely reflects the well-controlled system of feedback mechanisms regulating the NF-κB signalling pathway. 10.1371/journal.pone.0160834.g008Fig 8 PCA plot of principal components 1 and 2. PCA plot of principal components 1 and 2, colour-coded by observation category, i.e. control versus IL-1 stimulated and range of initial cytoplasmic fluorescence. The six categories are: IL-1 stimulated/0-1.5 = Blue, IL-1 stimulated/1.5-3.0 = Red, IL-1 stimulated/>3.0 = Black, control/0-1.5 = Yellow, control/1.5-3.0 = Green, and control/>3.0 = Purple. The plot shows separation of observations with initial fluorescence < 1.5 units from the rest of the data, with partial separation between the control and IL-1 stimulated observations within this group. There is also a limited degree of separation between observations with initial fluorescence values of 1.5-3.0 units from the rest of the data, however the amount of overlap between control and IL-1 stimulated is more significant here. Modelling Component-Level Interactions This medium-level abstraction decomposes the IL-1 stimulated NF-κB signalling pathway into its constituent molecular components. This level models an abstracted view of the various molecular interactions between the components, which together give rise to the emergent behaviours of the system. Modelling the Cascade of Interactions As per the system-level properties, the NF-κB signalling pathway can be described from a high-level perspective using cartoon diagrams to communicate the interactions between system components, and in this instance the diagram can also act as a network map and illustrate the sequence of interactions between components (see Fig 9). The UML communication diagram (see Fig 10) builds on this high-level cartoon to convey the network map in a more formalised way. 10.1371/journal.pone.0160834.g009Fig 9 Cartoon diagram of high-level interactions. Simplified cartoon diagram depicting the high-level interactions between the TLR or IL-1R superfamily of receptors, the co-receptors and adaptor proteins, and the protein kinases within the NF-κB canonical signalling pathway. Diagram developed from findings of [51–55]. 10.1371/journal.pone.0160834.g010Fig 10 UML communication diagram. UML communication diagram for the IL-1 stimulated NF-κB signalling pathway. Although portraying temporal interactions as per sequence diagrams (not shown), we believe that these diagrams are more intuitive for non-Computer Science audiences as they are more flexible in relation to the position of system components, thus allowing the positioning of components to approximate to the spatial locations within a Eukaryotic cell. Developed from reviews of [56, 57]. Briefly, the network commences with an extracellular ligand (signalling molecule) binding to a cell membrane receptor (which is a member of the TLR/IL-1 receptor superfamily). The receptor then dimerises, and co-receptors such as CD14 [58], MD2 [59] (in the case of TLR4, [60]) and TILRR [55, 61] (in the case of IL-1RI/IL-1AcP) help facilitate and amplify the receptor response. In situations where the Tollip adaptor protein binds, it mediates association of IRAK protein kinase to the IL-1 receptor complex, but then inhibits IRAK [52] and transduction of the signal down the signalling pathway. Conversely, in situations where the MyD88 adaptor protein binds, it mediates association of the receptor complex with IRAK protein kinase [62], which in turn activates TRAF6 through phosphorylation [63] for propagation of the signal. Once activated, TRAF6 continues signal transduction through activation of TAK1, which subsequently activates the IKK complex [53, 54]. The activated IKK complex phosphorylates NF-κB inhibitors, such as IκBα, which facilitates its dissociation from the NF-κB molecule within the complex [64]. The released IκBα undergoes a second modification called polyubiquitination [65], which then targets IκBα for rapid degradation by the proteasome. Conversely, the released NF-κB is able to translocate from the cytosol to the nucleus, where it is subsequently activated and upregulates the transcription of target genes. Modelling Activities Another complementary UML notation that provides a view of activities within the system instead of component interactions is the activity diagram, which through the use of swim-lanes may also be used to convey the location of activities (see Fig 11). As activity diagrams focus on activities and not components, they are able to convey the individual interactions (expressed as activities) that give rise to the emergent behaviour of the system. Furthermore, the focus on activities allows us to aggregate sets of individual interactions into functional modules. This is beneficial when domain modelling, as biological systems can generally be abstracted into groupings of components by functionality. With particular reference to the IL-1 stimulated NF-κB signalling pathway, we can separate the system into three functional modules relating to cell membrane receptor activation, activation of the NF-κB signalling module, and generation of new IκBα to dampen the response through negative feedback regulation. 10.1371/journal.pone.0160834.g011Fig 11 UML activity diagram. Full end-to-end UML activity diagram for the IL-1 stimulated NF-κB signalling pathway using the concept of swim-lanes to convey sub-cellular location of components. Developed from [39, 56, 57]. As per the cartoon network diagram (Fig 9) and UML communication diagram (Fig 10), the set of activities within the system begin with extracellular stimuli and the formation of the active receptor complex. The associated signal transduction then follows, with the first activity being the activation of IRAK, which then propagates the stimuli-related signal through the pathway via phosphorylation of intermediates. Upon phosphorylation of the IκBα inhibitor which is bound to NF-κB, the activity splits into two branches: a) phosphorylated IκBα is released from the NF-κB complex and becomes degraded via the proteasome, and b) the NF-κB dimer is released, binds to an importing nuclear receptor, is translocated from the cytosol to the nucleus, and is then activated. Once active, the NF-κB dimer may bind to the promoter region of an inflammatory response gene and initiate transcription, which ultimately generates new inflammatory response proteins. Modelling Individual Component Dynamics This level of abstraction provides the greatest detail within the domain model, through modelling the dynamics of individual components within the system. The final set of UML diagrams that represent the domain model, refer to low-level dynamics of individual components and use the state machine diagram notation. Fig 12 depicts a set of linked state machine diagrams for the receptor, intermediate components, IκBα, NF-κB, nuclear transporter, inflammatory gene, and inflammatory mRNA components of the signalling pathway. We believe the ability to link individual state machine diagrams into a single end-to-end diagram provides a powerful approach for domain modelling, as it allows the low-level dynamics of components to be captured in a single diagrammatic view of the system as a whole. 10.1371/journal.pone.0160834.g012Fig 12 Linked state machine diagrams. Linked series of state machine diagrams for the IL-1 stimulated NF-κB signalling pathway. The individual components have their own state machines, which are explicitly linked using UML join notations and embedded within a single large state machine that represents the cell. Here the cell has two states relating to dormant or active. Developed using [41, 52, 56, 57, 66–68]. It can be seen that the cell membrane receptor initially starts off in a dormant state, but may become active upon binding of extracellular stimuli, along with the co-receptor and MyD88 adaptor protein (defined using UML guard notation). Conversely, and as discussed previously, the cell membrane receptor may also become inhibited upon binding of Tollip. As defined in the previous cartoon and UML diagrams, following activation of cell membrane receptor, the extracellular signal is propagated through the signalling pathway through activation of intermediate components, culminating with activation of IKK. For the purposes of the domain model, we have abstracted away the granularity of these intermediate components (e.g. IRAK, TRAF6, TAK1 and IKK) to that of a generic intermediate component, which by default is dormant, but becomes active following phosphorylation as the signal is propagated through the transduction cascade. Within the NF-κB signalling module, the IκBα inhibitor molecule by default (i.e. following creation via transcription and translation) is unbound (free), but may probabilistically bind to NF-κB when it enters an interaction boundary, and therefore enters an inhibiting state. Following activation of the IKK enzyme, the IκBα releases the NF-κB dimer, to again enter the free state, but this time is degraded and removed from the system. Similarly, the NF-κB dimer, is by default in an inhibited state within the system due to IκBα inhibition. Following IKK-mediated release by IκBα, it becomes free, and able to translocate to the nucleus where it may become active (note the guard condition), to facilitate the upregulation of inflammatory gene transcription. Should the NF-κB dimer spontaneously unbind from the promoter region of the inflammatory gene, it will once again enter the free state (upon which it may probabilistically translocate back to the cytoplasm), or alternatively new IκBα molecules, this time within the nucleus may also bind to return the NF-κB dimer to an inhibited state, upon which the nuclear localisation sequence will be masked and it will be translocated out of the nucleus into the cytoplasm [39]. As per previous UML diagrams, binding of an IκBα molecule or NF-κB dimer to a nuclear membrane transporter, transitions the transporter protein from a dormant to an active state, for the translocation of the ligand from either the cytoplasm to the nucleus, or vice versa. Following translocation of an NF-κB dimer to the nucleus and its binding to the promoter region of an inflammatory gene, the gene transitions from a dormant to an actively being transcribed state for generation of mRNA. Upon creation, the new mRNA is translocated to the cytoplasm, where it is translated into new inflammatory protein by the ribosome. Modelling Numerical Aspects of the System The three different views outlined above provide a top-down perspective of the IL-1 stimulated NF-κB signalling pathway, which reflects the hierarchical nature of complex systems. The UML and cartoon-like diagrams used so far, have been useful for semi-formally defining the relationships and dynamics at the system, component, and intra-component levels, however they have not been able to appropriately capture the numerical aspects of the signalling pathway. For example, we have found diagrammatic notations to be deficient in modelling details regarding the ratios of NF-κB molecules (in free and inhibited states) against free IκBα molecules across the cytoplasmic and nuclear compartments. Similarly, we have been unable to convey nuclear translocation dynamics or details of IκBα degradation within UML in a form that would be intuitive to biologists. Table 1 therefore defines the key rates, ratios, and physical attributes associated with the IL-1 stimulated NF-κB signalling pathway. 10.1371/journal.pone.0160834.t001Table 1 The key rates, ratios and constants. High-Level Attribute Specific Attribute Value Cell Environment Cell Volume 2,000 μm3 Nucleus Volume 100 μm3 Approx. No. per Cell IL-1RI Receptors 5,000—10,000 RelA (NF-κB) 60,000 (Endogenous) IκBα 66,000 (Endogenous), ∼135,000 (Endogenous, cytoskeleton-bound) NF-κB Cytoplasmic:Nuclear Location 10:1 Bindable NF-κB: IκBα 1:1 ratio; ∼17% NF-κB ‘free’ in resting cells Total NF-κB: IκBα 1:3 (including cytoskeleton sequestered) IL-1 Stimulated ∼20% decrease in cytoplasmic NF-κB; ∼40-fold increase in nuclear NF-κB; ∼8-fold increase in transfected v endogenous NF-κB IκBα Phosphorylation Peaks at 10 min post IL-1 stimulation Ubiquitination Peaks at 30 min post IL-1 stimulation Degradation ∼40% degraded after 10 min; ∼60% degraded after 30 min; ∼80% degraded after 60 min Nuclear Translocation NF-κB flow to nucleus 40-60 molecules/sec (max IL-1 stimulation); Following nuclear NF-κB peak, takes ∼90 min to reach basal steady-state cytoplasm: nucleus ratio Shuttling Dissociation of NF-κB-IκBα complex within cytoplasm and independent import of subunits; Continuous process, i.e. steady-state dynamics of in- and out- fluxes; Lag time in nuclear translocation following IL-1 stimulation; Negligible NF-κB-IκBα complex translocation; IκBα translocation more rapid than NF-κB; After IL-1 stimulation, increased rate of nuclear to cytoplasmic translocation The key rates, ratios and constants of the IL-1 stimulated NF-κB signalling pathway. This table provides key details for the cell environment, approximate number of key molecules, IκBα biochemistry, nuclear translocation, and ratios of NF-κB and IκBα molecular states within the cytoplasm and nucleus. Developed from the work of [27, 29–32]. Discussion Biological systems are complex, with behaviours and characteristics that result from a highly connected set of interaction networks that function through time and space. As discussed previously, the IL-1 stimulated NF-κB signalling pathway is a complex intracellular network that manifests in stochastic and dynamic responses to inflammatory stimuli. The system-wide behaviours, generated as an inflammatory response to pathogenic invasion and other physiological perturbations, emerge through the cumulative effect of low-level intracellular interactions within an individual cell, being amplified across a population of immune response cells. As such, the inherent complexity of the signalling pathway and its associated stochasticity and dynamics, renders the process of domain modelling both time consuming and non-trivial in nature. Being analogous to a functional specification (from software engineering), the primary purpose of the domain model is to clearly and unambiguously capture our abstracted view of the functionality of the real-world domain, which will be incorporated within the future iterations of the resulting computational model. We found the iterative process of domain modelling to be extremely helpful in allowing the modeller to explore the biological domain (in conjunction with the domain expert) before development of the computational model. Once complete, and validated, the domain model acts as the functional specification for the computational model, and provides a comprehensive and transparent understanding of the domain that underpins both the scope of the computational model and the resulting in silico experimentation that will be performed as part of the exploration phase of the CoSMoS project. As such, the domain model is an essential project deliverable that provides an audit trail on how the real-world biology is linked, through abstractions, assumptions, and constraints to the functionality of the computational model. Furthermore, in this specific case, the actual process of developing the domain model in conjunction with the domain expert, facilitated a much more in-depth understanding of the domain than would have been gained through published literature alone. The domain model may be a collection of informal notes relating to relevant aspects of the domain, but may also include informal sketches (such as cartoons), more formal diagrams (such as those produced with UML), mathematical equations, scientific constants (e.g. biochemical rate constants), and physical descriptors (such as size, quantity, location, and speed). The key constraint of the domain model is that it should remain free from an implementation specific focus and should therefore not contain any reference to the programming languages or workarounds, which may be required during development of the simulator. As such, the domain model should be focused on the scientific domain, and not design considerations for the resulting computational model. Through this approach of using cartoon diagrams, UML notation, statistical techniques, and descriptions of rates, ratios and constants within a table, we have developed an accurate reproduction of the biological domain. There exists substantial quantities of literature on NF-κB signalling dynamics, and various aspects of the signalling dynamics are independently studied by a wide variety of labs. It is generally understood that representing every aspect of a real-world system in models and simulations is computationally intractable. As such, a subset of the properties and behaviours from the real-world system need to be defined for subsequent investigation. One of the primary purposes of the domain model is to capture this subset of real-world system properties, at the correct abstraction level to answer the questions of scientific interest; for example, the single-cell data relating to NF-κB pathway dynamics, mean that our future computational model should be at the level of subcellular interactions and biochemical reactions within a single cell. Due to the complex, stochastic nature of the IL-1 stimulated NF-κB signalling pathway, we have been unable to develop a single diagrammatic view that could capture the various components, interactions, and dynamics of the system. It has therefore been necessary to utilise a number of different cartoons and UML notations throughout our domain modelling exercise. These different diagrammatic views allow us to capture the initiation and propagation of the signalling pathway across the inherent hierarchies of the system (i.e. system-wide, component interactions, and individual component dynamics), which we believe to be a natural progression when domain modelling and reflects the concepts of hierarchy and modularity from systems biology [69]. The use of cartoon and UML diagrams were an essential first step towards development of our domain model, however in isolation they were not enough to provide a comprehensive model. In particular, they were unable to convey the dynamics of IκBα degradation (along with the associated NF-κB release and subsequent activation), or indeed model the quantitative aspects of the signalling pathway. We therefore used a number of descriptive and multivariate statistical techniques to complement the UML diagrams, in order to develop a more comprehensive domain model of the IL-1 stimulated NF-κB signalling pathway. Through this use of statistical techniques to complement UML, we have successfully reproduced the system interactions and stochastic dynamics found in biology. Two tailed χ2 goodness of fit tests were used due to the uncertainty about direction of difference of the observed versus expected data. The full dataset has been shown to approximate to a Negative Binomial distribution (see Figs 5 and 6), which is in keeping with the findings of White and Bennetts [70] and Bliss and Fisher [46] who advise (through statistical modelling of a number of biological systems) that biological populations, be that cell or organism level, often approximate very closely to negative binomial distributions. There is also a subset of observations within the 0-3.0 fluorescence units range however, which tend to Normality when using non-integer binning frequencies (not shown). As such, future statistical tests on the data, and indeed any simulation-level data produced from in silico experimentation, should be non-parametric in nature as these are applicable to any distribution, and do not assume normality. Furthermore, due to its non-parametric nature, the central measure used should be the median average, as this is not affected to the same extent from skewed data as the mean average [71]. Additionally, we have shown that multivariate techniques may be used to classify single-cell analysis observations into groups dependent on their initial fluorescence, and to separate control from IL-1 stimulated observations within ranges of fluorescence units. Through hierarchical cluster analysis, and analysing the various PCA plots (for example, the PC1 v PC2 plot in Fig 8), it can be deduced that there is evidence of partial separation of control versus IL-1 stimulated observations, beyond which the inherent variance associated with the data becomes too great. We accept that the large degree of variation is consistent with normal biology, however believe that for the purposes of our research, we should focus on a subset of experimental data. As the advantage of single-cell analysis is lost if you pool the data and calculate an average, and with the results of the above multivariate statistical tests in mind, we propose that a series of expression levels are used for development and calibration of our future computational model. This approach agrees with earlier findings by Carlotti et al [29] who advised that cells with high expression of the enhanced green fluoresecent protein and NF-κB (RELA) construct show impaired nuclear translocation dynamics, and that these aberrant cells mask the dynamics (at the population level) of cells expressing near-physiological amounts of the fusion protein. We also believe that in order to get rational results, each cell needs to form its own control (which was also the approach taken for the model of Pogson et al [28]), in order to eliminate the wide variations observed when averaging dynamics over multiple cells, and by implication simulations. Furthermore, it is believed that such an approach would yield more consistent results as cell time-course dynamics would be expressed as a percentage of initial fluorescence for each cell. As per Read et al [72] and Bersini [73], we agree that a subset of UML notations are able to efficiently represent elements of the domain model of biological systems (in our case the IL-1 stimulated NF-κB signalling pathway). We have found activity diagrams and communication diagrams particularly effective at depicting system-wide behaviours; communication diagrams to be effective at depicting relationships between components; and state machine diagrams effective at depicting low-level dynamics within individual components. Furthermore, we have found that activity diagrams are particularly effective when used in conjunction with swim-lanes to convey the location (e.g. cytoplasm or nucleus) of activities, and sequentially linked state machine diagrams are particularly effective at depicting the end-to-end state changes within a system. Although we have found UML to be particularly useful in these cases, it does have a number of deficiencies however. Along with the issues found by Read et al [72], we have discovered a number of additional areas where the current UML standards have deficiencies in modelling biology. For example, although UML facilitates detailed information to be depicted as attributes of individual components, it relies on the reader to unpick the multitude of diagrams to collate all of the information, for example parameter values (such as size of cell, and speed of movement of intracellular components) and rate constants (such as degradation of IκBα, and translocation of components across the nuclear membrane). We believe that a table of such information would provide a more effective mechanism to convey this information, than to over-engineer a UML diagram. Furthermore, although UML allows the range of individual objects to be depicted through multiplicity, in the form of a zero to many ‘0‥*’ association, this does not effectively convey the degree of simultaneous interactions between agents. Similarly, it is well understood that observations of genetically identical, individual cells in a standardised environment often display significant differences in their response to perturbations [44], thus leading to the large degree of inherent variation within biological populations (be they cells, organisms, or communities). At a molecular level, this may be due to the varying numbers of particular proteins within a population of cells. UML does not have the ability to depict this variation, and nor was it designed to. Conclusions In this article, we have presented a domain model of the IL-1 stimulated NF-κB signalling pathway using UML and statistical techniques. UML has been advocated as a modelling language for visualization, specification, construction and documentation of software systems [74]. Although it is not a programming language, we believe that together with a modelling approach (such as agent-based modelling) and programming language (such as Java or C), UML provides an excellent mechanism to develop models of complex dynamical systems. We agree with Cook [75] that “UML is likely to influence model-driven development for the foreseeable future”, but through adoption of a principled approach to development of our domain model, we have discovered that UML has a number of deficiencies when trying to convey the stochastic, heterogeneous nature of dynamics, within complex biological systems. This lack of functionality leads us to conclude that UML should not be seen as the only tool to be used in the domain modelling process. We address this problem by utilising a number of statistical techniques in order to gain a fuller understanding of the domain, and for scoping the abstraction of the domain to be taken forward into our future computational model. Likewise, UML does not currently have the ability to depict patterns within wet-lab data, which we believe is an essential component of the domain model for complex biological systems. It is generally agreed [76–78], that the principled design and development of a conceptual model (such as the Domain Model in the CoSMoS approach) is an essential step towards ensuring the right computational model is developed. As such, our domain model represented here will serve as evidence during validation and verification of the resultant computational model (Simulation Platform) that will be developed during the exploration phase of our CoSMoS project (forthcoming). Our multi-level domain model has taken a top-down approach by looking at the emergent system-wide behaviour, followed by component interactions, and finally the individual component dynamics. As such, within our domain of interest, the statistical techniques have been used at the system-level only, due to the wet-lab (in vitro) data being based on fluorescence at the single-cell level; the other levels within our multi-level domain model have therefore been developed using UML diagrammatic notations. We therefore believe that multi-level domain models developed using UML, benefit from the complementary views that emerge from statistical analysis of the underlying in vitro data. We acknowledge however, that the ability of statistical techniques to model any interplay between the different levels within a multi-level system, is reliant on characteristics of the empirical (i.e. wet-lab) data. Rumpe and France [79] advise that different stakeholders and modellers from different domains have varying interpretations of what constitutes an appropriate UML diagram. They further advise that as the UML specification allows the modeller a degree of flexibility through the use of semantic variations, diagrams can be tailored to better support the varied requirements of individual modellers, stakeholders, and their respective domains. We therefore suggest that the statistical techniques used within this case study, along with the various cartoon-like diagrams for modelling the expected behaviours of the system (see Fig 3) and physical containment of components (see Fig 4) represent an example semantic variation point for modelling complex intracellular signalling pathways. Finally, we believe that community and industry standards, such as UML, are important for improving the communication between developers and domain experts. The use of these standards, should make the reimplementation of models by different researchers (and labs) easier, and indeed should reduce the duplication of work, and more importantly reduce implementation errors, which may become introduced through reverse engineering of existing models and manual walkthroughs of published papers. We therefore believe the use of cartoon and UML diagrams to be an essential first step towards development of a domain model, which may be published alongside the results of in silico experimental papers; however in isolation they are not enough to provide a comprehensive model, which other researchers and labs may use to reproduce computational models. In particular, cartoons and UML diagrams have been unable to convey the dynamics of IκBα degradation (along with the associated NF-κB release and subsequent activation), or indeed model the quantitative aspects of the signalling pathway. We therefore conclude that the use of descriptive and multivariate statistical techniques to complement the UML diagrams, is essential for the development of comprehensive domain models of complex biological systems, such as the IL-1 stimulated NF-κB signalling pathway. Indeed, through our principled approach for domain modelling, we have accurately reproduced the stochastic nature of the real-world system using a diagrammatic and statistical approach. Supporting Information S1 Fig Dendrogram representing the clustering of single-cell analysis observations. Dendrogram representing the clustering of observations from [31] by hierarchical cluster analysis using the complete(-linkage) method. The boxes indicate that hierarchical cluster analysis identifies the three forced clusters as observations having an initial cytoplasmic fluorescence less than 3.0, between 3.0 and 8.0, and above 8.0 fluorescence units. (TIF) Click here for additional data file. S2 Fig Scree plot of the principal components from principal component analysis of the single-cell fluorescence data. Scree plot of the principal components from principal component analysis of observations from Yang et al [32]. Each bar corresponds to its respective principal component; bar heights are the variances of the principal components. (TIF) Click here for additional data file. S3 Fig Bi-plot of PC1 and PC2 from principal component analysis of the single-cell fluorescence data. Bi-plot of PC1 and PC2 from principal component analysis of observations from Yang et al [32]. This plot shows that measurements for times 0, 10 and 30 min contribute equally to the separation of PC1 due to their virtually equivalent arrow lengths. They are not fully parallel to the PC1 axis however, and therefore also contribute slightly to PC2. (TIF) Click here for additional data file. S4 Fig Plot of loadings for PC1 following principal component analysis. Plot of loadings for principal component 1 following PCA. PC1 was chosen because this is the component which contributes most to separation of the data. It can be seen that observations with initial fluorescence between 0-3.0 and >3.0 can be separated easily as the observation between 0-3.0 units have negative loadings and >3.0 have positive loadings. Furthermore, observations for cells with initial fluorescence between 0-1.5 tend to have relatively stable loadings (around -4.5), whereas those between 1.5-3.0 begin to have more variable loadings. (TIF) Click here for additional data file. S1 Table Control observations for data analysis. Subset of control observations from the single-cell analysis of Yang et al [32] that were used within our data analysis. (PDF) Click here for additional data file. S2 Table IL-1 stimulated observations for data analysis. Subset of IL-1 stimulated observations from the single-cell analysis of Yang et al [32] that were used within our data analysis. (PDF) Click here for additional data file. S3 Table χ2 test for control observations. χ2 test for control observations approximating to a negative binomial distribution. (PDF) Click here for additional data file. S4 Table χ2 test for IL-1 stimulated observations. χ2 test for IL-1 stimulated observations approximating to a negative binomial distribution. (PDF) Click here for additional data file. S5 Table Summary of principal component analysis of the single-cell fluorescence data. Summary of principal component analysis of the single-cell fluorescence data, showing the standard deviation, proportion of variance and cumulative proportion of variance for each principal component. (PDF) Click here for additional data file. S1 File Subset of Single-Cell Observations used within our Domain Model. (PDF) Click here for additional data file. S2 File Chi-squared (χ2) goodness of fit. (PDF) Click here for additional data file. S3 File Hierarchical Cluster Analysis. (PDF) Click here for additional data file. S4 File Principal Component Analysis. (PDF) Click here for additional data file. This research did not produce new empirical data, but instead used published biological data to develop a domain model of the IL-1 stimulated NF-κB signalling pathway, and to investigate the validity of using UML and statistical techniques when developing such models. The previously published articles on which this modelling research is based are referenced throughout the paper, and the subset of data used within our data analysis has been reproduced in S1 and S2 Tables. ==== Refs References 1 Kitano H (2002 ) Computational Systems Biology . Nature 420 : 206 –210 . 10.1038/nature01254 12432404 2 Kitano H (2002 ) Systems biology: A brief overview . Science 295 : 1662 –1664 . 10.1126/science.1069492 11872829 3 Kitano H (2004 ) Biological robustness . Nature Reviews Genetics 5 : 826 –837 . 10.1038/nrg1471 15520792 4 Alden K , Timmis J , Andrews PS , Veiga-Fernandes H , Coles M (2012 ) Pairing experimentation and computational modelling to understand the role of tissue inducer cells in the development of lymphoid organs . Frontiers in Inflammation 3 : 172 . 5 Andrews PS, Polack FAC, Sampson AT, Stepney S, Timmis J (2010) The CoSMoS process, version 0.1: A process for the modelling and simulation of complex systems. Tech. Rep. YCS-2010-453, University of York 6 Bown J , Andrews PS , Deeni Y , Goltsov A , Idowu M , Polack FAC et al (2012 ) Engineering simulations for cancer systems biology . Current Drug Targets 13 : 1560 –1574 . 10.2174/138945012803530071 22974398 7 Polack FAC, Andrews PS, Ghetiu T, Read M, Stepney S, Timmis J et al. (2010) Reflections on the simulation of complex systems for science. In: Proceedings of the Fifteenth IEEE International Conference on Engineering of Complex Computer Systems (ICECCS’10): 276–285. IEEE. 8 Polack FAC, Andrews PS, Sampson AT (2009) The engineering of concurrent simulations of complex systems. In: Proceedings of the Eleventh Conference on Congress on Evolutionary Computation (CEC’09): 217–224. IEEE. 9 Chaudron MRV , Heijstek W , Nugroho A (2012 ) How effective is UML modelling? an empirical perspective on costs and benefits . Software and Systems Modelling 12 : 571 –580 . 10.1007/s10270-012-0278-4 10 Object Management Group (2011) Unified modeling language superstructure specification v2.4. http://www.omg.org/spec/UML/2.4/Superstructure 11 Bersini H , Klatzmann D , Six A , Thomas-Vaslin V (2012 ) State-transition diagrams for biologists . PLOS ONE 7 (7 ): e41165 10.1371/journal.pone.0041165 22844438 12 Zhang L , Williams RA , Gatherer D (2016 ) Rosen’s (M,R) System in Unified Modelling Language . BioSystems 139 : 29 –36 . 10.1016/j.biosystems.2015.12.006 26723228 13 Read M , Timmis J , Andrews PS , Kumar V (2009 ) A domain model of experimental autoimmune encephalomyelitis In: Stepney S , Welch PH , Andrews PS , Timmis J (eds.) CoSMoS : 9 –44 . Luniver Press 14 le Novere N , Hucka M , Mi H , Moodie S , Schreiber F , Sorokin A et al (2009 ) The Systems Biology Graphical Notation . Nature Biotechnology 27 : 735 –741 . 10.1038/nbt0909-864d 19668183 15 Oda K , Kitano H (2006 ) A Comprehensive Map of the Toll-like Receptor Signaling Network . Molecular Systems Biology 2006 0015 . 16 Caron E , Ghosh S , Matsuoka Y , Ashton-Beaucage D , Therrien M , Lemieux S et al (2006 ) A Comprehensive Map of the mTOR Signaling Network . Molecular Systems Biology 6 : 453 . 17 Tiger CF , Krause F , Cedersund G , Palmer R , Klipp E , Hohmann S et al (2012 ) A Framework for Mapping, Visualisation and Automatic Model Creation of Signal Transduction Networks . Molecular Systems Biology 8 : 578 10.1038/msb.2012.12 22531118 18 Grizzi F , Chiriva-Internati M (2005 ) The complexity of anatomical systems . Theoretical Biology and Medical Modelling 2 : 26 10.1186/1742-4682-2-26 16029490 19 Baltimore D (2011 ) NF-κB is 25 . Nature Immunology 12 (8 ): 683 –685 . 10.1038/ni.2072 21772275 20 Naumann M , Scheidereit C (1994 ) Activation of nf-κb in vivo is regulated by mulitple phosphorylations . The EMBO Journal 13 : 4597 –4607 . 7925300 21 Ghosh S , May MJ , Kopp EB (1998 ) NF-κB and Rel proteins: Evolutionarily conserved mediators of immune response . Annual Review of Immunology 16 : 225 –260 . 10.1146/annurev.immunol.16.1.225 9597130 22 Hoffmann A , Levchenko A , Scott ML , Baltimore D (2002 ) The IκB-NF-κB signaling module: Temporal control and selective gene activation . Science 298 : 1241 –1245 . 10.1126/science.1071914 12424381 23 Nelson DE , Ihekwaba AEC , Elliott M , Johnson JR , Gibney CA , Foreman BE , et al (2004 ) Oscillations in NF-κB signaling control the dynamics of gene expression . Science 306 : 704 –708 . 10.1126/science.1099962 15499023 24 Shih VFS , Kearns JD , Basak S , Savinova OV , Ghosh G , Hoffmann A (2009 ) Kinetic Control of Negative Feedback Regulators of NF-κB/RelA Determines their Pathogen- and Cytokine-Receptor Signaling Specificity . Proceedings of the National Academies of Science 106 : 9619 –9624 . 10.1073/pnas.0812367106 25 Cheng Z , Taylor B , Ourthiague DR , Hoffmann A (2015 ) Distinct single-cell signaling characteristics are conferred by the MyD88 and TRIF pathways during TLR4 activation . Science Signaling 8 : ra69 10.1126/scisignal.aaa5208 26175492 26 Williams RA , Timmis J , Qwarnstrom EE (2014 ) Computational models of the NF-κB signalling pathway . Computation 2 (4 ): 131 –158 . 10.3390/computation2040131 27 Pogson M , Holcombe M , Smallwood R , Qwarnstrom EE (2008 ) Introducing spatial information into predictive NF-κB modelling—an agent-based approach . PLOS ONE 3 : e2367 10.1371/journal.pone.0002367 18523553 28 Pogson M , Smallwood R , Qwarnstrom EE , Holcombe M (2006 ) Formal agent-based modelling of intracellular chemical interactions . BioSystems 85 : 37 –45 . 10.1016/j.biosystems.2006.02.004 16581178 29 Carlotti F , Chapman R , Dower SK , Qwarnstrom EE (1999 ) Activation of nuclear factor κB in single living cells . Journal of Biological Chemistry 274 : 37941 –37949 . 10.1074/jbc.274.53.37941 10608861 30 Carlotti F , Dower SK , Qwarnstrom EE (2000 ) Dynamic shuttling of nuclear factor κB between the nucleus and cytoplasm as a consequence of inhibitor dissociation . Journal of Biological Chemistry 275 : 41028 –41034 . 10.1074/jbc.M006179200 11024020 31 Yang L , Chen H , Qwarnstrom E (2001 ) Degradation of IκBα is limited by a postphosphorylation/ubiquitination event . Biocehmical and Biophysical Research Communications 285 : 603 –608 . 10.1006/bbrc.2001.5205 32 Yang L , Ross K , Qwarnstrom EE (2003 ) RelA control of IκBα phosphorylation . Journal of Biological Chemistry 278 : 30881 –30888 . 10.1074/jbc.M212216200 12663663 33 Ghosh S , Matsuoka Y , Asai Y , Hsin KY , Kitano H (2012 ) Software for systems biology: From tools to integrated platforms . Nature Reviews Genetics 12 : 821 –832 . 34 R Core Team (2014) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria, http://www.R-project.org/ 35 Balci O (1994 ) Validation, verification, and testing techniques throughout the life cycle of a simulation study . Annals of Operations Research 53 : 121 –173 . 10.1007/BF02136828 36 Read M , Andrews PS , Timmis J , Kumar V (2014 ) Modelling biological behaviours with the unified modelling language: An immunological case study and critique . Journal of the Royal Society Interface 11 : 20140704 10.1098/rsif.2014.0704 37 Baeuerle PA , Henkel T (1994 ) Function and activation of NF-κB in the immune system . Annual Review of Immunology 12 : 141 –179 . 10.1146/annurev.iy.12.040194.001041 8011280 38 Baldwin AS (1996 ) The NF-κB and IκB proteins: New discoveries and insights . Annual Review of Immunology 14 : 649 –681 . 10.1146/annurev.immunol.14.1.649 8717528 39 Brasier AR (2006 ) The NF-κB regulatory network . Cardiovascular Toxicology 6 : 111 –130 . 10.1385/CT:6:2:111 17303919 40 Sen R , Baltimore D (1986 ) Multiple nuclear factors interact with the immunoglobulin enhancer sequences . Cell 46 : 705 –716 . 10.1016/0092-8674(86)90346-6 3091258 41 Baeuerle PA , Baltimore D (1988 ) IκB a specific inhibitor of the NF-κB transcription factor . Science 242 : 540 –546 . 10.1126/science.3140380 3140380 42 Hoffmann A , Baltimore D (2006 ) Circuitry of nuclear factor κB signaling . Immunological Reviews 210 : 171 –186 . 10.1111/j.0105-2896.2006.00375.x 16623771 43 Schooley K , Zhu P , Dower SK , Qwarnstrom EE (2003 ) Regulation of nuclear translocation of nuclear factor-κB relA: evidence for complex dynamics at the single-cell level . Biochemical Journal 369 : 331 –339 . 10.1042/BJ20020253 12350227 44 Tijskens LMM , Konopack P , Simcic M (2003 ) Biological variance, burden or benefit? Postharvest Biology and Technology 27 : 15 –25 . 10.1016/S0925-5214(02)00191-6 45 Elowitz MB , Levine AJ , Siggia ED , Swain PS (2002 ) Stochastic gene expression in a single cell . Science 297 : 1183 –1186 . 10.1126/science.1070919 12183631 46 Bliss CI , Fisher RA (1953 ) Fitting the negative binomial distribution to biological data . Biometrics 9 (2 ): 176 –200 . 10.2307/3001850 47 Bridges CC (1966 ) Hierarchical cluster analysis . Psychological Reports 18 : 851 –854 . 10.2466/pr0.1966.18.3.851 48 Fraley C , Raftery AE (1998 ) How many clusters? which clustering method? answers via model-based cluster analysis . The Computer Journal 41 (8 ): 578 –588 : 10.1093/comjnl/41.8.578 49 Pearson K (1901 ) On lines and planes of closest fit to systems of points in space . Philosophical Magazine Series 6 2 (11 ): 559 –572 . 10.1080/14786440109462720 50 Wold S , Esbensen K , Geladi P (1987 ) Principal component analysis . Chemometrics and Intelligent Laboratory Systems 2 : 37 –52 . 10.1016/0169-7439(87)80084-9 51 Shizuo A , Kiyoshi T (2004 ) Toll-like receptor signalling . Nature Reviews Immunology 4 : 499 –511 . 10.1038/nri1391 52 Burns K , Clatworthy J , Martin L , Martinon F , Plumpton C , Maschera B et al (2000 ) Tollip, a new component of the IL-1RI pathway, links IRAK to the IL-1 receptor . Nature Cell Biology 2 : 346 –351 . 10.1038/35014038 10854325 53 Doyle SL , O’Neill LAJ (2006 ) Toll-like receptors: From the discovery of NF-κB to new insights into transcriptional regulations in innate immunity . Biochemical Pharmacology 72 : 1102 –1113 . 10.1016/j.bcp.2006.07.010 16930560 54 Kawai T , Akira S (2006 ) TLR signaling . Cell Death and Differentiation 13 : 816 –825 . 10.1038/sj.cdd.4401850 16410796 55 Zhang X , Shephard F , Kim HB , Palmer IR , McHarg S , Fowler GJS et al (2010 ) TILRR, a novel IL-1RI co-receptor, potentiates MyD88 recruitment to control Ras-dependent amplification of NF-κB . Journal of Biological Chemistry 285 : 7222 –7232 . 10.1074/jbc.M109.073429 19940113 56 Hayden MS , West AP , Ghosh S (2006 ) NF-κB and the immune response . Oncogene 25 : 6758 –6780 . 10.1038/sj.onc.1209943 17072327 57 O’Neill LAJ , Dinarello CA (2000 ) The IL-1 receptor/toll-like receptor superfamily: Crucial receptors for inflammation and host defense . Immunology Today 21 : 206 –209 . 10.1016/S0167-5699(00)01611-X 10782049 58 Wright SD , Ramos RA , Tobias PS , Ulevitch RJ , Mathison JC (1990 ) CD14, a receptor for complexes of lipopolysaccharide (LPS) and LPS binding protein . Science 249 : 1431 –1433 . 10.1126/science.1698311 1698311 59 Shimazu R , Akashi S , Ogata H , Nagai Y , Fukudome K , Miyake K et al (1999 ) MD-2, a molecule that confers lipopolysaccharide responsiveness on Toll-like receptor 4 . Journal of Experimental Medicine 189 : 1777 –1782 . 10.1084/jem.189.11.1777 10359581 60 Poltorak A , He X , Smirnova I , Liu MY , van Huffel C , Du X et al (1998 ) Defective LPS signaling in C3H/HeJ and C57BL/10ScCr mice: Mutations in Tlr4 gene . Science 282 : 2085 –2088 . 10.1126/science.282.5396.2085 9851930 61 Zhang X , Pino GM , Shephard F , Kiss-Toth E , Qwarnstrom EE (2012 ) Distinct control of MyD88 adapter-dependent and Akt kinase-regulated responses by the interleukin (IL)-1RI co-receptor, TILRR . Journal of Biological Chemistry 287 (15 ): 12348 –12352 . 10.1074/jbc.C111.321711 22262840 62 Wesche H , Henzel WJ , Shillinglaw W , Li S , Cao Z (1997 ) MyD88: An adapter that recruits IRAK to the IL-1 receptor complex . Immunity 7 : 837 –847 . 10.1016/S1074-7613(00)80402-1 9430229 63 Cao Z , Xiong J , Takeuchi M , Kurama T , Goeddel DV (1996 ) TRAF6 is a signal transducer for interelukin-1 . Nature 383 : 443 –446 . 10.1038/383443a0 8837778 64 DiDonato JA , Mercurio F , Karin M (1995 ) Phosphorylation of IκBα precedes but is not sufficient for its dissociation from NF-κB . Molecular and Cellular Biology 15 : 1302 –1311 . 10.1128/MCB.15.3.1302 7862124 65 Brown K , Gerstberger S , Carlson L , Franzoso G , Siebenlist U (1995 ) Control of IκBα proteolysis by site-specific, signal-induced phosphorylation . Science 267 : 1485 –1488 . 10.1126/science.7878466 7878466 66 Baeuerle PA , Baltimore D (1988 ) Activation of DNA-binding activity in an apparently cytoplasmic precursor of the NF-κB transcription factor . Cell 53 : 211 –217 . 10.1016/0092-8674(88)90382-0 3129195 67 Karin, M., Ben-Neriah, Y.: Phosphorylation meets ubiquitination: The control of NF-κB activity. Annual Review of Immunology 18, 621–663 (2000) 68 Siebenlist U , Franzoso G , Brown K (1994 ) Structure, regulation and function of NF-κB . Annual Review of Cell Biology 10 : 405 –455 . 10.1146/annurev.cb.10.110194.002201 7888182 69 Sauro HM (2008 ) Modularity defined . Molecular Systems Biology 4 : 166 10.1038/msb.2008.3 18277382 70 White GC , Bennetts RE (1996 ) Analysis of frequency count data using the negative binomial distribution . Ecology 77 : 2549 –2557 . 10.2307/2265753 71 Siegel S (1957 ) Nonparametric statistics . The American Statistician 11 (3 ): 13 –19 . 10.1080/00031305.1957.10501091 72 Read M , Timmis J , Andrews PS , Kumar V (2009 ) Using UML to model EAE and its regulatory network In: Andrews PS , Timmis J , Owens NDL , Aickelin U , Hart E , Hone A , Tyrrell AM (eds.) ICARIS . LNCS 5666: 4 –6 Springer . 73 Bersini H (2012 ) UML for ABM . Journal of Artificial Societies and Social Simulation 15 : 9 10.18564/jasss.1897 74 Booch G , Rumbaugh J , Jacobson I (1999 ) The Unified Modeling Language User Guide Reading, MA : Addison-Wesley . 75 Cook S (2012 ) Looking back at UML . Software and Systems Modelling 11 : 471 –480 . 10.1007/s10270-012-0256-x 76 Lacy LW, Randolph W, Harris B, Youngblood S, Sheehan J, Might R et al. (2001) Developing a consensus perspective on conceptual models for simulation systems. In: Proceedings of the 2001 Spring Simulation Interoperability Workshop. 77 Wang W , Brooks R (2007 ) Improving the understanding of conceptual modelling . Journal of Simulation 1 : 153 –158 . 10.1057/palgrave.jos.4250024 78 Robinson S (2011 ) Conceptual modeling for simulation: definition and requirements In: Conceptual Modeling for discrete-event simulation . Robinson S , Brooks R , Kotiadis K , van der Zee D-J (eds) Boca Raton, FL : CRC Press . 79 Rumpe B , France R (2011 ) Variability in UML language and semantics . Software and Systems Modelling 10 : 439 –440 . 10.1007/s10270-011-0210-3

PMC005xxxxxx/PMC5003380.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757097610.1371/journal.pone.0159297PONE-D-16-09343Research ArticleMedicine and Health SciencesEndocrinologyEndocrine DisordersDiabetes MellitusMedicine and Health SciencesMetabolic DisordersDiabetes MellitusSocial SciencesEconomicsHealth EconomicsMedicine and Health SciencesHealth CareHealth EconomicsMedicine and Health SciencesPeople and PlacesGeographical LocationsAsiaChinaSocial SciencesEconomicsEngineering and TechnologyManagement EngineeringRisk ManagementInsuranceBiology and Life SciencesAnatomyBody FluidsBloodBlood SugarMedicine and Health SciencesAnatomyBody FluidsBloodBlood SugarBiology and Life SciencesPhysiologyBody FluidsBloodBlood SugarMedicine and Health SciencesPhysiologyBody FluidsBloodBlood SugarMedicine and Health SciencesHematologyBloodBlood SugarSocial SciencesEconomicsMicroeconomicsUrban EconomicsEconomic Burden in Chinese Patients with Diabetes Mellitus Using Electronic Insurance Claims Data Economic Burden in Chinese DM Patientshttp://orcid.org/0000-0003-3171-0163Huang Yunyu 12Vemer Pepijn 13Zhu Jingjing 2Postma Maarten J. 134Chen Wen 2*1 Unit of PharmacoTherapy, -Epidemiology & -Economics, Department of Pharmacy, University of Groningen, Groningen, The Netherlands2 School of Public Health, Fudan University, Shanghai, China3 Institute of Science in Healthy Aging & healthcaRE (SHARE), University Medical Center Groningen (UMCG), University of Groningen, Groningen, The Netherlands4 Department of Epidemiology, University Medical Center Groningen (UMCG), University of Groningen, Groningen, The NetherlandsKhamseh Mohammad Ebrahim EditorInstitute of Endocrinology and Metabolism, ISLAMIC REPUBLIC OF IRANCompeting Interests: The authors have declared that no competing interests exist. Conceived and designed the experiments: YH WC. Analyzed the data: YH PV. Wrote the paper: YH PV MJP. Supervision of data collection: WC. Standardized the data: YH JZ. * E-mail: wenchen@fudan.edu.cn29 8 2016 2016 11 8 e01592979 3 2016 30 6 2016 © 2016 Huang et al2016Huang et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Background There is a paucity of studies that focus on the economic burden in daily care in China using electronic health data. The aim of this study is to describe the development of the economic burden of diabetic patients in a sample city in China from 2009 to 2011 using electronic data of patients’ claims records. Methods This study is a retrospective, longitudinal study in an open cohort of Chinese patients with diabetes. The patient population consisted of people living in a provincial capital city in east China, covered by the provincial urban employee basic medical insurance (UEBMI). We included any patient who had at least one explicit diabetes diagnosis or received blood glucose lowering medication in at least one registered outpatient visit or hospitalization during a calendar year in the years 2009–2011. Cross-sectional descriptions of different types of costs, prevalence of diabetic complications and related diseases, medication use were performed for each year separately and differences between three years were compared using a chi-square test or the non-parametric Kruskal-Wallis H test. Results Our results showed an increasing trend in total medical cost (from 2,383 to 2,780 USD, p = 0.032) and diabetes related cost (from 1,655 to 1,857 USD) for those diabetic patients during the study period. The diabetes related economic burden was significantly related to the prevalence of complications and related diseases (p<0.001). The overall medication cost during diabetes related visits also increased (from 1,335 to 1,383 USD, p = 0.021). But the use pattern and cost of diabetes-related medication did not show significant changes during the study period. Conclusion The economic burden of diabetes increased significantly in urban China. It is important to improve the prevention and treatment of diabetes to contribute to the sustainability of the Chinese health-care system. The authors received no specific funding for this work. Data AvailabilityData are available from the third party, Zhejiang Province Human Resources and Social Security Department, China. Data is only available upon request from researchers who meet the criteria for access to confidential data. The requests can be sent to the corresponding author (wenchen@fudan.edu.cn) for further information.Data Availability Data are available from the third party, Zhejiang Province Human Resources and Social Security Department, China. Data is only available upon request from researchers who meet the criteria for access to confidential data. The requests can be sent to the corresponding author (wenchen@fudan.edu.cn) for further information. ==== Body Background As a common chronic disease, diabetes is costly to health care systems in nearly all countries [1]. People with diabetes have more outpatient visits, use more medications, have a higher probability of being hospitalized, and are more likely to require emergency and long-term care than people without the disease [2]. In China, an estimated 20.8 million people had diabetes mellitus (DM) in 2000 [3] and this number increased markedly to 92.4 million in 2008 [4]. The updated data showed that the number of adult people with diabetes in China was 109.6 million in 2015 and this number is predicted to increase to 150.7 million by 2040 [5], approximately 11% of the Chinese adult population. This implies a considerable economic burden of diabetes and its complications making it an important public health challenge. In China, electronic health information systems were developed at hospital and city level in urban areas, which could improve the quality of care and help to support the decision making. Access to valid electronic heath data is necessary for improvement of effectiveness evaluation, disease management and reimbursement policy-making. Unfortunately, a general lack of standardized, reliable and systematic coding of diagnoses and prescriptions in the Chinese electronic health information systems is common [6]. Optimal control of diabetes disease can mitigate costs of diabetes and complications. Although the Chinese Diabetes Society has published a guideline of prevention and treatment for Type 2 Diabetes Mellitus [7], the guideline is often not followed in real-world treatment and use of medication for diabetes [8]. This may lead to non-optimal control of the course of disease and reduce the affordability of treatment and care for diabetic patients. Existing domestic public health studies of diabetes in China mainly focus on prevalence and disease burden [9, 10]. Specific effectiveness studies and economic evaluations of diabetes related medications were also reported for Chinese patients [11–17]. However, data used in these studies were usually collected by household surveys or clinical trials and were usually based on single hospitals, which may lack data precision or external validity. There is a paucity of studies that focus on the economic burden in daily care using electronic health data. The aim of this study is to describe the development of the economic burden of diabetic patients in a sample city in China from 2009 to 2011 using electronic data of patients’ claims records. The economic burden assessed includes diabetes complications and diabetes related diseases. As the potentially largest component of the costs, medication use and its related cost will be specifically detailed in the analyses. Methods Study design and cohort definition This study is a retrospective, longitudinal study in an open cohort of Chinese patients with diabetes between 2009 and 2011. The patient population consisted of all people living in a provincial capital city in east China (HZ), covered by the provincial urban employee basic medical insurance (UEBMI), one of the two nationwide urban basic medical insurances. This includes employees from provincial companies and governmental organizations in the sample city. We have chosen this sample city (HZ), as it is seen as a role model for other cities for their well-organized health information system [18], which ensures the quality of electronic data. The sample city is also a typical representative of eastern urban China considering the social economic development level. The records of all outpatient visits and hospitalizations in the electronic medical insurance claims database were tracked for inclusion. We included any patient who had at least one explicit diabetes diagnosis or received blood glucose lowering medication in at least one registered outpatient visit or hospitalization during per calendar year in the years 2009–2011. Blood glucose lowering medications were identified by the strings of ‘A10A’ (insulins and analogues) and ‘A10B’ (blood glucose lowering drugs excluding insulins) contained in Anatomical Therapeutic Chemical (ATC) codes Patients who received blood glucose lowering medications but had diagnosis of polycystic ovary syndrome (PCOS) other than diabetes were excluded since evidence showed that some blood glucose lowering drugs were used for PCOS cases [19]. All diagnosis and prescription were set by doctors and recorded in the electronic database, which were trusted to reflect the valid and reliable information. Data extraction Detailed claims records of outpatient visits and hospitalizations in each year were extracted for each included patient. Data were extracted in two types of datasets for each year. The first dataset contained the general information of each doctor visit, including patients’ internal ID (which can be tracked during different years), gender and age, visit type (outpatient visit or hospitalization), serial number of visit, date of visit, level of institution, length of stay, diagnosis and costs. Cost information contained total cost for the visit and the allocated costs, including 1) cost covered by the insurance scheme, 2) cost reimbursed by civil servant subsidies, and 3) cost covered by patients themselves, labeled out-of-pocket (OOP) payment. The second dataset contained detailed information for every treatment, examination, test, material and medications issued or prescribed during each visit, including name of issued items, form, unit price, quantity and reimbursement type in the insurance list. Dose information was not provided for any of the prescribed medication. Data standardization In China, data management in electronic data systems generally lack standardization [20]. Although the Chinese name of diagnoses and medication names were recorded in the original datasets, those names could be various and without unified coding due to physician preference or a difference in structure of information system between hospitals. To standardize, we categorized diagnoses and recoded diabetes related medications based on their Chinese names for further analyses. We categorized the diagnoses into 11 groups, which included four major complication groups, six diabetes related disease groups and a group for acute complications (S1 Table). This categorization was based on the Chinese guideline of prevention and treatment for diabetes[7]. The acute complications were identified by an experienced endocrinologist from one of the tertiary hospitals affiliated to Fudan University, Shanghai, China. All extracted diagnoses were screened for full name and terms indicating related diseases (e.g., “kidney” indicating nephropathy, “eye” indicating retinopathy) by the first and third author independently. Disagreements and indetermination were discussed with the endocrinologist to reach a consensus. Both biomedicine (called Western medicine in China) and Chinese traditional medicine play important roles in patients’ health care in China. Biomedicine was recoded into ATC codes based on the Chinese names of medication in prescriptions. All unique biomedicine names in the datasets were recoded and double-checked by 12 medical students in two rounds. All medication was categorized into six groups (S2 Table), based on the Chinese guideline of prevention and treatment for diabetes: insulins and analogues (ATC codes starting with ‘A10A’), all other blood glucose lowering drugs (ATC codes starting with ‘A10B’), antihypertensive medications (ATC codes starting with ‘C02’, ‘C03’, C07’, ‘C08’, ‘C09’) and lipid modifying medications (ATC codes starting with ‘C10’) within the biomedicine category, other biomedicine and Chinese traditional medicine. Due to the lack of official coding standards and systematic clinical evidence of effectiveness for diabetes, Chinese traditional medicines were analyzed as a single category. Data analysis Cross-sectional descriptions of different types of costs, prevalence of diabetic complications and related diseases, medication use were performed for each year separately. Different costs were defined as follows: Total medical cost: total cost of all outpatient visits and hospitalizations recorded in the insurance database in one calendar year for included patients. Diabetes related cost (DM cost): cost occurring in diabetes related medical visits (DM visits). Outpatient visits and hospitalizations were categorized as “diabetes related” when: (a) the diagnosis explicitly indicated diabetes or diabetic complications or related disease; OR (b) diabetes related biomedicines (as defined above) were prescribed during the visit. Overall medication cost during DM visits: cost of all biomedicines and Chinese traditional medicines prescribed during the diabetes related visits. Diabetes related biomedicine cost (DM biomedicine cost): cost of diabetes related biomedicines. The reported costs in 2010 and 2011 were adjusted to 2009’s price using the Chinese official reported consumer price index (3.3% in 2010, 5.4% in 2011 [21]). All costs were reported in US dollars using the exchange rate of each year (1 US dollars = 6.831, 6.770 and 6.459 Chinese Yuan in 2009, 2010 and 2011, respectively [21]) to make the results more comparable to other published studies [1, 22, 23]. Differences of hospitalization rate, prevalence of complications and related diseases and DM biomedicine use between three years were compared using a chi-square test. Cost data and percentages were compared by the non-parametric Kruskal-Wallis H test. Differences between any two years for above-mentioned outcomes were compared by a post hoc pairwise comparison adjusted for multiplicity. The association of DM costs (log transferred due to skewed distribution) with number of complications and time was tested by multivariate linear regression with an interaction of these two factors. Data preparation and statistical analyses were performed using Stata SE Version 14.0 (Stata Corporation, College Station, TX). Ethics Statement The data used in this study was anonymized before authors had access to the data. For research using anonymous electronic medical records no ethics committee approval is needed in China. Results Patient population The studied cohort included 1,668, 1,797 and 2,078 patients from 2009 to 2011, respectively (Table 1). Almost all (97.8% of patients in 2009) were included in all three years. Approximately 65% patients were male. 80.9%, 79.6% and 76.2% patients were aged from 45 to 75 years old in 2009, 2010 and 2011, respectively. From 2009 to 2011, the number of diabetic patients gradually had an apparent increasing trend in the sample city. 10.1371/journal.pone.0159297.t001Table 1 Demographic information of sampled patients. 2009 2010 2011 Total patients 1,668 1,797 2,078 Gender male 1,066(63.9%) 1,152(64.1%) 1,352(65.1%) female 602(36.1%) 645(35.9%) 726(34.9%) Age group 0–30 7(0.4%) 7(0.4%) 10(0.5%) 30–45 122(7.3%) 124(6.9%) 160(7.7%) 45–60 638(38.3%) 675(37.6%) 772(37.2%) 60–75 711(42.6%) 754(42.0%) 822(39.6%) 75+ 190(11.4%) 237(13.2%) 314(15.1%) Total annual medical cost For all patients, the annual total cost per patient increased from 2,383 to 2,780 USD from 2009 to 2011 (p = 0.032), which was mainly caused by difference between 2009 and 2011 (adjusted p = 0.006) (Table 2). Annual DM cost averagely accounted for 70% of the total cost in 2009 and 2010 and increased from 1,655 to 1,857 USD in the study period, but decreased to 67.2% in 2011 (p = 0.001, adjusted p-value for 2009 vs 2011 = 0.047). Approximately 30% of the total costs were paid by patients themselves and the average percentage of this OOP payment decreased from 29.8% in 2009 to 27.7% in 2011 (p = 0.001, adjusted p-value for 2010 vs 2011 = 0.001, 2009 vs 2011 < 0.001). 10.1371/journal.pone.0159297.t002Table 2 Average annual costs per patient (USD). 2009 2010 2011 p-value of 3-year comparison Multiplicity adjusted p values 09vs10 10vs11 09vs11 Annual costs per patient Number of patients 1,668 1,797 2,078 Total medical cost 2,382.5 2,556.2 2,780.1 0.032 0.424 0.313 0.006 OOP payment (% of total medical cost) 781.7(29.8%) 807.8(29.1%) 822.3(27.7%) 0.001* 1.000 0.331 0.047 DM cost (% of total medical cost) 1,655.0(70.3%) 1,719.4(70.0%) 1,856.5(67.2%) <0.001* 1.000 0.001 <0.001 Annual costs per patient of patients with hospitalizations Number of patients (% of total patients) 287(17.2%) 361(20.1%) 420(20.2%) 0.038 0.030 0.924 0.019 Total medical cost 6,301.4 6,249.9 7,093.9 0.558 1.000 0.456 0.412 OOP payment (% of total medical cost) 1,972.5(32.2%) 1,989.1(33.3%) 2,032.8(31.8%) 0.820* 1.000 0.798 1.000 DM cost (% of total medical cost) 4,768.2(75.9%) 4,480.8(74.3%) 5,295.9(72.0%) 0.073* 1.000 0.437 0.058 Hospitalization cost (% of total medical cost) 3,791.4(52.5%) 3,514.1(51.2%) 4,626.3(53.2%) 0.596* 1.000 0.701 1.000 Annual costs per patient of patients without hospitalizations Number of patients (% of total patients) 1,381(82.8%) 1,436(79.9%) 1,658(79.8%) 0.038 0.030 0.924 0.019 Total medical cost 1,568.0 1,627.7 1,687.3 0.221 0.543 0.940 0.052 OOP payment (% of total medical cost) 534.2(29.3%) 510.8(28.0%) 515.6(26.6%) <0.001* 0.745 0.545 0.036 DM cost (% of total medical cost) 1008.0(69.1%) 1,025.2(69.0%) 985.3(66.0%) 0.001* 1.000 0.002 0.001 * Comparison of percentages of total medical cost The proportion of patients with hospitalizations increased from 17.2% to 20.2% (p = 0.038, adjusted p-value for 2009 vs 2010 = 0.030, 2009 vs 2011 = 0.019) during the study period. For patients needing hospitalization, the economic burden was much heavier. The average annual cost per patient with at least one hospitalization in a year (6,301 USD in 2009) was more than four-fold the costs per patient with only outpatient visits (1,568 USD in 2009). Annual hospitalization cost increased from 3,791 in 2009 to 4,626 USD in 2011, but this increase was not significant (p = 0.596). The hospitalization cost was the major drive for the annual costs, which accounted for more than 50% of annual total cost. Patients needing hospitalization had a higher percentage of OOP payment and DM cost comparing to their counterparts without hospitalizations. In fact, these two percentages decreased from 2009 to 2011 in patients having no hospitalizations (p<0.001 and 0.001, respectively), but not in patients needing hospitalization. Prevalence of diabetic complications and related diseases Based on recorded diagnoses, 60.9% of DM patients had at least one diabetic complication or related disease in 2009 and this percentage increased to 71.2% in 2011 (p<0.001) (Table 3). The percentage of patients with only one complication or related disease decreased from 22.2% to 21.5% from 2009 to 2011. Percentages of patients with two or more complications and related diseases all increased. Hypertension was the related disease of highest prevalence in the included patients, followed by cardio- and cerebral-vascular diseases and hyperlipidemia. 10.1371/journal.pone.0159297.t003Table 3 Prevalence of diabetic complications and related diseases. Number of DM patients (% of total patients)* 2009 2010 2011 Total patients 1,668 1,797 2,076 Patients with complications 1015(60.9%) 1159(64.5%) 1479(71.2%)**     With one complication 371(22.2%) 404(22.5%) 447(21.5%)     Two complications 274(16.4%) 333(18.5%) 442(21.3%)     Three complications 189(11.3%) 222(12.4%) 315(15.2%)     Four or more complications 181(10.9%) 200(11.1%) 275(13.2%) Prevalence of different diseases     Hypertension 733(43.9%) 853(47.5%) 1,082(52.1%)     Cardio- and cerebral vascular diseases 429(25.7%) 539(30.0%) 701(33.7%)     Hyperlipidemia 294(17.6%) 314(17.5%) 422(20.3%)     Neuropathy 275(16.5%) 288(16.0%) 412(19.8%)     Retinopathy 236(14.2%) 297(16.5%) 382(18.4%)     Nephropathy 146(8.8%) 157(8.7%) 242(11.7%)     Acute complications 96(5.8%) 99(5.5%) 122(5.9%)     Fatty liver 82(4.9%) 66(3.7%) 105(5.1%)     Diabetic foot 6(0.4%) 7(0.4%) 31(1.5%)     Hyperuricemia 10(0.6%) 13(0.7%) 22(1.1%)     Troisier-Hanot-Chauffard syndrome - 1(0.1%)) 5(0.2%) * Since patients can have more than one complications, the proportions didn’t add up to 100%; sorted on % in 2011 ** Chi-square test of three year comparison of number of patients with complications: p<0.001 Diabetes related costs of patients with complications and related diseases The number of complications and related diseases was significantly related to the direct medical burden of diabetes treatment, with more diseases causing higher costs (p<0.001 in all groups with one or more complications comparing to the reference group with no complications) (Fig 1). The average annual DM cost didn’t show a significant growing trend in 2010 comparing to 2009 as reference (p = 0.383), but increased significantly from 2009 to 2011 (p = 0.035). There was no interaction effect between number of complications and time. 10.1371/journal.pone.0159297.g001Fig 1 Diabetes related costs for patients with complications and related diseases (USD). Diabetes related medication use and costs Overall medication cost during DM visits per patient increased from 1,335 to 1,383 USD in the studied years (p = 0.021, adjusted p-value for 2009 vs 2011 = 0.039) (Table 4). This increase was mainly due to increased cost for biomedicines (p = 0.012). The vast majority of DM costs was accounted by medication cost (90%), although this percentage slowly decreased overtime (p<0.001, adjusted p-value for 2010 vs 2011 = 0.001, 2009 vs 2011 < 0.001). 10.1371/journal.pone.0159297.t004Table 4 Diabetes related medication use and costs (USD). 2009 2010 2011 p-value of 3-year comparison Multiplicity adjusted p values 09vs10 10vs11 09vs11 Annual overall medication cost during DM visits per patient 1,335.4 1,367.7 1,383.4 0.021 1.000 1.000 0.719     %of DM cost 91.2% 90.1% 88.2% <0.001* 0.147 0.001 <0.001     Annual biomedicine cost per patient 1,144.3 1,188.1 1,185.3 0.012 0.856 1.000 0.732     Annual traditional medicine cost per patient 222.5 215.5 239.9 0.685 1.000 0.367 0.476 Annual DM biomedicine cost 736.5 746.9 698.9 0.284 0.401 0.057 1.000     %of overall medication cost during DM visits 68.7% 68.7% 69.3% 0.750* 1.000 1.000 1.000 Blood glucose lowering biomedicines     Number of users (% of total patients) 1,638(98.2%) 1,763(98.1%) 2,040(98.2%) 0.978 0.838 0.884 0.945     Annual cost per patient (% of overall medication cost during DM visits) 476.8 (47.7%) 481.3 (47.7%) 437.1 (46.5%) 0.295* 1.000 0.642 0.608 Antihypertensive biomedicines     Number of users (% of total patients) 1,164(69.8%) 1,290(71.8%) 1,497(72.0%) 0.268 0.195 0.861 0.130     Number of RAAS inhibitor users (% of total patients) 872(52.3%) 957(53.3%) 1,083(52.1%) 0.754 0.565 0.479 0.922     Annual cost per patient (% of overall medication cost during DM visits) 306.8 (25.4%) 301.5 (24.9%) 285.0 (25.9%) 0.467* 1.000 0.493 1.000 Lipid-modifying biomedicines     Number of users (% of total patients) 527(31.6%) 588(32.7%) 751(36.1%) 0.008 0.478 0.026 0.004     Annual cost per patient (% of overall medication cost during DM visits) 174.5 (12.4%) 182.4 (12.2%) 178.4 (13.1%) 0.318* 1.000 0.559 0.874 * Comparison of percentages DM biomedicine cost didn’t show a significant trend (p = 0.284). The cost of blood glucose lowering biomedicines, including insulins and analogues and all other blood glucose lowering biomedicines, accounted for a major part (47% on average) of overall medication cost during DM visits. Approximately 70% of all patients used antihypertensive medicines, with the costs accounting for 25% of overall medication costs during DM visits. The percentage of hypertensive biomedicine users and the proportion of these medication cost in the overall medication cost during DM visits remained stable from 2009 to 2011. The percentage of patients using lipid-modifying biomedicines was much lower in 2009 (31.6%), but increased to 36.1% in 2011. Discussion To our knowledge, this is the first study analyzing the direct economic burden in diabetic patients using longitudinal data from electronic medical insurance claims database. Our results showed an increasing trend in total medical cost and diabetes related cost for diabetic patients during the study period. The diabetes related economic burden was significantly related to hospitalization and the prevalence of complications and related diseases. The overall medication cost during diabetes related visits also increased from 2009 to 2011. But the use pattern and cost of all three groups of DM biomedicine didn’t show significant changes during the study period. For Chinese patients, studies on economic burden of diabetes from the patient perspective emerged after 2000 [10]. Those studies usually adopted a cross-sectional design and data were collected in the hospitals or recalled by patient interviews [23–28]. Although the expansion of the number of studies in different cities enhanced the crosswise comparison, existing studies didn’t conduct historical comparison usually due to the cost of data collection. In this study, economic burden could be compared over three years in basically the same population (97.8% of patients in the first year could be followed for three years) using electronic health records. Because the UEBMI scheme was highly developed in the sample city (HZ) and covered a stable population, the increase of included patients from 2009 to 2011 indicates a rapid growth of diabetes prevalence in urban China. Considering that our inclusion criteria excluded those patients who had no registered diagnosis or records of hypoglycemic drug use but may have diabetes, there actually may be more diabetic patients in this insured population. The age distribution of studied population was in accordance with the conclusions of diabetes prevalence in the existing literature [22]. The gap of patient numbers between male and female was mainly due to the gender imbalance in the employers working in provincial companies and organizations. Although the prevalence of different diabetic complications and related disease haven’t been systematically investigated nationwide in China, the Chinese guideline of prevention and treatment for diabetes summarized the epidemiological situation of those diseases using evidence from individual studies [7]. Compared to those data, our sample showed a mildly higher prevalence of hypertension (52.1% in 2011), similar prevalence of cardio- and cerebral-vascular disease (33.7% in 2011) and lower prevalence of nephropathy, retinopathy, neuropathy and diabetic foot. The differences may be partly because of our strict diagnosis categorization based on screening of diagnosis name and terms for those diseases with varied pathogeneses and symptoms. Especially for neuropathy, studies found that 60%-90% of diabetic patients may have this complication but 30%-40% of them have no symptoms and thus are undiagnosed [7]. Both those undiagnosed patients and limitation of our diagnosis categorization may underestimate the prevalence. Data from 2013 estimated that the worldwide average annual cost of treating and managing diabetes was 1,437 USD per person [5]. Results in this study showed that the average annual DM cost per patient increased from 1,655 to 1,857 USD. This trend reflected that diabetes caused increasing economic burden in the medium and large-sized cities in China. One study concluded that the average annual growth rate of the direct medical cost of DM was 19.90% during 1993 to 2003 in China, which was much larger than the average annual growth rate of GDP (12.77%) [29]. The average annual growth rate of DM cost in our studied population (5.9% from 2009 to 2011) seemed not to exceed the rapid growth of GDP per capita in the sample city (15.33% from 2009 to 2011 [30]). But when comparing to the disposable income per capita in the sample city (3,937, 4,290 and 4,792 USD in 2009, 2010 and 2011, respectively [30]), the OOP medical cost can still be considered a heavy burden for the diabetic patients (approximately 20% of disposable income). Our results confirm the decisive role of hospitalization in direct medical costs reported before [10, 22, 31]. The results also showed an increase in percentage of patients needing hospitalization. The high rates of hospitalization underline the poor condition of prevention for diabetes in China and contribute to the increase of total medical cost. Patients needing hospitalization usually had more serious conditions (higher percentage of total medical cost for DM cost) and faced higher OOP payment burden. Comparing to the disposable income per capita, OOP cost for patients needing hospitalization accounted for over 40% of the disposable income, a great problem for the individual. Our study showed that DM cost increased with the number of complications. This result underlines the evidence from other studies showing that complication-induced direct medical cost causes a high economic burden in diabetic patients [22, 27]. The longitudinal data showed a significant growth of percentage of patient with at least one complication. The linear relationship was statistically significant between the number of complications and DM cost. Medications were proved to be the main form of treatment in terms of costs for controlling blood glucose and other diabetes complications and related diseases [32]. Our results also showed an increase of overall medication cost occurring during DM visits. But medication cost percentage of total DM cost decreased. This may be explained by the promotion of a comprehensive medicine policy after launch of the Chinese new health reform in 2009, which regulated physicians’ prescription behavior to provide inexpensive generic DM biomedicines to improve the medication use. Meanwhile, cost of diabetes related biomedicines didn’t show a growing trend. Therefore, the increasing of total medical cost, DM cost and overall medication cost may be explained by other reasons. More serious course of disease, more hospitalizations, more drugs and treatments for concomitant diseases may all contribute to cause the increased costs. The International Diabetes Federation Clinical Guidelines Task Force’s global guideline for T2DM calls for the use of a statin and a renin-angiotensin-aldosterone system (RAAS) inhibitor in all persons with T2DM, regardless of lipid and blood pressure levels [33]. Our results showed a low number of these medications users, although the percentage of lipid-modifying medication users increased significantly from 2009 to 2011. China still needs to improve the prevention and treatment pathway for diabetes and thus prevent the negative impacts induced by growth of future hospitalizations, disability, mortality, and medical care costs. Our study used the electronic medical data instead of reported or surveyed data. Because insurance identity cards are mandatorily used when patients see a doctor, we believe these electronic data to be accurate and credible. These data also contained information on medical visits occurring in all levels of health institutions, which was the main limitation in individual hospital based studies. Despite this strength, this study still had limitations. First, the definition of diabetes patients in our study excluded those patients who had no DM visits registered in the electronic database but may have diabetes. In addition, our data didn’t differentiate type 1 and type 2 diabetes, although the prevalence of type 1 diabetes in China was estimated as lowest around the world [34]. Second, our definition of DM cost included all costs during diabetes and complications based on the diagnosis name. Although other diseases irrelevant to diabetes, e.g. trauma, were excluded, we assumed those complications that could in principal have other causes than diabetes was diabetes related. For example, a patient with a hospitalization for nephropathy, which could be caused by high blood pressure alone, was assumed to be caused by diabetes, since all patients in our study were defined as having diabetes. This is an overestimation, but we expected this overestimation to be very small. Third, we believe that the information for the electronic data is trustworthy based on the well-structured information system, but there could be possibilities of mistakes in the raw data level. Furthermore, the procedure of recoding prescriptions and diagnoses may introduce misclassification of medications and diagnoses. Fourth, our data didn’t contain the population other than provincial UEBMI beneficiaries. Compared to the other basic medical insurance scheme in urban areas, urban residents’ basic medical insurance (URBMI), UEBMI has higher reimbursement level for both outpatient and inpatient visits. In addition, URBMI covers unemployed adults and elderly people with poorer economic status. Both higher reimbursement level and better economic status for UEBMI beneficiaries would encourage them to use more health services and spend more money on health. Thus the results may estimate higher economic burden comparing to burden for patients with other insurance scheme. Lastly, the indirect costs in diabetes have important societal influence. However, our electronic data is not suitable to calculate indirect costs. The impact of indirect costs will be an important topic for future research. Conclusion The economic burden of diabetes increased significantly in urban China, with a significant OOP burden for patients needing hospitalization. As diabetes prevalence is expected to grow in the future, economic burden of diabetes will continue to weigh heavily on health budgets. For the relatively new system of universal health coverage in China, the increasing economic burden of diabetes will aggravate the contradiction between limited health resources available and increasing health care demand. Therefore, it is important to improve the prevention and treatment of diabetes to contribute to the sustainability of the Chinese health-care system. Supporting Information S1 Table Categories of diagnoses for data standardization. (DOCX) Click here for additional data file. S2 Table Diabetes related biomedicines included in the data analyses. (DOCX) Click here for additional data file. The assistance with diagnoses coding and examination provided by Dr. M. Xia of Zhongshan Hospital, Shanghai, China, was greatly appreciated. ==== Refs References 1 Zhang P , Zhang X , Brown J , Vistisen D , Sicree R , Shaw J , et al Global healthcare expenditure on diabetes for 2010 and 2030 . Diabetes research and clinical practice . 2010 ;87 (3 ):293 –301 . Epub 2010/02/23. 10.1016/j.diabres.2010.01.026 .20171754 2 American Diabetes Association . Economic costs of diabetes in the U.S. in 2012 . Diabetes care . 2013 ;36 (4 ):1033 –46 . Epub 2013/03/08. 10.2337/dc12-2625 ; PubMed Central PMCID: PMCPmc3609540.23468086 3 Wild S , Roglic G , Green A , Sicree R , King H . Global prevalence of diabetes: estimates for the year 2000 and projections for 2030 . Diabetes care . 2004 ;27 (5 ):1047 –53 . Epub 2004/04/28. .15111519 4 Yang W , Lu J , Weng J , Jia W , Ji L , Xiao J , et al Prevalence of diabetes among men and women in China . The New England journal of medicine . 2010 ;362 (12 ):1090 –101 . Epub 2010/03/26. 10.1056/NEJMoa0908292 .20335585 5 International Diabetes Federation. Diabetes Atlas 6th edition. 2014 Update 2014 [cited 2014]. Available from: http://www.idf.org/sites/default/files/Atlas-poster-2014_EN.pdf. 6 Shu T , Liu H , Goss FR , Yang W , Zhou L , Bates DW , et al EHR adoption across China's tertiary hospitals: a cross-sectional observational study . International journal of medical informatics . 2014 ;83 (2 ):113 –21 . Epub 2013/11/23. 10.1016/j.ijmedinf.2013.08.008 .24262068 7 Chinese Diabetes Society . Chinese guideline of prevention and treatment for Type 2 Diabetes Mellitus . Chin J Diabetes Mellitus . 2014 ;6 (7 ):447 –98 . 8 Ji L , Wu J , Chen Y , Rong H , Hou J , Feng B , et al Analysis of use of hypoglycemic medication of patients with type 2 diabetes mellitus in Chinese middle and big cities . Chin Pharm J . 2012 ;47 (9 ):736 –38 . 9 Zhang D , Tang Z , Li L , Gu L , Xue D . A systematic review on prevalence rate of diabetes in 2001–2010 in China . Shanghai Journal of Preventive Medicine . 2012 ;24 (9 ):492 –95 . 10 Zheng Y , Ji L , Wu J . Cost·of-illness studies of diabetes mellitus in China: A systematic review . Chin J Endocrinol Metab . 2012 ;28 (10 ):821 –25 . 11 Xie X , Vondeling H . Cost-utility analysis of intensive blood glucose control with metformin versus usual care in overweight type 2 diabetes mellitus patients in Beijing, P.R. China . Value in health: the journal of the International Society for Pharmacoeconomics and Outcomes Research . 2008 ;11 Suppl 1 :S23 –32 . Epub 2008/04/17. 10.1111/j.1524-4733.2008.00363.x .18387063 12 Palmer JL , Gibbs M , Scheijbeler HW , Kotchie RW , Nielsen S , White J , et al Cost-effectiveness of switching to biphasic insulin aspart in poorly-controlled type 2 diabetes patients in China . Advances in therapy . 2008 ;25 (8 ):752 –74 . Epub 2008/08/16. 10.1007/s12325-008-0080-4 .18704282 13 Palmer JL , Beaudet A , White J , Plun-Favreau J , Smith-Palmer J . Cost-effectiveness of biphasic insulin aspart versus insulin glargine in patients with type 2 diabetes in China . Advances in therapy . 2010 ;27 (11 ):814 –27 . Epub 2010/11/10. 10.1007/s12325-010-0078-6 .21061114 14 Yang L , Christensen T , Sun F , Chang J . Cost-effectiveness of switching patients with type 2 diabetes from insulin glargine to insulin detemir in Chinese setting: a health economic model based on the PREDICTIVE study . Value in health: the journal of the International Society for Pharmacoeconomics and Outcomes Research . 2012 ;15 (1 Suppl ):S56 –9 . Epub 2012/02/01. 10.1016/j.jval.2011.11.018 .22265068 15 Gao L , Zhao FL , Li SC . Cost-utility analysis of liraglutide versus glimepiride as add-on to metformin in type 2 diabetes patients in China . International journal of technology assessment in health care . 2012 ;28 (4 ):436 –44 . Epub 2012/09/26. 10.1017/s0266462312000608 .23006540 16 Deng J , Gu S , Shao H , Dong H , Zou D , Shi L . Cost-effectiveness analysis of exenatide twice daily (BID) vs insulin glargine once daily (QD) as add-on therapy in Chinese patients with Type 2 diabetes mellitus inadequately controlled by oral therapies . Journal of medical economics . 2015 ;18 (11 ):974 –89 . Epub 2015/07/03. 10.3111/13696998.2015.1067622 .26134916 17 Geng J , Yu H , Mao Y , Zhang P , Chen Y . Cost effectiveness of dipeptidyl peptidase-4 inhibitors for type 2 diabetes . PharmacoEconomics . 2015 ;33 (6 ):581 –97 . Epub 2015/03/05. 10.1007/s40273-015-0266-y .25736235 18 Wei F, Xu Y. HZ was elected as the role model city for exellecnt electronic social security system. hangzhoucom. 2013. 19 Lord JM , Flight IH , Norman RJ . Insulin-sensitising drugs (metformin, troglitazone, rosiglitazone, pioglitazone, D-chiro-inositol) for polycystic ovary syndrome . The Cochrane database of systematic reviews . 2003 ;(3 ):Cd003053. Epub 2003/08/15. 10.1002/14651858.cd003053 .12917943 20 Li Q , Huang X , Gao R , Lu F . Discussion on developing a data management plan and its key factors in clinical study based on electronic data capture system . J Chin Integr Med . 2012 ;10 (8 ):841 –46 . 21 National Bureau of Statistics of China. China statistical yearbook 20112012. 22 Hu H , Sawhney M , Shi L , Duan S , Yu Y , Wu Z , et al A systematic review of the direct economic burden of type 2 diabetes in china . Diabetes therapy: research, treatment and education of diabetes and related disorders . 2015 ;6 (1 ):7 –16 . Epub 2015/02/06. 10.1007/s13300-015-0096-0 ; PubMed Central PMCID: PMCPmc4374079.25652613 23 Wang W , McGreevey WP , Fu C , Zhan S , Luan R , Chen W , et al Type 2 diabetes mellitus in China: a preventable economic burden . The American journal of managed care . 2009 ;15 (9 ):593 –601 . Epub 2009/09/15. .19747024 24 Chen X , Tang L , Chen H , Zhao L , Hu S . Assessing the impact of complications on the costs of Type 2 diabetes in urban China . Chin J Diabetes . 2003 ;11 (4 ):238 –41 . 25 Qiu Y , Ye L , Li X , Lu W . Eeonomie burden of type 2 diabetes mellitus in Shanghai . Chinese Health Resources . 2005 ;8 (2 ):69 –71 . 26 Zhang Z , Chen J , Tang Z , Hu R , Lu B . Study on direct health expenditure of diabetes mellitus in China . Chinese Health Resources . 2007 ;10 (3 ):162 –68 . 27 Wang W , Fu CW , Pan CY , Chen W , Zhan S , Luan R , et al How do type 2 diabetes mellitus-related chronic complications impact direct medical cost in four major cities of urban China? Value in health: the journal of the International Society for Pharmacoeconomics and Outcomes Research . 2009 ;12 (6 ):923 –9 . Epub 2009/10/14. 10.1111/j.1524-4733.2009.00561.x .19824187 28 Yang W , Zhao W , Xiao J , Li R , Zhang P , Kissimova-Skarbek K , et al Medical care and payment for diabetes in China: enormous threat and great opportunity . PloS one . 2012 ;7 (9 ):e39513 Epub 2012/10/11. 10.1371/journal.pone.0039513 ; PubMed Central PMCID: PMCPmc3458850.23049727 29 Liu K , Wang M . Economic burden of main chronic non-infectious diseases in China . Chinese Health Economics . 2005 ;24 (10 ):77 –80 . 30 HZ Bureau of Statistics. HZ Statistical Yearbook 2012 2012 [cited 2012 20 Nov]. Available from: http://www.hzstats.gov.cn/web/ShowNews.aspx?id=iyuczsWwYFQ=. 31 Hu S , Liu G , Xu Z , Li D , Hu Y . Current Status of Epidemic and Economic Burden of Diabetes Mellitus in China . Chinese Health Economics . 2008 ;27 (8 ):5 –8 . 32 He M , Li B , Bian Y . The structure analysis of hospitalization costs on the diabetics of in-patient . Chinese Health Economics . 2009 ;28 (11 ):22 –4 . 33 IDF Clinical Guidelines Task Force. Global guideline for type 2 diabetes. 2012. 34 Karvonen M , Viik-Kajander M , Moltchanova E , Libman I , LaPorte R , Tuomilehto J . Incidence of childhood type 1 diabetes worldwide. Diabetes Mondiale (DiaMond) Project Group . Diabetes care . 2000 ;23 (10 ):1516 –26 . Epub 2000/10/07. .11023146

PMC005xxxxxx/PMC5003381.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 27571423PONE-D-16-0264610.1371/journal.pone.0161630Research ArticlePhysical SciencesPhysicsThermodynamicsEntropyEarth SciencesGeographyHuman GeographyHuman MobilitySocial SciencesHuman GeographyHuman MobilityBiology and Life SciencesBehaviorComputer and Information SciencesSystems ScienceDwell TimePhysical SciencesMathematicsSystems ScienceDwell TimeResearch and Analysis MethodsSimulation and ModelingAgent-Based ModelingComputer and Information SciencesSystems ScienceAgent-Based ModelingPhysical SciencesMathematicsSystems ScienceAgent-Based ModelingEngineering and TechnologyTelecommunicationsResearch and Analysis MethodsSimulation and ModelingEarth SciencesGeographyHuman GeographyUrban GeographySocial SciencesHuman GeographyUrban GeographyA Theoretical Basis for Entropy-Scaling Effects in Human Mobility Patterns Theoretical Entropy Scaling in Human MobilityOsgood Nathaniel D. 12http://orcid.org/0000-0001-8969-670XPaul Tuhin 1*Stanley Kevin G. 1Qian Weicheng 1 1 Dept. of Computer Science, University of Saskatchewan, Saskatoon, SK, Canada 2 Dept. of Community Health and Epidemiology, University of Saskatchewan, Saskatoon, SK, Canada Sendiña-Nadal Irene Editor Universidad Rey Juan Carlos, SPAIN Competing Interests: The authors have declared that no competing interests exist. Conceived and designed the experiments: NDO TP KGS WQ. Performed the experiments: TP WQ. Analyzed the data: NDO TP KGS WQ. Contributed reagents/materials/analysis tools: NDO TP KGS WQ. Wrote the paper: NDO TP KGS WQ. Conceived the initial derivation: NDO. * E-mail: tuhin.paul@usask.ca2016 29 8 2016 11 8 e016163020 1 2016 9 8 2016 © 2016 Osgood et al2016Osgood et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Characterizing how people move through space has been an important component of many disciplines. With the advent of automated data collection through GPS and other location sensing systems, researchers have the opportunity to examine human mobility at spatio-temporal resolution heretofore impossible. However, the copious and complex data collected through these logging systems can be difficult for humans to fully exploit, leading many researchers to propose novel metrics for encapsulating movement patterns in succinct and useful ways. A particularly salient proposed metric is the mobility entropy rate of the string representing the sequence of locations visited by an individual. However, mobility entropy rate is not scale invariant: entropy rate calculations based on measurements of the same trajectory at varying spatial or temporal granularity do not yield the same value, limiting the utility of mobility entropy rate as a metric by confounding inter-experimental comparisons. In this paper, we derive a scaling relationship for mobility entropy rate of non-repeating straight line paths from the definition of Lempel-Ziv compression. We show that the resulting formulation predicts the scaling behavior of simulated mobility traces, and provides an upper bound on mobility entropy rate under certain assumptions. We further show that this formulation has a maximum value for a particular sampling rate, implying that optimal sampling rates for particular movement patterns exist. http://dx.doi.org/10.13039/501100000038Natural Sciences and Engineering Research Council of Canada356043-2008Stanley Kevin G. http://dx.doi.org/10.13039/501100000038Natural Sciences and Engineering Research Council of Canada327290-06Osgood Nathaniel D. KGS received funding (Grant Number: 356043-2008) from "Natural Sciences and Engineering Research Council of Canada". Funder’s website: http://www.nserc-crsng.gc.ca/. NDO received funding (Grant Number: 327290-06) from "Natural Sciences and Engineering Research Council of Canada". Funder’s website: http://www.nserc-crsng.gc.ca/. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are provided as Supporting Information file (S1 Data).Data Availability All relevant data are provided as Supporting Information file (S1 Data). ==== Body Introduction The importance of understanding how humans move through, consume and interact with the space they inhabit is a central tenet of geography, urban planning, architecture, and many other social sciences. Being able to concisely represent the quality of human movement through space allows practitioners in these disciplines to design better cities, buildings, and policies. Traditionally, human motion was studied using the pen-and-paper tools of the anthropologist, including retrospective surveys, direct observation, ethnography, or self-report through interviews or diaries. While these techniques have provided remarkable insight into human mobility, particularly into its cognitive aspects, they are limited in spatio-temporal resolution, and are prone to observer or reporter bias, and can be time consuming. Technological advances in localization have opened new opportunities for analyzing human mobility [1] [2]. Electronically mediated population tracking is a practical alternative to traditional pen and paper techniques. Inexpensive loggers or smartphone apps can use the Global Positioning System (GPS) to record trajectories through space [3] [4] [5]. While GPS-based systems provide exceptional positioning quality and coverage when outdoors, they can be unreliable in institutional buildings or in terrain where sky views are blocked. GPS-based data acquisition can also be more cumbersome as participants have to be recruited, potentially outfitted with appropriate equipment and debriefed. An alternate approach is to mine cell tower or WiFi router contact traces through time to generate trajectories by representing the locations of the device and, therefore, the person, as the locations of the towers or routers to which the device is connected (e.g., [6]). In proximity-based representations, space is implicitly represented as a sequence of polygons, derived from the Voronoi diagram of the beacons. While these representations can be easier to obtain, as cell or router contact records are often maintained by telecommunication companies or institutions, they are also often characterized by a heterogeneous spatial decomposition (based on the Voronoi diagram structure) and intermittent sampling, as records are often only generated for active connections (calls, texts, or data transmission). These technologically mediated localization systems provide much higher spatial and temporal fidelity than traditional methods, are less prone to bias, but are divorced from the cognitive processes underlying the decision making. The additional spatio-temporal resolution can be a double edged sword, as traditional statistical analysis techniques suitable for analyzing survey responses are no longer sufficient for characterizing such data. To address the overabundance and complexity of the data, researchers have looked at visualization methods or statistical metrics to represent the important components of the data more concisely. Binned or aggregate statistical representations are popular. Heatmaps, visualizations of the two dimensional frequencies of parameters of interest, are a standard method of aggregating location over time and space (e.g., [7, 8]). Space is typically binned at a specific resolution, then location data is accumulated for each bin. Aggregate distributions of secondary measures can also be useful to summarize high fidelity data. Aggregate measures such as visit frequency, trip duration, trip length, and radius of gyration have been previously reported in the literature [9–12]. In all of these representations, spatio-temporal variation is marginalized over some variable, destroying important information about the structure of the variability. However, several researchers have observed simple and reproducible patterns and a high degree of spatial and temporal regularity in visited locations of humans [13–16]. In their seminal paper, Song et al. [9] proposed the entropy rate of a mobility pattern as a metric of variability or predictability in human behaviour. By discretizing the world, and providing a label to each discretized location, a trajectory through space could be converted into a string of location labels or symbols. As a string, this representation could be summarized by the entropy rate, which is closely related to the compressibility of the string. People with a great deal of regularity in their schedules would be represented by a lower entropy rate than people whose spatio-temporal habits were less predictable. This metric had the advantage of providing a measure of the regularity of spatio-temporal habits of a population as a single number. Song et al.’s original work has been extended to other aspects of human behavior, including social contact and activity in both complete and moving average implementations [17] [18] [19]. According to Shannon’s original definition, entropy is calculated directly from a random variable or distribution [20] [21]. Entropy could be calculated for aggregated distributions such as trip length or dwell time, but that representation does not capture the empirical entropy rate for the trajectory string. To approximate entropy rate empirically, lossless compression algorithms are generally employed [22]. In particular, the Lempel-Ziv 78 (LZ) algorithm has been shown to provide asymptotic estimates for the entropy rate of a string as the length of the string goes to infinity [9] [22] [23]. Following the example established in Song et al.’s original paper, researchers estimate the entropy rate of a mobility string through LZ compression, although shortcomings with this approach have been noted [24]. Employing the methodology originally proposed by Song et al., it is possible to use LZ compression to approximate the entropy rate of a person’s trajectory. However, the entropy rate calculated for this path is not universal, as it depends on the spatial and temporal resolution with which the path is sampled. That is, the resolution of binning and the regularity and rate of sampling impact the entropy rate calculated from the LZ compression technique [24] [25] [26]. Meaningful comparisons of entropy rates between different people or populations can only occur if those rates were calculated from strings with identical spatial and temporal resolution. This implies that meaningful comparison of mobility entropy across experiments is not possible in general, as the experimental protocol changes. It further implies that comparing different individuals in the same dataset could be problematic if there is heterogeneity in the geographic bin size or sampling rate; for example, in a study comparing the mobility of rural and urban populations through cell phone records, where the rural Voronoi cells were systemically and significantly larger than their urban counterparts. Because mobility entropy rate is a useful metric, some researchers have studied or proposed empirical methods of describing variations in spatio-temporal scale [24] [25] [26]. However, empirical models can be difficult to generalize, as specific models may be tightly tied to the datasets from which they were derived. In this paper, we provide a theoretical derivation of a scaling law for mobility entropy rate calculated through Lempel-Ziv compression. This derivation is theoretically valid for non-overlapping trajectories which can be represented as a series of line segments navigated at constant velocity over a regular four-connected grid. This scaling model shows excellent agreement with simulated trajectories, even when those trajectories violate assumptions underlying the derivation. Analysis of the mathematical properties of the model yields several key findings. First, variation with spatio-temporal scale is an inevitable consequence of the LZ approximation. Second, mobility entropy rate at any spatio-temporal scale can be represented by four parameters: the length of the trajectory, the velocity of each segment and the spatial and temporal scales. Third, the model has a unique maxima with respect to the temporal sampling rate, implying that there is a natural sampling rate for a given trajectory which maximally captures the information it encodes. Finally, the performance of this model indicates it might be possible to express mobility entropy rates measured with different experimental configurations at common resolutions, allowing comparison between disparate populations and experiments, allowing mobility entropy rate to be employed to its full potential as a metric. Analysis Problem Structure Our derivation relies upon the performance of Lempel-Ziv (LZ) compression in approximating the entropy rate of mobility, the most common method for estimating entropy rate based on the seminal work of Song et al. [9]. As many other researchers have noted [22, 23], this approximation makes strong assumptions about the behavior of the string, notably that it represents a stationary ergodic process, and is sufficient long for the algorithm to converge. While these assumptions may be violated in practice, the approximation is widely used in the literature. Examining the extent to which this approximation scales will provide valuable insight into the interpretation of existing and future results using this approximation, independent of whether the underlying assumptions are correct. We constrain our derivation to the behavior of the LZ approximation for patterns of movement only, and do not explicitly consider parameters such as location dwell time. That is, our analysis is most suited to datasets concerned with trips or trajectories, and will not necessarily apply to datasets which capture prolonged periods of rest. The derivation problem then becomes examining how LZ compression functions for a set of paths. The most fundamental assumption required for this examination is the definition of a path. We define a human mobility path as a series of piecewise linear two dimensional segments, navigated at a constant velocity. We assume that these paths are executed over a discretized space, as is common in the literature. For convenience, authors of [9, 10, 26] have used non-uniform Voronoi decompositions of the space, as these decompositions flow naturally from the cell tower or WiFi router locations. However, these datasets are characterized by irregular boundaries and variable cell sizes, greatly complicating mathematical derivation of scaling properties. Instead, for tractability, we have chosen a regular grid approximation, which is more appropriately used when discretizing higher fidelity tracked datasets obtained through GPS trackers or smartphone locations [4, 5, 15, 27]. Finally, we assume that paths are sampled regularly in time, again consistent with GPS tracking, rather than the stochastic data arrival associated with cellular call records. Because we assume that we are starting with a high-fidelity source like GPS traces, interpolation of locations between timesteps is not required. As an agent traverses the discretized space, their locationing system will emit symbols (represented as letters in examples for convenience) corresponding to the label of the grid cells at their measured locations, creating a single dimensional string representing their trajectory through the two dimensional space. Because we assume a piecewise linear path through regular grids, sampled at regular intervals, we can begin to analyze how traversing these grids would appear. For a path parallel to either axis of the grid, the agent will emit a sequence of symbols characterized by repetition of the current grid cell. For constant velocity paths through multiple grid cells, this will lead to a uniform repetition of symbols, based on agent speed and cell size (e.g., ‘AAAABBBBCCCCDDDD’ for one speed and ‘AABBCCDDEEFFGGHH’ for an agent traveling twice as fast). However, if the path is not parallel to the grid cells’ axes, then the agent may clip edges of cell (e.g. ‘AAAAABCCCCC’) changing the string and the entropy rate. As defining all possible arbitrary paths through cells is not mathematically tractable, we assume that agent must traverse the entire cell. This is the strongest assumption that we make, and the most likely to fail when applied to empirical data. This assumptions has the additional impact of forcing paths to be bin-sized aligned; individual line segments must have a length that is an integer multiple of the bin size. Finally, we assume that each line segment traces a unique path through space, and crosses no other segment. While on the surface this seems like a limiting assumption, made to facilitate derivation, we mean to eliminate strongly repeating trajectories, like orbits, which would significantly depress the entropy rate as calculated from the LZ approximation. We expect that crossing but non-overlapping paths, as reported in works such as [11], would have entropy rate approximations close to the unique path case, because while individual symbols might repeat, we would not expect to observe the repetition blocks of multiple symbols. We limit the analysis to a sampling regime that will return sensible answers. Specifically, we consider regimes for bin width (resolution) and sample period in which scaling is meaningful. Our assumptions can be summarized as: Path: we assume that the path can be sufficiently well approximated as a series of line segments. Velocity: we assume a non-zero constant velocity v for each line segment dvi/dt = 0. Accuracy: we assume that a given location measurement offers perfect accuracy, but relax this assumption in additional analysis. Measurement Density: we assume that measurements are made with sufficiently high resolution devices so as to support a spatial decomposition into square bins of characteristic length W and a regular temporal sampling of period T, with no need for interpolation. Connectedness: we assume that agents traverse the square bin or block in a classic four-connected manner, that is that participants only move in the cardinal directions though a block and traverse the entirety of the block, implying that the time to traverse a block is always W/v. Scale: we consider a mesoscopic sampling regime with the following characteristics: Spatial: the bin size is no bigger than the extent of the smallest line segment in the path. Temporal: no cells crossed by the path are skipped due to undersampling: T ≤ W/v. Independence: we assume that each segment traces a unique and independent path from all previous segments. This assumption is necessary for tractability, but eliminates repetition (and, therefore, reductions in entropy rate) at an inter-path segment level. Repetition would decrease entropy rate, so we expect that this assumption pushes our derivation towards an upper bound. Termination: we assume that each sequence of location symbols terminates with a unique symbol. In the subsequent sections, we derive scaling behavior from the process of Lempel-Ziv compression, under the above assumptions. For readability, derivations are summarized in the main text. For detailed step-by-step derivations, please refer to S1 Appendix. Single Segment Derivation We begin by considering a single line segment of length x traversed at constant velocity v parallel to one grid axis, then extend this to multiple non-overlapping line segments. The path requires t=xv time to traverse. Given our assumptions, the traversal of each grid cell will require at least one sampling period T and possibly more, resulting in one of more instances of each cell-symbol being emitted as the agent crosses the cell. Because the agent traverses each cell in its entirety, and in a four-connected manner, it takes the same amount of time to cross each cell. The results is a series of repeated symbols representing each of the cells that the segment passes through, where the number of repeats per cell is given by Lb=WvT and the total length of the string is L=xvT. From [9, 22, 24], the LZ-derived entropy rate of a string S of length L is given by (1L∑i=0L-1Λi)-1lnL(1) as L → ∞, where i is the index of a character in the string (with the first character being at i = 0), and Λi is the length of the minimum substring beginning at i such that this substring has not previously been observed in the prefix of S terminating at position i, and L is the length of the string. When scaling the spatial and temporal resolution for simplicity, we consider inter-sample periods given by T = T02−m(m ≥ 0), and bin sizes as W = W02n(n ≥ 0), where W0 and T0 are governed by our assumptions bounding the bin size and sampling rate. The values T0 and W0 are not necessarily fixed constants, but instead vary with the parameters and the choice of v, x (for W0) and T. Practically, there are bounds for each, given the method of localization employed, but in our formulation, W and T are parameters to some degree controlled by the experimenter, while x and v are properties of the observed agents. Structure of the Sampled Sequence Both the temporal inter-sampling rate T and the spatial scale W affect the structure of the sampled sequence. The sequence has a total length of L=xvT symbols, but is composed of xW blocks each consisting of Lb=WvT uniform repeating symbols. The number of symbols per block is an interaction between W, T, and v. Larger blocks take longer to traverse, leading to more repeated symbols. For W = x, the sampled string consists of a single, homogeneous, block of L symbols. For our lower bound of W = vT, this sampled sequence of length L consists of LLb blocks, each consisting of a single unique symbol. Because we assume non-overlapping paths, the binned values associated with different blocks are distinct. Because the sampled values within a given block are homogeneous, and because the sample value within the block is unique, the values of Λi all follow a regular pattern, which depends only on the index within the block, and not on the index within the sampled string as a whole. That is, we will have LLb unique symbols and blocks, with each symbol repeating Lb times within its block. Thus, Λi=Λimod2n, given the structure of our downsampling. We can thus decompose 1L∑i=0L-1Λi=1L∑b=1xvT2n∑j=02n-1Λj(2) The terms in the outer sum (over b) correspond to the number of blocks, which is also the number of unique symbols xLb. The index terms in the inner sum (over j) correspond to the number of repetitions in a block of length Lb = 2n. To derive this sum, we consider two distinct cases: the positions in the first half of the block, and those in the latter half of the block. The pattern for the Λj in the first half of the block is a simple rising sequence. Regardless of the block, the first sample in the block (i.e., j = 0) is a unique character not previously seen in the string, and thus ∀j = 0Λj = 1. Similarly, for all blocks of length of at least 2, the second sample in the block concatenated with its following symbol (in this or the next block) has not previously been seen in the string, and thus ∀j = 1Λj = 2. Using similar reasoning, the lambda values continue to rise within the block up to the index of j=2n2. Thus ∀j≤2n2Λj=j+1. That is, for indices up to the halfway point through the string, the substring starting at that point and including j additional subsequent characters (and thus of length j + 1) consists purely of repetitions of the same character associated with this block, of successively larger lengths, and has not previously been seen. We consider now the cases of the Λj in the second half of the block, noting the assumption above of a unique terminating symbol following characters in the final block. For characters at indices just beyond the midpoint of their block (i.e., j=2n2=2n-1), there is a minimum unique string consisting of the character at that point, 2n2-1=2n-1-1 additional identical characters beyond that point lying within the same block, and then (additionally) the first character of the next block, thus yielding a unique total string length starting at position j of 2n−1 + 1 = j + 1, as given by the formula above. For the indices in the following 2n−1 − 1 positions of the string (i.e., for 2n−1 < j ≤ 2n − 1), we are dealing with a strictly decreasing integer sequence, terminating in 2. This reflects the fact that for index j, the uniform symbol prefixes beginning at index point j have all previously been seen within this block, and the smallest unique string consists of the prefix beginning at the current point (index j), proceeding through the end of the block, and including one character beyond the end of that block (which has not yet been previously encountered within the string). For a character at position j (zero-based) within the block, this yields a string length of (2n − j) + 1. Thus, we have ∀j > 2n−1 Λj = (2n − j) + 1. To summarize, Λj will be an arithmetic sequence, starting at 1, until just beyond the midpoint is reached; and then decreasing until the final value of 2 (e.g., 1,2,…,Lb2,Lb2+1,Lb2,Lb2-1,…,2). Given this per-block total, and that there are xvT2n blocks, we have: ∑j=02n-1Λj=∑j=02n2(j+1)+∑k=02n2-1(k+1)=22n4+2n Having a closed form expression for Λj and the equivalence in Eq 2, we can now derive an expression for Λi. 1L∑i=0L-1Λi=vTxxvT2n22n4+2n=2n-2+1 Substituting Λi into the equation for LZ compression-based entropy rate Eq (1), the estimated entropy rate of the string is: H(W,T)=2n-2+1-1lnxvT=lnxvT(2n-2+1) Because the number of symbols is related to the width of the cell and sampling rate, and as we have assumed the minimum width W0 = vT to ensure at least one sample per cell H(W,T)=4W0lnxvTW+4W0 and, therefore, H(W,T)=4lnxvTWvT+4(3) Where x and v are independent properties of the path in question, and W and T are parameters that are intrinsic to the methods and apparatus of a particular experiment. That a scaling law exists containing only four terms, two controlled by the experimenter, and two determined by the path, is one of the key findings of this work. While choice of units will affect the size of the x, v, T and W0 terms, we note that the governing terms xvT and WvT are distinguished by being of unit dimension; thus the entropy rate expression is also of unit dimension, and invariant to unit change. The first of these expressions is the total length of the sampled string; the latter is the number of samples required to cross a bin. This result suggests that for a single line segment, the entropy rate of strings sampled at different resolutions according to bin widths W and temporal inter-sample spacing of T should scale proportional to O(4lnxvTWvT+4). Somewhat counter-intuitively, the entropy rate for a sequence of non-overlapping line segments of total length x, which are traversed in four-connected manner, is identical to the single line segment derivation above. Consider two cases: a single line segment of length x, and a snaking series of line segments also collectively of length x, which are selected in four-connected manner, but randomly picking a non-overlapping direction at every bin. The single segment linear path induces a string containing LLb unique symbols, each repeating Lb times, as described above, and is, therefore, described by Eq (3). The snaking path induces a string with exactly the same structure. Each transit of a bin produces Lb symbols. At the end of each bin transit, a new batch of Lb symbols begins, starting with a never before seen character. At the end of the path, in accordance with our assumptions, a unique symbol is emitted. This applies to any mixture of line segment lengths traversed at constant velocity, as long as they are multiples of W, and do not overlap. Any set of paths that generate a repeating structure like the structure for a single line segment will exhibit entropy scaling behavior described by Eq (3). Intuitively, the straight line trajectory should have a lower entropy rate than the snaking trajectory because the trajectory can be described by a simple mathematical function. However, the entropy rate of the sequence is evaluated independently of the rule used to generate it. This apparent incongruence between the apparent and actual entropy rates for trajectories is subtle, and outside the scope of this work. However, a further investigation into the role of context into human mobility entropy rate estimation, along the lines of [24], appears warranted. This formulation extends to any number of dimensions as long as the decomposition of that space is a hypercube, and transiting of the hypercube happens hyperface to hyperface along equidistant paths across the hypercube, which is essentially the higher-dimensional generalization of the four-connected path we have assumed. Because the compression—and, therefore, the entropy rate calculation—happens only on the trajectory, which is a single dimensional manifold, as long as the structure of the symbols generated by the trajectory remains the same, the above analysis will hold, and the scaling law will apply. In the case of higher dimensional spaces, W is the single dimensional edge length of the hypercube, and v is the velocity through the hypercubes. Because opposite faces of a hypercube will be W distance apart, by definition, the straight line trajectory through a hyperspace will have the same symbol structure, and, therefore, the same entropy rate scaling behavior as above. Because there also must exist a path of distance W between adjacent faces of the hyperplane, the non-overlapping path argument above also applies. Therefore, Eq (3) holds, in general for spaces of arbitrary dimension, decomposed as hypercubes, for non-overlapping paths. The scaling law exhibits some degree of upper-boundedness against some, but not all, of the assumptions. In particular, paths characterized by repetition will decrease the overall entropy rate by introducing inter-block repetition, that LZ will detect and compress. Violations of the scale assumptions will also decrease entropy, as bin sizes larger than the smallest line segment will cause line segment concatenation with a cell, and, therefore, longer repeating blocks. Similarly, skipping cells due to undersampling will not increase the entropy, as a maximal condition of each symbol in the string being new and unique will already have been reached. However, the addition of noise can disrupt the sequences described here, potentially increasing entropy rate, as expected for additive noise processes. Allowing non-four-connected paths could also increase the entropy in some cases, particularly as cell size increases and clipping becomes more likely, although whether the entropy rate increases or decreases is dependant on the interaction of path and spatial discretization. Scaling Law Behavior When proposing scaling laws, it is often useful to examine their limiting behavior. The proposed law is well behaved in the limits for the experimenter controlled parameters. As T tends to zero, while the length of the string rises, each bin will also be sampled by an ever larger number of repetitions and the entropy rate goes to zero. By contrast, the limit of H(W, T) as T → ∞ is negative infinity. However, this bound does not make sense semantically, because it represents the entropy rate of mobility patterns which are never sampled, which violates our assumption about sampling. As W approaches zero, entropy rate tends towards a maximum value lnxvT, which represents the log of the number of symbols sampled, or the entropy rate of a series of distinct symbols of the given length. As W → ∞, entropy rate approaches zero, which is sensible, as the entire string would consist of a repetition of the same location symbol. The proposed law is also well behaved in the path description parameters. As v → 0, H(W, T) also goes to zero, as we have a path composed of a single repeating symbol. As v → ∞, (putting aside relativistic effects), the entropy rate goes to negative infinity, which, as in the case of T, corresponds to a path that is never sampled, and violates our assumptions about sampling. At a minimum, L must be at least one, or there is no string, and LZ will return the compression of a single symbol, likely a poor approximation of the entropy rate. As the string becomes infinitely long, with infinite no. of distinct blocks, the entropy rate approaches infinity, as would be appropriate. A natural question is whether the scaling law has any maxima or minima with respect to W or T, as this would imply sampling regimes which might be considered optimal. This behavior can be investigated using the partial derivatives. The partial derivative of H(W, T) with respect to W is ∂H∂W=-4vtln(xvT)(WvT+4)2(4) The derivative does not have a root with respect to W, so there are no minima or maxima along the W axis for the scaling relationship, implying that no sampling dimension is preferred. Examining the partial derivative of the entropy rate scaling with respect to T yields ∂H∂T=4vW+16vT-4vWln(xvT)(4Tv+w)2(5) which has a sequence of roots for a given (v, W, x) at T=W4vW(4xeW)(6) Where e is the natural basis and W is the Lambert W function, which is not solvable analytically, but is readily approximated numerically. This function is defined for W > 0 and v > 0, which is strictly true in our formulation, as W is a distance, and v is a ratio of distance and time. This implies that for certain values of (x, v, W), there exists a sampling rate corresponding to maximum entropy rate. Sampling beyond this rate will lead to repetition, decreasing the entropy rate. Sampling below this rate will result in removing information, also lowering the entropy rate. This finding is a central outcome of the scaling law, as it implies that there exists an optimal temporal sampling regime for a given spatial resolution and mobility pattern. Entropy Rate of Paths with Mixtures of Velocities While the previous section derived the scaling behavior of the entropy rate of a non-overlapping piecewise linear path, this analysis is unnecessarily limiting for practical application. We seek here to derive an entropy rate for a sequence of non-overlapping line segments traversed with varying velocity. Considering non-overlapping paths as before, Eq (3) provides a starting point to examine how entropy rate might sum for non-overlapping paths of straight line segments through space. We begin by noting that changes in speed undertaken between two samples occuring within the same spatial bin are not observable, being below the spatial sampling rate. The number of symbols emitted when transiting the cell is proportional to the time it takes to cross the cell, divided by the sampling rate. The time taken to cross the cell can be trivially represented as the width of the cell divided by the average speed within the cell, from the definition of average speed (vc¯=WT). Given that speed changes within a cell are averaged by the emission of symbols, we need only concern the derivation with inter-cell velocity variability. Given the same linear four-connected path, covering a distance x, consider the case where a fraction α is made at velocity βv, and fraction (1 − α) is made at velocity γv, yielding a time-averaged velocity of v¯=vαβ+(1-α)γ. The string length is L′=αxβvT+(1-α)xγvT=xv¯T(7) The total entropy rate is then (step-by-step derivation is provided in S1 Appendix): 1L′∑i=0L′-1Λi-1lnL′=1L′∑b=1αxβvT2n22n4+2n+∑b=1(1-α)xγvT2n22n4+2n-1lnL′ and therefore H(W,T)=4lnxv¯TWv¯T+4(8) which is the same expression as in Eq (3), but including time averaged rather than constant velocity. This derivation is generally valid, subject to bounds on the velocity which maintain that at least one symbol per cell must be recorded, and no cells can be skipped by changing velocity. Impact of Spatial Uncertainty As most entropy rate calculations of interest will be performed on empirical data, it is important to consider the impact of measurement noise on scaling behavior. If measurement noise dominates, then the scaling behavior described here is of limited utility. However, if the measurement noise has well-behaved statistical properties, it may be possible to derive an expected entropy rate considering these impacts. We seek here to consider the effects of spatial noise on the entropy rate estimates, as we expect timing estimates to be much finer grained than human motion. We assume a GPS-like positioning system, with position error estimates that are normally distributed around the true value μ with standard deviation σ, employing the classic zero mean Gaussian noise model. The probability that a given measurement (a sample from that distribution) lies further than distance d from the mean is given by 1-erf(dσ2). Now consider taking a measurement at the center point of a generic square bin of physical width W. The probability, p, of a measurement lying outside the distance to the boundary (W2)—and, thus, returning an erroneous spatial bin, and associated symbol—is given by Eq (9), where draws from this distribution are considered independent. p=1-erfW22σ(9) By incorporating the above noise model, and applying a number of further assumptions, the entropy rate can be approximated as (step by step derivation is provided in S1 Appendix): H(W,T)=lnxvT1p+1pLb1+2(1-p)(1-p)Lb2-1p-(1-p)Lb2(10) Recall that L=xvT and Lb=WvT, where the total path length is x, physical bin width is W, the velocity is v, and inter-sampling period is T. We can further expand Eq (10) by substituting WvT for Lb, and Eq (9) for p. If the agent travels distance x with a mixture of velocities, v in Eq (10) gets substituted by the time-averaged velocity v¯. Erroneous symbols generated through noise processes may come from a bin traversed earlier in the trajectory, a bin that will be traversed later in the trajectory, or from a bin that will not be encountered by the trajectory. While the occurrence of an erroneous reading in either of the first two categories will yield repetitions (thus, preventing the relevant substrings from being entirely unique), an occurrence of the latter will not. Specifically, we believe that it is considerably more likely that the formula in Eq (10) will underestimate the entropy rate in practice, as large enough noise to be effective will disrupt the repetition of symbols, and, therefore, increase entropy rate. However, it is possible to imagine pathological behavior where noise would, for the entire duration it takes to traverse a bin width W at v, perturb the measurement in the direction of the next bin on the trajectory, returning a double length sequence of symbols and thus decreasing the entropy rate. However, for a symmetric error distribution like a Gaussian, we anticipate that this behavior should be rare. Fig 1 compares the entropy rate measures with (generally top) and without (generally bottom) noise for 5 < = W < = 200, 0.5 < = T < = 10, v¯=1, and x = 1000. Absent noise, the entropy rate is generally lower over wide ranges of medium and large spatial scales and sampling periods when compared with the estimate of entropy rate with noise. However, at small physical scales and longer sampling periods, the entropy rate absent noise can lead to sequences of entirely unique symbols, whereas there is some repetition in the presence of noise—and, therefore, somewhat lower entropy rate. Assuming a standard deviation of 30m for GPS, these two entropy rate estimates exhibit a high degree of disparity, particularly for physical scales of around 40−80m. By contrast, the entropy rate estimates with and without noise approach each other asymptotically as the spatial aggregation scale increases, as expected. 10.1371/journal.pone.0161630.g001Fig 1 Entropy Rate measures with (generally top) and without noise (generally bottom). Methods To provide a semi-empirical validation for the model, we compared the results of the theoretical model with the results from two widely employed and stylized simulated models of human mobility. A single agent traversed a simulated field with a constant speed (v) while following the employed motion models, and agent locations on the grid were recorded according to the spatial and temporal sampling rates. The maximum and minimum sampling periods were set to 512s and 1s, respectively. We collected 64 samples for max(T) = T0 = 512 s; therefore, making the number of samples 64 × 2m for T = T02−m. To collect 64 samples at T0 = 512s, the agent in the theoretical model had to traverse 64vT0 = 65536m. For other models where the agent moved in a square field, we set the diagonal length of the field to 64vT0 to make their comparison with the theoretical model sensible. The minimum value of W for a combination of v and T is vT, and the maximum value of W is 64vT0. Each models was applied with and without power law distributed dwelling at nodes, and (for each such variant) with and without additive noise. The two empirical motion models are: Random Waypoint Motion Model: in this model, 100 unique waypoints were drawn uniformly from the field described above. The waypoints described a fully connected graph; that is, the agent could go from a waypoint to any other waypoint. This allows crossing paths, which we assumed absent in the theoretical derivation for simplicity. Transitions from one waypoint to another were drawn uniformly. However, because waypoints were drawn uniformly, the probability of repeated path sequences is low. We investigated transitions with and without dwell time. For transitions with dwell time, dwell time was drawn from a power law distribution with the exponent of −1.8 and maximum dwell time was set to 17 hours, consistent with [10]. Power Law-based Motion Model: in this model, the agent selected an angular direction uniformly from a set {5k°:k∈N+and5≤5K≤355}, and drew the distance for the next step from a power law distribution, which is typically observed in empirical datasets (e.g., [10]). Draws were constrained to ensure that the agent remained in the field. The distance was limited to 0.8 times the characteristic length of the field. Movement directions were resampled until a destination inside the field was generated. In these experiments, −1.55 was chosen as the power law exponent, consistent with reported empirical findings [10]. For the dwell time variant, we employed the same distribution as for the Random Waypoint model. We also considered an additive measurement noise model. Each of the above scenarios was run once without any additive noise and once for the noise model. Simple zero mean Gaussian additive measrement noise model was considered, consistent with simple noise models of GPS location measurements. Noise was added to the signal after the agent moved but before simulated measurement took place. A moderate (σ = 10m) noise level was selected consistent with commodity GPS systems. A theoretical entropy rate was calculated from Eq (3), and compared to the empirical measurement calculated according to Eq (1). Several aspects of these simulated motion models depart from the assumptions made when deriving our scaling law. First, each model permits crossing paths, leading to repeated symbols, although are unlikely to produce cyclic paths. Second, we have included variants which include measurement noise and dwelling, neither of which are explicitly accounted for in Eq (3). Third, the models can lead to clipping effects explicitly ruled out when deriving Eq 3. Given that the paths were generated in simulation, we have precise control over the sampling rates, bin widths, path length and agent velocity and can, therefore, explicitly calculate the scaling law, and compare them against the Lempel-Ziv derived entropy rates from the trajectory records. Employing bin widths of W=W02n=vT2n, we can simplify Eq (3) into Eq (11). H(W,T)=4ln(L)2n+4(11) We use the coefficient of determination (R2 metric) to understand how well the theoretical curves fit with those from the empirical simulation models, including the model that applies Eq (1) to the sequences of the theoretical model. The definition of R2 is given in Eq (12), where f1, f2, …, fn are the predicted values for y1, y2, …, yn. R2 values were calculated in R software environment. R2=1-∑inyi-fi2∑inyi-1n∑inyi2(12) We ran the simulations on a Linux-based computing cluster with 96 computational nodes, each having 2 x eight-core Intel E5-2650L (1.8GHz) or Intel E5-2640L (2.0 GHz) Xeon Processors, and 32GB RAM. Jobs were submitted to the cluster through the Torque scheduler. Refer to S1 Data for the relevant data and code required to generate the data. Results We seek to determine how well the scaling law behaves when compared to paths absent non-Gaussian measurement noise, participant non-compliance and other effects that may be present in empirical data, which might obfuscate the underlying behavior and make comparisons more difficult. Some of the simulated systems here are noise free, but do allow for repeating symbols and cell clipping. Analyzing the behavior of these simulated systems against the theoretical scaling model could provide insight into the impact of breaking these key assumptions on the proposed scaling law’s predictions. Fig 2 presents the comparison between the theoretical model and power law-based models with and without dwelling, and with no added measurement noise in the sequences. In the model without dwelling, the scaling law provides exceptional agreement with the simulation. At very large W, the empirical entropy rate exceed the theoretical, as clipping effects begin to dominate. As the bin width increases, more repetitions occur in the string. Therefore, entropy rate goes down. The theoretical model considers regular patterns of string. However, because of stochastic nature of empirical strings, the effect of large bin width may be less dominant in lowering the entropy rate than is the case for the theoretical model. This is why the entropy rate of the empirical models in Fig 2 for large W exceeds that of the theoretical model. As an example, consider two 64-character strings from the alphabet {‘0’, ‘1’}, which are expressed, using regular expression, as /0{32}1{32}/ and /1{3}0{31}1{30}/. Here, the second string has a higher entropy rate. The first string has the structure assumed by the theoretical model, while the second indicates a clipped trajectory. The latter may appear as the representation of a trip, at a large bin width, which is derived from power law-based trip segment lengths and dwell times. 10.1371/journal.pone.0161630.g002Fig 2 Theoretical Model Generated Sequence Entropy Rate Vs. LZ Entropy Rate of Sequence Obtained from Power Law Models. Fig 3 presents the comparisons between the theoretical model and the noise-free random waypoint-based models with and without dwelling. Similar to the power law based empirical model, entropy rates at large bin widths exceed those of the theoretical model. However, the effect of dwelling is less pronounced than power law-based models, because fewer constraints were placed on the trip length in the random waypoint model. The trip segments, therefore, were longer and fewer trip segments (2 to 5 segments as compared to 186 to 292 for the power law model in the conducted experiments) were required to obtain the desired numbers of location samples. This resulted in fewer dwell occurrences in the random waypoint model than their power law counterparts. The theoretical model shows admirable agreement for the entropy rate scaling behavior for both synthetic mobility models. Deviation from theoretical behavior is apparent for very small and very large values of W. 10.1371/journal.pone.0161630.g003Fig 3 Theoretical Model Generated Sequence Entropy Rate Vs. LZ Entropy Rate of Sequence Obtained from Random Waypoint Models. To show the effects of added measurement noise to the power law and random waypoint based models on entropy rate, Fig 4 presents the entropies of the sequences obtained from these models, with dwelling enabled, alongside the entropies of their noisy versions for σ = 10m, a value typical for consumer GPS systems. Fig 4 shows that the introduced zero mean Gaussian noise does not significantly alter the entropy rate, particularly as grid size increases. The probability that a given measurement falls outside the current grid cell, given the accuracy of GPS systems, is small for the sizes of cells considered. Smaller cells would be more susceptible to noise deviations, and might show greater impact on entropy rate, but that impact would be predominantly sensor noise and not the phenomenon of interest. While compensating for noise using more complex models such as Eq (10) may be possible, a simpler solution in some circumstances would be to use bin sizes larger than the expected error, but that still capture the phenomenon of interest. 10.1371/journal.pone.0161630.g004Fig 4 Theoretical Model Generated Sequence Entropy Rate Vs. LZ Entropy Rate of Power Law and Random Waypoint Models with and without Noise, and with Dwelling. Fig 5 compares the curves generated by the theoretical and simulation models. For each simulation model, we compare the curves, relating entropy H to W for different values of T, with the corresponding curves of the theoretical model. Each boxplot in Fig 5 is generated with the R2 values of fitting the theoretical curves to the curves of the simulation models over all T. All but the power law with dwelling model show exceptional fit quality (in excess of 0.9), and even the poorer fitting models have an R2 of about 0.8. The shortcomings of the R2 metric on non-linear models notwithstanding, these results provide us with additional confidence in the fit quality visually evident in the previous figures. 10.1371/journal.pone.0161630.g005Fig 5 Fitness of Theoretical Curves to Simulation Models. Explanation of Results The theoretical model provides a surprising degree of agreement with the synthetic mobility models, suggesting that the mechanics of compression have a great deal to do with the scaling behavior reported in the literature. Our derivation indicated that, subject to our assumptions, the scaling model should form an upper bound on the entropy rate, as any deviations from a unique straight line path would reduce repetition in the string, and, therefore, increase the entropy rate. However, when the theory deviates from the prediction, it almost always underestimates the entropy rate calculated from Lemple-Ziv compression. This is primarily due to violations of two of our assumptions, made to make the mathematics tractable. First, while we assumed a unique termination character during our derivation, we did not supply a unique termination character at the end of strings built from the simulation. This has the counterintuitive result of increasing the estimated entropy rate. Consider a sequence of four symbols. If all symbols are the same, ∑i=0L-1Λi=8 under our assumption, compared to ∑i=0L-1Λi=3 according to Eq (1). Therefore, theoretical entropy rate drops faster than the LZ-entropy for larger W. Second, we assumed that the agent traversed the entirety of each block that it encountered; however, this is not necessarily the case in practice. For example, a path which traverses cell A, clips cell B and traverses cell C could have a corresponding location string of ‘AAAAAAAACCBBBBBB’, whereas the theory implicitly assumes that the path must be ‘AAAAAAAABBBBBBBB’. While this assumption was reasonable at small W, at larger scales, real paths are less likely to transit in a four connected manner. This effect also demonstrates that there are representational effects in the compression calculation. With grid and travel path at arbitrary relative orientations, paths which clip the edge of a cell are possible, and increasingly likely with increasing cell size, increasing the entropy rate at larger scales beyond the theoretical prediction. However, despite these shortcomings, the predicted values showed excellent agreement with the empirical values computed from LZ compression on simulated paths. These results are encouraging for extending our model to incorporate real empirical data, which is confounded by missing data, varying sample sizes and non-Gaussian noise processes. This model should provide a firm theoretical basis for continuing work to address the more difficult situations encountered in real data. Discussion In this paper, we have described a methodology for estimating the differences in predicted entropy rates over different spatial and temporal scales, with and without Gaussian noise, grounded in the theoretical behaviour of the Lempel-Ziv compression algorithm typically used to the calculate mobility entropy rate. We have demonstrated that scaling behaviour is to be expected and is inversely proportional to the spatial scale, and proportionate to the logarithm of the sampling rate. From these derivations, we were able to demonstrate that there is a predicted sampling rate of maximal entropy rate, which can be calculated using the Lambert W function. This theoretical model was validated against models of simulated movement, and found to provide excellent fits for stylized results, but with declining impact at very large or small spatial scales where our assumptions begin to break down. These results are important for a number of reasons. First, we establish a strong theoretical foundation for mobility entropy rate scaling behavior observed and reported by a number of other authors [24, 26]. Based on an analysis of the behavior of Lempel-Ziv compression on the kinds of strings created by agents moving through space, we were able to demonstrate that the mobility entropy rate scaling behavior could be described with only four terms: the length of the path, the average velocity of the agent, the width of the spatial bin, and the period of the sampling rate. Because the scaling law encodes both parameters related to agent motion (x, v) and experimental design (W, T), we can conclude that the scaling depends both on agent behavior and the mathematical realization of that path. This finding is important, as it indicates that the scaling behavior encodes the mobile agent’s behavior, and is not purely an artifact of mathematics, and, therefore, is itself a potentially useful metric. This finding also opens a clear opportunity to separate the two components of entropy rate scaling, providing the ability to isolate the behavioral fingerprint represented in the data. Second, the scaling law is general, subject to the assumptions. Because the trajectory compressed using Lempel-Ziv itself is a single dimensional manifold, as long as the space decomposition and path definition is analogous to the four-connected path described in the assumptions, the scaling law is valid. Similarly, because LZ compression does not distinguish between symbols, only symbol order, any non-overlapping path that crosses the entirety of a cell along only cardinal directions is also valid. We note that while describing the trajectories of people was our primary motivation, this derivation applies to the trajectory of any agent moving through space, subject to our assumptions. Third, the structure of the equation indicates that the differences matter. As shown in the results and in previous works [24, 26], changing the scale of measurement can have a significant impact on the resulting entropy rate calculation. Directly comparing mobility entropy rates from experiments with differing spatial and temporal resolutions is not meaningful. Estimates of entropy rate at a common spatio-temporal resolution, either using the upper bound estimate here, or through an empirical estimate, would be required. This outcome is particularly important for spatial scale, as it implies that the results for studies with heterogeneous cell sizes may be confounded by scaling effects, particularly if the frequency of visits to cells of different sizes is significantly different for different participants. Finally, the scaling law has a maximum value with respect to T, implying that there is a preferred sampling rate for a given spatial and velocity profile. This is an obvious point to use as a common comparator between datasets. Datasets with similar entropy rate maxima will likely have more similar scaling properties than those that do not. This property is also potentially useful for researchers designing data collection studies, as they could use anticipated average velocity, trip length and spatial bin size to identify a preferred sampling period T. Limitations and Future Work The primary limitation in this work is the set of assumptions which made the theoretical analysis tractable. By assuming that the agent was always in motion, and that the path contained no repetitions, and through use of a simple noise model, we have constrained the generalizability of the findings. However, the model matched well against simulated systems, and is relatively straightforward to calculate. The primary goal of any future work should be to extend our results to encapsulate a more broadly representative model of human mobility and noise processes. The second major limitation of our assumptions was that the discretization of space was based on equally dimensioned square grid cells. While this is a reasonable assumption, in practice, researchers have employed cellular tower records to provide the discretization of space (e.g. [10]), leading to a distribution of cell sizes based on the Voronoi diagram of the cell towers’ spatial configuration. The irregularity of the cell tower configuration could potentially exacerbate cell clipping effects, and make the entropy rate dependent on the path the agent takes though the cell. A more sophisticated analysis treating both cell shape and path orientation as independent random variables might address these issues; however, that analysis requires a substantial additional body of research. Similarly, time scales from call records are not constant and depend on individual calling patterns. Extending our work so that spatial resolution and sampling rate can also be represented as random variables would be an important step forward. Finally, we validated our scaling law against simulated mobility models. The model provided surprisingly good fits given the strength of the assumptions, and the fact that both simulated systems violated those assumptions. However, the stylized mobility models employed, while popular, have been shown to be imperfect representations of human mobility [11, 28]. It is a priority to validate the scaling law against actual mobility data. Concluding Remarks The findings presented here provide a theoretical explanation for the scaling behavior observed in calculations of mobility entropy rate from strings of locations using Lempel-Ziv compression. These results, while based on stylized assumptions, provided a useful approximation of scaling behavior for a wide variety of simulated paths, knowing only the average velocity, even under simulated sensor noise. The theory and simulated results provided close agreement for a wide range of spatial and temporal sampling scales, only breaking down at relatively large (corresponding to long repetitions of single symbols) or very small (corresponding to strings of unique symbols) spatial scales, indicating that our assumptions are plausibly valid. The entropy rate scaling formulation has a maximum at a particular sampling frequency, implying that optimal sampling regimes for given trajectories should exist and are in principle approximatable. This work is an important step in transforming mobility entropy rate from a scientific curiosity into a reliable workhorse of modern mobility and spatial behavior studies. By extending this work to emprical data and less stylized mobility assumptions, a scale-free mobility entropy rate formulation may be derived. Supporting Information S1 Appendix Detailed Scaling Law Derivation. (PDF) Click here for additional data file. S1 Data Relevant data and program code to generate the data. (ZIP) Click here for additional data file. We would like to acknowledge the Natural Sciences and Engineering Research Council of Canada for providing funding, and Dr. Mark A. Smith of Sandia National Laboratories for initial discussions regarding entropy rate scaling effects. ==== Refs References 1 Isaacson M , Shoval N . Application of Tracking Technologies to the Study of Pedestrian Spatial Behavior . The Professional Geographer . 2006 ;58 (2 ):172 –183 . 10.1111/j.1467-9272.2006.00524.x 2 Brown BB , Werner CM , Tribby CP , Miller HJ , Smith KR . Transit use, physical activity, and body mass index changes: objective measures associated with complete street light-rail construction . American journal of public health . 2015 ;105 (7 ):1468 –1474 . 10.2105/AJPH.2015.302561 25973829 3 Cho E, Myers SA, Leskovec J. Friendship and Mobility: User Movement in Location-based Social Networks. In: Proceedings of the 17th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. KDD’11. New York, NY, USA: ACM; 2011. p. 1082–1090. Available from: http://doi.acm.org/10.1145/2020408.2020579. 4 Hashemian MS, Stanley KG, Knowles DL, Calver J, Osgood ND. Human network data collection in the wild: the epidemiological utility of micro-contact and location data. In: Proceedings of the 2nd ACM SIGHIT International Health Informatics Symposium. ACM; 2012. p. 255–264. 5 Hashemian M, Knowles D, Calver J, Qian W, Bullock MC, Bell S, et al. iEpi: an end to end solution for collecting, conditioning and utilizing epidemiologically relevant data. In: Proceedings of the 2nd ACM international workshop on Pervasive Wireless Healthcare. ACM; 2012. p. 3–8. 6 Eagle N , Pentland A . Reality mining: sensing complex social systems . Personal and ubiquitous computing . 2006 ;10 (4 ):255 –268 . 10.1007/s00779-005-0046-3 7 Draghici A, Agiali T, Chilipirea C. Visualization system for human mobility analysis. In: RoEduNet International Conference-Networking in Education and Research (RoEduNet NER), 2015 14th. IEEE; 2015. p. 152–157. 8 Pu J, Xu P, Qu H, Cui W, Liu S, Ni L. Visual analysis of people’s mobility pattern from mobile phone data. In: Proceedings of the 2011 Visual Information Communication-International Symposium. ACM; 2011. p. 13. 9 Song C , Qu Z , Blumm N , Barabási AL . Limits of predictability in human mobility . Science . 2010 ;327 (5968 ):1018 –1021 . 10.1126/science.1177170 20167789 10 Song C , Koren T , Wang P , Barabási AL . Modelling the Scaling Properties of Human Mobility . Nature Physics . 2010 ;6 (10 ):818 –823 . 10.1038/nphys1760 11 Lee K , Hong S , Kim SJ , Rhee I , Chong S . SLAW: A New Mobility Model for Human Walks In: INFOCOM 2009 , IEEE ; 2009 p. 855 –863 . 12 Rhee I , Shin M , Hong S , Lee K , Kim SJ , Chong S . On the Levy-Walk Nature of Human Mobility . Networking, IEEE/ACM Transactions on . 2011 6 ;19 (3 ):630 –643 . 10.1109/TNET.2011.2120618 13 Paul T, Stanley K, Osgood N, Bell S, Muhajarine N. Scaling Behavior of Human Mobility Distributions; 2016. Accepted. 14 Mathew W, Raposo R, Martins B. Predicting future locations with hidden markov models. In: Proceedings of the 2012 ACM Conference on Ubiquitous Computing. ACM; 2012. p. 911–918. 15 Chon Y, Shin H, Talipov E, Cha H. Evaluating mobility models for temporal prediction with high-granularity mobility data. In: Pervasive computing and communications (percom), 2012 ieee international conference on. IEEE; 2012. p. 206–212. 16 González MC , Hidalgo CA , Barabási AL . Understanding individual human mobility patterns . Nature . 2009 ;458 (7235 ):238 –238 . 10.1038/nature07850 17 Xu KS . Predictability of social interactions . arXiv preprint arXiv:13061271 . 2013 ;. 18 Baumann P, Santini S. On the use of instantaneous entropy to measure the momentary predictability of human mobility. In: Signal Processing Advances in Wireless Communications (SPAWC), 2013 IEEE 14th Workshop on; 2013. p. 535–539. 19 Jensen BS, Larsen JE, Jensen K, Larsen J, Hansen LK. Estimating human predictability from mobile sensor data. In: Machine Learning for Signal Processing (MLSP), 2010 IEEE International Workshop on; 2010. p. 196–201. 20 Roy BN . Fundamentals of classical and statistical thermodynamics . John Wiley & Sons ; 2002 . 21 Shannon CE . A Mathematical Theory of Communication . The Bell System Technical Journal . 1948 ;27 (3 ):379 –423 . 10.1002/j.1538-7305.1948.tb01338.x 22 Kontoyiannis I , Algoet PH , Suhov YM , Wyner AJ . Nonparametric entropy estimation for stationary processes and random fields, with applications to English text . Information Theory, IEEE Transactions on . 1998 ;44 (3 ):1319 –1327 . 10.1109/18.669425 23 Rodriguez-Carrion A, Garcia-Rubio C, Campo C, Das SK. Analysis of a fast LZ-based entropy estimator for mobility data. In: Pervasive Computing and Communication Workshops (PerCom Workshops), 2015 IEEE International Conference on. IEEE; 2015. p. 451–456. 24 Smith G, Wieser R, Goulding J, Barrack D. A refined limit on the predictability of human mobility. In: Pervasive Computing and Communications (PerCom), 2014 IEEE International Conference on. IEEE; 2014. p. 88–94. 25 Lin M, Hsu WJ, Lee ZQ. Predictability of individuals’ mobility with high-resolution positioning data. In: Proceedings of the 2012 ACM Conference on Ubiquitous Computing. ACM; 2012. p. 381–390. 26 Qian W , Stanley KG , Osgood ND . The Impact of Spatial Resolution and Representation on Human Mobility Predictability In: Web and Wireless Geographical Information Systems . Springer ; 2013 p. 25 –40 . 27 Research M. GeoLife: Building social networks using human location history; 2014. [Online; accessed 01-July-2014]. http://research.microsoft.com/en-us/projects/geolife/. 28 Hossmann T, Spyropoulos T, Legendre F. Putting contacts into context: Mobility modeling beyond inter-contact times. In: Proceedings of the Twelfth ACM International Symposium on Mobile Ad Hoc Networking and Computing. ACM; 2011. p. 18.

PMC005xxxxxx/PMC5003382.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757120210.1371/journal.pone.0161822PONE-D-16-05504Research ArticleSocial SciencesSociologyEducationSchoolsBiology and Life SciencesGeneticsGenomicsGenomic MedicinePeople and PlacesGeographical LocationsEuropeDenmarkResearch and Analysis MethodsResearch DesignSurvey ResearchQuestionnairesSocial SciencesSociologyEducationSymposiaBiology and Life SciencesEvolutionary BiologyPopulation GeneticsBiology and Life SciencesGeneticsPopulation GeneticsBiology and Life SciencesPopulation BiologyPopulation GeneticsBiology and Life SciencesNeuroscienceCognitive ScienceCognitive PsychologyAcademic SkillsLiteracyBiology and Life SciencesPsychologyCognitive PsychologyAcademic SkillsLiteracySocial SciencesPsychologyCognitive PsychologyAcademic SkillsLiteracyPeople and PlacesPopulation GroupingsProfessionsTeachersSpitting for Science: Danish High School Students Commit to a Large-Scale Self-Reported Genetic Study Genomic Literacy in Danish High SchoolsAthanasiadis Georgios 12*Jørgensen Frank G. 3Cheng Jade Y. 1Kjærgaard Peter C. 245Schierup Mikkel H. 126Mailund Thomas 121 Bioinformatics Research Centre, Aarhus University, 8000, Aarhus, Denmark2 Centre for Biocultural History, Aarhus University, 8000, Aarhus, Denmark3 Tørring Gymnasium, 7160, Tørring, Denmark4 Department of Culture and Society, Aarhus University, 8000, Aarhus, Denmark5 The Natural History Museum of Denmark, University of Copenhagen, 1471, Copenhagen, Denmark6 Department of Bioscience, Aarhus University, 8000, Aarhus, DenmarkUddin Monica EditorUniversity of Illinois at Urbana-Champaign, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: PCK MHS TM FGJ. Data curation: GA JYC. Formal analysis: GA JYC TM. Funding acquisition: PCK MHS TM. Investigation: JYC FGJ TM. Methodology: PCK MHS TM FGJ. Software: JYC TM. Supervision: PCK MHS TM. Visualization: GA. Writing – original draft: GA. Writing – review & editing: GA FGJ PCK MHS TM. * E-mail: athanasiadis@birc.au.dk29 8 2016 2016 11 8 e01618227 2 2016 14 8 2016 © 2016 Athanasiadis et al2016Athanasiadis et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Scientific outreach delivers science to the people. But it can also deliver people to the science. In this work, we report our experience from a large-scale public engagement project promoting genomic literacy among Danish high school students with the additional benefit of collecting data for studying the genetic makeup of the Danish population. Not only did we confirm that students have a great interest in their genetic past, but we were also gratified to see that, with the right motivation, adolescents can provide high-quality data for genetic studies. National Lottery Funds (Danske Spil)Kjærgaard Peter C. Danish Ministry of Education and the Centre for Ciocultural History (Aarhus university)This project and related research were supported from the National Lottery Funds (Danske Spil, granted to Prof. Peter C. Kjærgaard), The Danish Ministry of Education and the Centre for Biocultural History, Aarhus University. The funders had no role in the study design, data collection or analysis. Data AvailabilityGenetic and phenotypic data have been deposited at the European Genome-Phenome Archive (EGA; https://ega-archive.org) under accession number EGAS00001001868.Data Availability Genetic and phenotypic data have been deposited at the European Genome-Phenome Archive (EGA; https://ega-archive.org) under accession number EGAS00001001868. ==== Body Introduction Genomic literacy among the general public remains a hot topic in the era of precision medicine and translational genomics [1–4], and it encompasses knowledge about the biochemistry and structure of the genome, basic rules of heredity, the polygenic nature of common diseases, the interaction between genes and environment etc. In addition to imparting basic scientific knowledge, genomic literacy covers ethical, legal, and social implications (ELSI) that are essential in order for individuals to make informed decisions about disease [5] or to understand genetic ancestry. Initiatives promoting genomic literacy can involve any segment of society. In a typical scenario, researchers partner up with public entities with the support of third-party funding agencies [6]. Denmark has a long tradition of supporting such partnerships. Every year in December, the Danish Ministry of Education invites groups involved in scientific outreach and education to apply for funding from the profits of Danske Spil–the national lottery in Denmark. The selection of applicants is straightforward, allowing work on the ground to start within a six-month timeframe. Inspired by the increasing need for genomic literacy, we designed Where Are You From? (http://hvorkommerdufra.dk/)–a high school genomic project aiming at (i) demystifying genetic concepts regarding human ancestry; (ii) building bridges between academia and young students interested in pursuing a scientific career; and (iii) collecting high-quality data for genetic analysis. Such data can provide insights into e.g. the genetic affinity of the Danish population with other European populations or historical patterns of admixture and migration, thus improving our knowledge about the genetic history of Denmark [7]; they can also help us build and validate genetic predictors of anthropometric traits like height and body mass index, thus providing a translational framework for improving genetic risk assessment [7]. In brief, the project consisted of collecting DNA and anthropometric data, and included two symposia at the Aarhus University campus, where basic genetic concepts and preliminary scientific results were presented to the students. Materials and Methods Outreach Once funds from Danske Spil were secured, we published an official advertisement in a Danish magazine for high school biology teachers, inviting schools from across Denmark to participate in our project [8]. The response was overwhelming: teachers from as many as 40 out of a total of 168 high schools in the entire country expressed initial interest in volunteering for the project (Fig 1). Most of the partners were general high schools (gymnasium or STX in Danish) with the exception of two technical high schools (teknisk gymnasium or HTX). Because of the educational structure in Denmark, STX students have a greater than average interest in science and technology. This implies that, to our knowledge, our project was one of the largest gatherings of students with a potential interest in genetics in the history of Danish education. 10.1371/journal.pone.0161822.g001Fig 1 Flowchart of the Where Are You From? project. The flowchart displays recruitment, dropout and exclusion rates, as well as tentative reasons for the latter two. The project included two symposia that took place at the Aarhus University campus–one before and one after data collection (Fig 2)–and covered several topics of genomic analysis applied to humans. In particular, during the first symposium, the students had the chance to learn about basic concepts of human evolution, which included natural selection, the use of mtDNA and Y-chromosome haplotypes to track migration patterns, the use of recombination to study human history and relatedness, the admixture of anatomically modern humans with Neanderthals and Denisovans, but also about genotyping technologies and their potential use in association mapping. During the second symposium, we revisited many of the concepts from the first symposium and we presented preliminary results of the genetic analysis of the students’ DNA. 10.1371/journal.pone.0161822.g002Fig 2 Timeline of the Where Are You From? project. Data collection We asked students to prepare an Oragene•Dx® saliva collection kit from Genotek (Ottawa, Canada) for DNA analysis during science class under the supervision of their teachers and to respond to an online questionnaire about their age; their own, their parents’ and their grandparents’ place of birth; their family’s level of education; and some basic anthropometric traits (S1 Table). We outsourced genotyping to 23andMe (Mountain View, CA, USA) on the grounds of cost-effectiveness and user-friendliness. Initially, students activated their 23andMe profiles, but long before the company returned any genetic results to the users, we changed passwords for all profiles thus preventing access to health-related information outside this project’s scope. We also minimized our own access to health-related information by assigning the entire database handling to a single individual. Consent and ethics approval All students gave their written informed consent to participate in the project, signed either by themselves (>18 years old) or by their parents/guardians (S2 Table). The institutional review board of Aarhus University approved the study. The Regional Research Ethics Committee of Central Jutland informed us orally that no additional medical ethics approval would be required provided that our project entailed no health-related research. Indeed, our questionnaire did not include any health-related questions, and participants had no access to health-related information from their 23andMe profiles (see previous section). A full guideline for medical ethics approval for research in Denmark can be found at http://www.rm.dk/sundhed/faginfo/forskning/de-videnskabsetiske-komiteer/ (in Danish). Results Participation rates Of the 40 schools that initially contacted us, 36 engaged in the activities of the project–i.e. data collection and attendance at the symposia (Table 1). Dropout was primarily due to calendar inconvenience or lack of travel funds. A total of ~1,100 students from the 36 schools volunteered for our project, but because there were not enough funds to cover the genotyping of all of them, we prioritized ~800 while trying to maximize representation of all regions of Denmark (Fig 3). However, regardless of whether they were sampled or not, all students were invited to the outreach activities. After genotyping was completed, we had access to genetic data from 722 kits and information from 662 questionnaires, with 600 students from 33 schools providing both saliva samples and answers to the questionnaire (Fig 4A). To avoid coercion, we did not investigate the reasons for dropping out. 10.1371/journal.pone.0161822.t001Table 1 Name, geographic coordinates and number of questionnaire respondents (N) for each of the 36 schools participating in the Where Are You From? project. High school name Latitude Longitude N respondents Aalborg Katedralskole 57.05 9.91 18 Aarhus Statsgymnasium 56.16 10.17 13 Brønderslev Gymnasium 57.28 9.94 16 Egaa Gymnasium 56.21 10.27 20 Esbjerg Gymnasium 55.49 8.49 22 Fredericia Gymnasium 55.58 9.75 20 Frederiksbjerg Gymnasium 55.68 12.53 24 Frederiksværk Gymnasium og HF 55.97 12.01 8 Grenaa Gymnasium 56.41 10.89 11 Haderslev Katedralskole 55.26 9.49 29 Herningsholm HTX 56.15 8.99 6 Himmerlev Gymnasium 55.66 12.11 18 Horsens Gymnasium 55.84 9.83 30 Køge Gymnasium 55.46 12.17 14 Langkær Gymnasium 56.19 10.11 - Mariagerfjord Gymnasium 56.63 9.81 17 Marselisborg Gymnasium 56.14 10.2 19 Mulernes Legatskole 55.41 10.43 19 Nakskov Gymnasium og HF 54.84 11.14 12 Nørresundby Gymnasium og HF 57.07 9.94 22 Nykøbing Katedralskole 54.77 11.88 41 Ribe Katedralskole 55.33 8.76 25 Risskov Gymnasium 56.19 10.21 15 Rødkilde Gymnasium 55.71 9.55 5 Roskilde Gymnasium 55.64 12.08 26 Silkeborg Gymnasium 56.18 9.6 37 Skive Gymnasium 56.55 9.03 19 Sorø Akademi 55.43 11.56 39 Tørring Gymnasium 55.86 9.48 15 Varde Gymnasium 55.63 8.5 19 Vejles Tekniske Gymnasium 55.71 9.54 12 Vestfyns Gymnasium 55.27 10.11 26 Viborg Gymnasium og HF 56.46 9.45 10 Viby Gymnasium 56.11 10.15 7 Viby Teknisk Gymnasium 56.18 10.19 17 Virum Gymnasium 55.79 12.48 11 Total N 662 10.1371/journal.pone.0161822.g003Fig 3 Geographic distribution of the 36 Danish high schools participating in the Where Are You From? project. Color code reflects the five administrative divisions of Denmark. Number of schools per division is shown in brackets. Jutland (and Aarhus in particular) high schools were overrepresented in the sampling, due to their proximity to the symposium site in Aarhus University campus. 10.1371/journal.pone.0161822.g004Fig 4 Summary of sample recruitment. (A) Six hundred students from 33 different schools provided both genetic and questionnaire data. (B) Histogram of % questionnaire completeness (N = 662). (C-E) Box plots (median and interquartile range) of % questionnaire completeness by sex, age and highest parental education level (1 = unskilled; 2 = skilled; 3 = short-term higher education; 4 = long-term higher education). Whiskers represent data within 1.5 times the interquartile range and circles represent outliers. Per-individual completeness was defined as the number of fields completed with valid answers over the total number of fields. The age of participants ranged from 15 to 20 (median = 18) and their female-to-male ratio was 2.16. Questionnaire completeness was particularly high with 566 out of 662 (85.5%) students delivering at least 80% complete questionnaires (Fig 4B). Interestingly, female students delivered more complete questionnaires than male students (Mann-Whitney U = 52881, p = 0.01; Fig 4C), but no significant clustering was observed by age or parental education background (Kruskal-Wallis H test, p > 0.05; Fig 4D and 4E). Genetic and phenotypic data have been deposited at the European Genome-Phenome Archive (EGA; https://ega-archive.org) under accession number EGAS00001001868. Modern Danish society accommodates different ethnic and cultural groups and this was also reflected in our sample, where ~4% of the participants were born in countries like Afghanistan, China, Ethiopia, Finland, Germany, Greenland, Iraq, Jordan, Korea, Kosovo, the Netherlands and Zambia. This number went up when we looked at grandparental origin, where 14.6% of the participants had at least one grandparent born outside Denmark. Four hundred seven (407) participants had their four grandparents born in Denmark (a typical inclusion criterion when analyzing historical patterns of genetic variation), whereas 131 participants had their four grandparents born in more specific regions (Table 2). 10.1371/journal.pone.0161822.t002Table 2 Number of participants representing six well-defined geographic regions in Denmark. Region N (all four of the grandparents born in the same region) N (at least three of the grandparents born in the same region) Capital Region 4 14 Zealand 21 36 Funen 6 13 South Jutland 24 37 Central Jutland 48 89 North Jutland 28 48 Total 131 237 Discussion In many aspects, high schools are an ideal platform for promoting scientific literacy [9]. Experience has shown that high school students combine all the useful characteristics needed for successful scientific training: a genuine interest in novel technologies; the maturity to reflect upon ethical, legal, and social aspects of genetics; and the intellectual capacity and availability for such undertakings [10]. Partnerships with high schools, however, must be fast paced so that most participants can receive feedback on their involvement while they are still students. In this spirit, Where Are You From? was coordinated and delivered in an astonishingly short time (Fig 2), making a clear point that good scientific outreach does not need to be lengthy or prohibitively expensive. Perhaps the best indicator of the success of our project was the warm response from students and their teachers, and the extensive publicity it drew in the media with coverage in nation-wide primetime television news channels. Indeed, approximately 600 students and their teachers–some traveling as far as 300 km–signed up for the second symposium, in which we presented preliminary results on genetic ancestry and related demographic topics. Even though this number is smaller compared to the first symposium (~780 registered attendees), it still reflects a high level of commitment to the project’s activities. During the second symposium, comprehension of the genomic concepts presented was assessed in situ by use of simple “fun questions”, which the students answered by use of clickers. A diploma with information about Y-DNA and/or mtDNA haplogroups and the percentage of Neanderthal ancestry was awarded to each participant. In addition, filmed and edited material from the two symposia was almost immediately released online (http://hvorkommerdufra.dk/materiale/) for use by schools across Denmark. Despite the disengagement of a few partners in the course of the project (Fig 1), we managed to sample approximately one in 9,350 inhabitants of Denmark. To put this in context, a similar project in the UK (People of the British Isles) sampled one in 16,000 inhabitants [11]. With the help of the genetic data collected, we have studied the fine-scale population structure and historical genetic influences in Denmark [7]. The merit of our project lies in that (i) it enhances genomic literacy in high school students, one of the most dynamic and malleable demographics of the population and (ii) it points out that students, with the guidance of their teachers, can be as good a source of genetic data as other segments of the general population. Our high school students showed a commitment that is comparable to that of adult participants in medical studies [12]. As already mentioned, we took great care in securing voluntary and fully informed student enrolment. Yet, because the activity was carried out in a school context, we cannot discard sporadic cases of student participation due to real or, more probably, perceived pressure by peers and/or teachers. Finally, we believe that future scientific outreach projects with a focus similar to that of Where Are You From? can benefit largely from incorporating more types of environmental and disease-related data. As showcased by the Precision Medicine Initiative, launched in early 2015 in the USA, such an endeavor is nowadays possible thanks to the increased user connectivity through social media and mobile devices that provide real-time measurements of variables such as glucose concentration, blood pressure and cardiac rhythm [4]. Aided by the technological and social advancements, and contingent on informed consent, future outreach projects and online initiatives, such as our very own TheHonestGene.org, can therefore be the ideal platform for recruiting participants for large-scale genomic studies. Conclusions In sum, the high response rate of high school students to an online questionnaire in the context of a genetic outreach activity is a nontrivial finding, given that little is known about the engagement of adolescents in sample recruitment. Our endeavor sends out an encouraging message to geneticists who intend to design new population genetic projects, promoting, at the same time, genomic literacy among the youngest members of society. Supporting Information S1 Table The online questionnaire answered by all participants in the Where Are You From? project. (DOCX) Click here for additional data file. S2 Table English translation of the participation statement signed by high school students or their parents/guardians. (DOCX) Click here for additional data file. This project and related research were supported from the National Lottery Funds (granted to Prof. Peter C. Kjærgaard), The Danish Ministry of Education and the Centre for Biocultural History, Aarhus University. The funders had no role in the study design, data collection or analysis. The authors would like to thank the high school students and their teachers for participating in the project. ==== Refs References 1 Kung JT , Gelbart ME . Getting a head start: the importance of personal genetics education in high schools . Yale J Biol Med . 2012 ;85 : 87 –92 . 22461746 2 Mirnezami R , Nicholson J , Darzi A . Preparing for precision medicine . N Engl J Med . 2012 ;366 : 489 –491 . 10.1056/NEJMp1114866 22256780 3 Wolf SM , Burke W , Koenig BA . Mapping the Ethics of Translational Genomics: Situating Return of Results and Navigating the Research-Clinical Divide . J Law Med Ethics J Am Soc Law Med Ethics . 2015 ;43 : 486 –501 . 10.1111/jlme.12291 4 Collins FS , Varmus H . A new initiative on precision medicine . N Engl J Med . 2015 ;372 : 793 –795 . 10.1056/NEJMp1500523 25635347 5 Wolf SM . Return of individual research results and incidental findings: facing the challenges of translational science . Annu Rev Genomics Hum Genet . 2013 ;14 : 557 –577 . 10.1146/annurev-genom-091212-153506 23875796 6 Donovan MS . Generating improvement through research and development in education systems . Science . 2013 ;340 : 317 –319 . 10.1126/science.1236180 23599484 7 Athanasiadis G , Cheng JY , Vilhjálmsson BJ , Jørgensen FG , Als TD , Le Hellard S , et al Nationwide genomic study in Denmark reveals remarkable population homogeneity . Genetics . 2016 ; 8 Jørgensen FG , Mailund T , Schierup M , Kjærgaard PC . Hvor kommer du fra? En undersøgelse af danskernes genetiske historie . Biofag . 2013 ;3 : 18 –20 . 9 Science in schools . Nature . 2013 ;497 : 287 –288 . 10 Munn M , Skinner PO , Conn L , Horsma HG , Gregory P . The involvement of genome researchers in high school science education . Genome Res . 1999 ;9 : 597 –607 . 10413399 11 Winney B , Boumertit A , Day T , Davison D , Echeta C , Evseeva I , et al People of the British Isles: preliminary analysis of genotypes and surnames in a UK-control population . Eur J Hum Genet EJHG . 2012 ;20 : 203 –210 . 10.1038/ejhg.2011.127 21829225 12 Athanasiadis G , Malouf J , Hernandez-Sosa N , Martin-Fernandez L , Catalan M , Casademont J , et al Linkage and association analyses using families identified a locus affecting an osteoporosis-related trait . Bone . 2014 ;60 : 98 –103 . 10.1016/j.bone.2013.12.010 24334171

PMC005xxxxxx/PMC5003383.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757097310.1371/journal.pone.0161361PONE-D-16-13545Research ArticleBiology and Life SciencesEcologyEcosystemsForestsEcology and Environmental SciencesEcologyEcosystemsForestsEcology and Environmental SciencesTerrestrial EnvironmentsForestsBiology and Life SciencesOrganismsPlantsTreesSprucesBiology and Life SciencesOrganismsPlantsTreesBiology and Life SciencesOrganismsPlantsTreesBirchesBiology and Life SciencesOrganismsFungiPeople and PlacesPopulation GroupingsEthnicitiesNorwegian PeopleBiology and Life SciencesOrganismsAnimalsInvertebratesArthropodaInsectsBiology and Life SciencesEcologyForest EcologyEcology and Environmental SciencesEcologyForest EcologyAssessment of the Main Natural Disturbances on Norwegian Forest Based on 20 Years of National Inventory Characterisation of Natural Disturbances in Norwayhttp://orcid.org/0000-0003-3829-5759Díaz-Yáñez Olalla 1*Mola-Yudego Blas 12Eriksen Rune 2González-Olabarria José Ramón 31 School of Forest Sciences, University of Eastern Finland, Joensuu, Finland2 Norwegian Institute of Bioeconomy Research, Ås, Norway3 Forest Sciences Centre of Catalonia, Solsona, SpainCarcaillet Christopher EditorEcole Pratique des Hautes Etudes, FRANCECompeting Interests: The authors have declared that no competing interests exist. Conceptualization: ODY BMY JGO. Data curation: ODY BMY RE JGO. Formal analysis: ODY BMY JGO. Funding acquisition: ODY BMY. Methodology: ODY BMY JGO. Project administration: ODY BMY. Software: ODY. Supervision: BMY JGO. Validation: ODY BMY JGO. Visualization: ODY. Writing – original draft: ODY BMY JGO. Writing – review & editing: ODY BMY RE JGO. * E-mail: olalla.diaz@uef.fi29 8 2016 2016 11 8 e01613614 4 2016 4 8 2016 © 2016 Díaz-Yáñez et al2016Díaz-Yáñez et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.The re-measurement of permanent forest inventories offers a unique opportunity to assess the occurrence and impact of forest disturbances. The present study aims at exploring the main forest damages in Norway based on the extensive data of several consecutive national forest inventories during the period 1995–2014. Five of the most common disturbance agents in Norway are selected for analysis: wind, snow, browsing, fungus and insect damage. The analyses focuses on the frequency and variation along time, the average damage at stand level and the spatial patterns of damage occurrence, resulting in a characterization of the damage produced by disturbances in Norway. The highest damage occurrences by disturbance agent are due to browsing, snow and wind. Snow presents a decreasing temporally trend in damage frequency in the studied period. By forest type, mature and intermediate birch forest are found to be more affected by snow damage, whereas mature spruce forest is by wind damage. The results from this study provide support to the hypothesis that damages by autumnal moth (Epirrita autumnata) on birch are more common in mature stands. No major attacks from bark beetle (Ips typographus) are found, probably related to the lack of major storm damages in the period. Forest types susceptibility to fungus has no apparent variation over time except in the last years, as increased occurrence is observed on mature spruce stands probably correlated with warmer than average periods. Browsing damage causes the most severe losses, as expected, in young stands, and is allocated mainly on the most productive forests. Although some of the disturbances present locally moderate effects, the results show no major disturbances threatening Norwegian forests in the studied period. Finally, the Norwegian national forest inventory demonstrates its reliability as a basis to understand the occurrence and effects of major natural disturbances. SIS-RedclimDoctoral School of the University of Eastern Finlandhttp://orcid.org/0000-0003-3829-5759Díaz-Yáñez Olalla We are grateful to the forest doctoral school of the University of Eastern Finland for their kind economic support, as well as to the SIS-Redclim, funded by the Norwegian Institute for Bioeconomy Research. Data AvailabilityThe authors confirm that all data underlying the findings of this study are fully available from The Norwegian Institute of Bioeconomy Research, partly openly accessible at http://www.skogoglandskap.no/temaer/statistikk_fra_landsskogstakseringen and https://ssb.no/en/jord-skog-jakt-og-fiskeri/statistikker/lst/aar?fane=arkiv#content and partly by contacting the institute through Aksel Granhus (aksel.granhus@nibio.no). The data used in this study are third-party data. Data are collected as part of a national monitoring program. The use of data is restricted by the data use policies of the national monitoring program of which they are part. Data are available upon request and in accordance with the guidelines of the monitoring program of which they are part.Data Availability The authors confirm that all data underlying the findings of this study are fully available from The Norwegian Institute of Bioeconomy Research, partly openly accessible at http://www.skogoglandskap.no/temaer/statistikk_fra_landsskogstakseringen and https://ssb.no/en/jord-skog-jakt-og-fiskeri/statistikker/lst/aar?fane=arkiv#content and partly by contacting the institute through Aksel Granhus (aksel.granhus@nibio.no). The data used in this study are third-party data. Data are collected as part of a national monitoring program. The use of data is restricted by the data use policies of the national monitoring program of which they are part. Data are available upon request and in accordance with the guidelines of the monitoring program of which they are part. ==== Body 1. Introduction Natural disturbances are a key factor in forest dynamics [1], and a reason for the potential distribution of terrestrial vegetation [2]. At the same time, disturbances are considered as a major threat to forest resources [3] and associated ecosystem services [4]. The impact either positive or negative, of natural disturbances greatly depends on their regime, and how the affected ecosystems are adapted to cope with the intensity and recurrence of those disturbances. High intensity (i.e. stand level or larger) events are usually attracting most of the attention of society and, arguably, the scientific community. Nevertheless, disturbances have an important impact on the evolution of a forest even when their intensity and associated severity are mild. Understanding e.g. the establishment of a stand after a severe disturbance, or the diversification of species and structure coming from a frequent but mild disturbance, provides the forester with information that can be included in management plans and objectives. This assertion can be applied either, if the objective is to emulate through management the evolution pathways of a natural forest [5,6], account for expected losses when defining a management plan [7], or combine economic and risk mitigation objectives [8]. In this context, the occurrence, susceptibility and impact of natural disturbances on forest ecosystems has been studied at different temporal and spatial scales, relying on a variety of methodological approaches and data sources [4,9]. In general, disturbance regimes are defined by the frequency of the events, their spatial and temporal distribution and impact on forest [10,11]. Even if the components of a disturbance regime depend largely on the type of disturbance [12], the spatial and compositional characteristics of the forest also has an important role on modifying both the extent and severity of disturbances [13]. The variability on disturbance regimes presents a challenge when large areas and multiple disturbance types are to be analyzed, as the spatial and temporal frames should be large enough to reflect variations on the spatial patterns of the events recurrence. For instance, remote sensing tools have shown their usefulness to capture information on the spatial distribution and impact of different types of disturbances at the required spatial and temporal scales [14–16]. However, remote sensing tools rely on variations on the vegetation structure and vitality, being traditionally better adapted to capture large-scale, stand replacing forest disturbances [17]. In addition, they require from previous information about the type of disturbance causing the observed variation on forest conditions. Field assessments on forest health, on the other hand, offer the possibility of identifying the disturbance agents, and provide reliable information about the rate of damage, even when it is small or located on non-dominant forest strata. This valuable information comes at a high economic cost if large areas are to be monitored over an extended period of time. National Forest Inventories (NFIs), although not designed for the specific purpose of assessing forest health, are considered a potential source of information that can produce and complement disturbance related information [18,19]. Although NFIs are not spatially continuous, they provide a good approximation at the national scale on the spatial distribution and temporal evolution of forest types and associated goods and services [20]. When measurements of forest health are included on the NFIs design, they become an excellent source of data for assessing the influence of forest characteristics on the occurrence and severity of natural disturbances. Examples on the use of NFIs for assessing occurrence and impact of disturbances can be found for pests or diseases [21–23], fire [24–27], wind and snow [28–31], or game related damage [32,33], among others. The present study aims at exploring forest disturbances in Norway by their specific agent, with emphasis on Snow, Wind, Browsing, Insect, and Fungus related damage. The main focus is to identify spatial and temporal patterns on the occurrence of forest disturbances based on data from four consecutive measurements of the Norwegian NFI, entailing 1995–2014. Additionally, it provides an overall insight on the occurrence and impact of natural disturbances on the Norwegian forest, analysing the susceptibility of different forest types to be affected by disturbances, and an overall assessment of the damage levels. 2. Materials and Methods 2.1. Data sources The data for the analysis was based on the Norwegian National Forest Inventory (NFI) collected during the 7th, 8th, 9th and 10th measurements, corresponding to the period 1995–2014. The Norwegian NFI is a systematic permanent inventory, where plots allocated on a 3x3 km grid are measured every 5 years. For each permanent plot, variables were measured at three levels: stand (1000 m2 around the plot center), plot and tree level (on a circular plot of 250 m2). Forest damage was measured at stand level. The plots included in the analysis were those with damage measurements available and located in productive forest (i.e. expected yield over 1 m3 ha-1 year-1 over bark [34]). During 1995–2005 damage was not recorded in plots that were in regeneration stage. For plots divided by a stand border (e.g. a plot has one part on forest and another on water, or on two forest types with very different productive parameters), only the forested part was considered or the first division when both parts were forested areas. In total 34 263 plot measurements were considered for the analysis, 8052, 8423, 8895, 8893 in the 7th, 8th, 9th and 10th NFI respectively; these plots entailing most of the forested parts of the country (Fig 1). The country was divided into regions as defined in the official statistics of forest condition and resources in Norway [34] (Fig 1). These regions are considered to have relatively similar forests, topography and climate (Table 1). The northernmost area of the country, Finnmark, was not included as the measurements in this county have started during the 9th NFI with different grid in non boreal forest areas. 10.1371/journal.pone.0161361.g001Fig 1 Studied area. Left: Studied area and regionalization used in the analysis. The counties were grouped in six main regions, except the northernmost county (Finnmark) that was not included in the analysis. Right: Spatial distribution of the national forest inventory plots by dominant species (spruce, pine, birch and mixed). Map border lines adapted from [35], original licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). 10.1371/journal.pone.0161361.t001Table 1 Climatic characterization of the plots included in the analysis, by region. Region tm(°C) tmin(°C) tmax(°C) Psum N Region 1 3 -0.7 6.7 664.7 8636 Region 2 2.4 -1.5 6.3 691.2 6490 Region 3 4.9 2 7.7 1049.6 5161 Region 4 5.5 2.8 8.1 1636.6 4347 Region 5 3.4 0 6.8 990.9 4889 Region 6 2.5 -0.5 5.5 1040.4 4740 Data based on the averages of the normal climatic period 1960–1990. tm: annual mean temperature tmin: minimum temperature, tmax: maximum temperature Psum: annual precipitation. N: Total number of plots included by region. Following the NFI instructions, damage from different disturbance agents were recorded in each plot at each inventory following the criteria: it had an effect in the future economic development of the stand, it compromises the regeneration, or it represents a relevant decrease in the volume production or wood quality [34]. Damage was only recorded when it was detected to have occurred within 5 years prior to the NFI measurement. In each plot, at least one disturbance agent was associated to the observed damage (Table 2). In the cases that several disturbance agents were reported at the same time in the same plot, the analysis only considered those ranked as the main disturbance agent (i.e. that occasioned the higher damage loss in the stand, as defined in [34]). Also at plot level, damage was quantitatively defined as a loss relative to volume, number of trees or crown loss according to the damage type and disturbance agent (Table 2) [34]. 10.1371/journal.pone.0161361.t002Table 2 Description of the disturbance agents and their associated damage as measured in the Norwegian National Forest Inventory. Disturbance agent Damage Wind Volume of trees blowdown by wind as a percentage of total volume. Snow Number of trees with snow break/blowdown in percentage of the total number of trees. Drought Volume of dead trees as a percentage of total volume. Frost Percentage of stand crown mass that is dead. Fire Volume of dead trees as a percentage of total volume. Landslide Number of trees with break / blowdown as a percentage of total number of trees. Browsing In old forest, it is the percentage of stand crown mass that is grazed away. In young forest it refers to the percentage of dead or damaged future trees as a percentage of the original number of trees. Insects (includes: Bark beetle (Ips typographus), European pine sawfly (Neodiprion sertifer), Autumnal moth (Epirrita autumnata), Weevil (Curculionidae) and insect not specified) Percentage of stands crown dead or fell off/ grazed away except for Bark beetle damage that corresponds with volume of dead trees as a percentage of total volume. Fungus (includes: Spruce needle rust fungus (Chrysomyxa abetis), Lophodermium needle cast pine (Lophodermium) and fungus not specified) Percentage of stand crown that is dead or discoloured for Spruce needle rust fungus and other fungus no specified. And in young forest, percentage of dead future trees as a percentage of the original number of trees. Mechanical (includes: Forest operations or damage by animal) Volume of trees damaged in percentage of the total volume. Mouse, beaver or ungulates Volume of damaged trees as a percentage of total volume in old forest. For young forest it is the percentage of dead future trees from original number of trees. Damage cause not known Percentage of stands total crown that is dead. The plots were divided according to their dominant species into four forest types. The criterion for assigning the forest type was the relative abundance of the main species expressed in volume distribution for older forest and crown cover for young forest. Therefore, plots were classified as spruce (spruce > 70%), pine (pine > 70%), birch (birch > 70%) and mixed forests (dominated by any or several of the previous species and with more than 10% of broadleaves). Plots were also divided according to their stand development class in three categories: young, intermediate and mature (corresponding with development class categories I and II, III and IV and V respectively in the Norwegian NFI [34]). The change from young to intermediate development class is mainly defined when the trees’ mean diameter exceeds 10 cm in spruce and pine forest, and 7–8 cm in birch forest. 2.2. Methodological approaches Five of the most common disturbance agents in Norway were selected for further analysis: wind, snow, browsing, fungus and insects. The analyses focused on the frequency of disturbance occurrence (presence of damage regardless of its impact), its temporal and spatial variation. The variation of the relative frequency of disturbance occurrence along the timeframe of the data was first visually explored by disturbance agent and region. In order to identify possible trends or peaks, linear and polynomial models were developed for each disturbance, identifying those with significant levels (p <0.05). The level of damage was also assessed quantitatively at plot level, and it was compared between regions. The spatial analysis included the identification of areas with the highest concentration of plots affected by the occurrence of each of the disturbance agents. In this case, a geo-statistical approach based on kernel methods was used [36,37]. Kernel methods allow estimating the probability of occurrence of events in a continuous space. In general, the method requires to define a bandwidth parameter that will be use to aggregate the events. In our case, the events were defined as plots presenting damage by one of the disturbance agents selected. The approach taken was based on an adaptive bandwidth method [38] that varies the bandwidth size in relation with the estimated local amount of data (i.e. the size of the bandwidth is inversely related to the amount of data). Although fixed bandwidths are more common because of its simple and effective application [39,40]; it is widely known that they can present problems when dealing with populations showing high spatial inhomogeneity. A fixed large bandwidth will miss the finer variations of highly dense areas and a narrow one will increase the intensity function for isolated points. The global bandwidth needed to obtain the varying bandwidths was calculated based on the over-smoothing factor [41] and the pilot bandwidth was calculated with a leave-one-out least-squares cross-validation (following Bowman and Azzalini, 1997 [42]). Considering the large border effect that the Norwegian geography can have on the estimates, edge corrected adaptive densities were applied. For the Kernel calculations, the R package sparr [43] was used. The frequencies of disturbance occurrence by development class of the forest (i.e. young stands, intermediate and mature) and dominant species were studied using contingency tables. The contingency tables were defined comparing the observed number of plots presenting damage by disturbance agent to the expected frequency values when the number of plots is divided proportionally between development classes. The difference was considered relevant when the observed value per cell was equal or larger than 1.5 the expected value (one-tailed), and otherwise it was assumed no effect of development class by species on the occurrence of the studied disturbance agent. Cells with large difference between observed and expected make a larger contribution to the Chi-squared test. The Chi-squared test was calculated against the null hypothesis that the different disturbance agents have the same effect on all forest types and development classes’ combinations (at the 0.05 level). The characteristics of the stand prior to the damage identification were based on the period before the damage was measured, for example if the disturbance occurrence was measured in the period 2000–2004 (8th NFI) the stand characteristics before the damage corresponds with the information obtained during the period 1995–1999 (7th NFI) in the same plot. From the initial set of 34 263 plot measurements, all of them were used for temporal damage frequency analysis by disturbance agent. Relative frequencies of damage occurrence by disturbance were calculated relative to the number of plots measured in each year. The Norwegian NFI measures one fifth of all the plots in each year, therefore relative yearly estimations are possible to obtain. Damage was analyzed using the plots affected by one of the selected natural disturbances and with measurements of damage impact on the forest (2557 plots); these measurements cover the period (2000–2014) because prior to 2000 only information on disturbance occurrence (damage or undamaged for each plot) had been recorded. Finally, in the contingency analysis were also included plots with species composition and development class information before the damage (23 767 plots). 3. Results Damage was detected in 3771 plots out of 34 263 studied plots during the period 1995–2014 (about 11%). In relation to the total plots available per forest type, birch forest was the most affected type, followed by mixed and spruce forests (17%, 10% and 8% of all plots, respectively). The highest disturbance occurrences corresponded to: browsing, snow and wind. By forest type, spruce forests were found to be more affected by snow, wind, fungus and browsing damage; pine forest was mainly affected by browsing, fungus and wind, birch forest by snow, insect attacks and browsing, and mixed forest by browsing, snow and wind. The time series showed some trends in the relative frequency of disturbance occurrences over time (Fig 2). At country level, the occurrence of snow and wind declined over time (linear trend, p<0.001 in both cases), although presenting a small increase in 2012–2014. Regionally, snow damage showed a similar trend in each of the 6 the regions (linear trend, p<0.001 for all regions), whereas wind showed a decreasing trend only in some regions (linear trend, p<0.001 only for regions 4, 5 and 6) as well as browsing (linear trend, p = 0.02 for region 6). Curves with peaks of frequencies were observed for browsing (significant for regions 3 and 5), and fungus (significant for region 6). In the case of snow and insect damage there were large occurrence pikes in region 6, corresponding to the years 1997, 2003 and 2013. 10.1371/journal.pone.0161361.g002Fig 2 Frequencies of damage occurrence by disturbance agent along 1995–2014. Top: Relative to country level (All). Bottom: Relative to each region. Each disturbance type showed a specific spatial distribution (Fig 3). Snow related damages were mainly located in the southern regions (region 1 and 2) and in the northern region (region 6), being specially clustered in mountainous areas. Wind related damage was more frequent on the western regions (region 4 and 5) and in mountainous areas facing the Atlantic side. Browsing damage was clearly aggregated in southeast regions (regions 1–3), forming two main hotspots. Insect related damage was most common in the northernmost of region 6, forming disperse hotspots. The occurrence of fungal damage was mainly present in the south of the country, with three main hotspots: in the coast, at the Swedish border and in the mountain areas of region 2. 10.1371/journal.pone.0161361.g003Fig 3 Damage occurrence spatial analysis. Spatial analysis of the locations with the highest concentration of damage occurrence (large maps) and average damage percentage by region (smaller maps) by disturbance agent during the period 1995–2014 (excluding Finnmark). For the large maps, darker colors represent higher concentrations of damaged plots, and the scale is relative for each map. For the small maps, the scale is defined in percentage of the total numbers of National Forest Inventory plots observed in the region. Map border lines adapted from [35], original licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). It must be taken into account that damage due to insect and fungus responds to different fungal agents that are specific to a species, and the maps represent only the general occurrence of fungal damages altogether. Therefore, the contingency tables aimed at observing the damage by both species composition and development classes (Table 3), in order to identify difference occurrence rates for different species or for different development stages related to the stand’s age. The calculated Chi-squared was 4117.6 (d.f. = 55) with a p-value < 0.001, presenting a highly significant departure from the null hypothesis, meaning that the different damage agents affect differently to different forest types and development classes. For instance, the observed frequency for snow damage was considerably higher from expected values in mature and intermediate birch forests. The occurrence of wind damage was higher than expected in mature spruce forest (regions 2, 3 and 5). The occurrence of browsing damage was higher than expected in young stands of all forest types except in spruce stands. In spruce stands, all stages of development appeared to be less susceptible to browsing damage, except during the period 2005–09 where young spruce stands followed the same trend as other young forest types. Mature and intermediate birch forest were more susceptible to insect damage than other forest types. None of the forest types was found more susceptible of fungus damages over the whole study period. Although, during the last years spruce seemed to be more prone to suffer from fungus attacks. The effect of each of the disturbances agents, on the different forest types, was also tested showing that individually, the damage agents also affect differently to different forest types and development classes. The Chi-squared results were 458.46 for snow, 222.82 for wind, 2339 for browsing, 726.57 for insect and 70.44 for fungus, (d.f. = 11 and p-value < 0.001 for all cases). 10.1371/journal.pone.0161361.t003Table 3 Contingency table presenting observed and calculated expected stand damage by disturbance agent. Observed/Predicted Development class Spp. Composition Snow Wind Browsing Insect Fungus No damage Mature Spruce 4644 5522 1374 2829 3727 20512034 Pine 549 1825 1184 333 3331 24472296 Birch 13236 2518 661 13224 723 15411681 Mixed 4440 3420 969 1627 1425 19471882 Intermediate Spruce 6058 5329 1298 1039 3636 27812692 Pine 342 1621 872 528 3726 20881967 Birch 8227 614 846 8918 1117 11821257 Mixed 6166 2633 16112 1344 3741 32313086 Young Spruce 925 113 6943 017 2916 11821177 Pine 210 05 10518 17 127 415488 Birch 815 07 11225 810 99 606678 Mixed 1452 026 42089 535 2933 22062439 Highlighted, those presenting a relevant deviation between observed and expected. Finally, when the level of damage per affected plot was evaluated, it was found that the disturbance agent causing more severe losses was browsing, and especially on those plots located in regions 1–4 (Fig 4). The mean percentage of damage losses caused by snow, wind and fungus damages were not significantly different between regions. Most of the losses due to insect damage were in region 6 where the species composition is typically dominated by birch. 10.1371/journal.pone.0161361.g004Fig 4 Average damage level observed per region (2000–2014) and disturbance agent. The dashed line represents the minimum damage value observed in the data (5%). The bands represent 2 x standard error associated to the mean. The damage is defined according to the National Forest Inventory instructions (see Table 2). 4. Discussion Long-term damage analysis enhances the understanding of natural disturbances in forest areas [30] and it can help forecast the probability of damage given different timeframes. Previous studies have demonstrated the reliability of NFI data to assess the occurrence of the most important disturbances and their impact on forest resources [21,23,44], making NFI data a good basis for a characterization of disturbances in the forest. Variables from NFI have also been used to better understand forest factors related to occurrence or damage of different natural disturbances [24,28,32]. The present study evaluates the temporal evolution and spatial aggregation patterns of disturbance occurrence across Norway, based on several consecutive NFI datasets entailing all productive forest and showing its state and development (stock, species composition, health) over a 20 year period. The availability of four NFI measurements provides an excellent opportunity to evaluate similarities and divergences between disturbances, considering their temporal and spatial aggregation patterns. When evaluating the evolution of disturbance occurrence, we observed clear variations between disturbance agents. The occurrence of snow and wind related damages appeared to peak in different years, but snow had a much higher peak between 1996–1997, and lower ones after this period. It was not possible to identify a clear fixed interval, which agrees with other European records where snow damaging events varied from each winter to longer cycles [45]. As expected, location seems to play a major role on defining the susceptibility to suffer from snow and wind damage [46]. As many other mountainous countries, Norway, presents a regionalized climate that can influence the spatial distribution of affected plots. Southern areas of the country are often influenced by southern weather systems whereas northern areas are influenced by oceanic clime, due to the proximity of the north Atlantic streams on the west. We found that the occurrence of snow and wind damages tended to be aggregated in mountainous areas, with no major differences along the north-south axis. Typically, northern trees are more adapted to this type of damage and also have a slower growth, allowing them to better adapt to wind and snow damage events [28]. When defining which forest types were more susceptible to be affected by wind and snow, our analysis showed that snow damage was more frequent on mature and intermediate birch forest, whereas wind affectation was associated to mature spruce, and both wind and snow damages were rare on pine forest. One factor that can explain this is the higher exposure of birch stands to the snow due to their location in mountainous areas. Another factor relates with structural parameters, e.g.: lower spacing between trees, compared to conifers. Our results partially disagree with previous studies indicating that snow and wind damage are more common in Scots pine and Norway spruce than birch [47,48]. However, for a similar diameter class, Scots pine would be more resistant to uprooting than spruce and birch stands, due to its deeper roots and better anchorage [47]. In addition to higher exposures due to location, one potential reason for the susceptibility of birch to snow damage could be that in Norway birch forest is usually shorter in height than spruce forest and their stems are easily bent or blow down by snow load or avalanches. Birch also has a tendency to regenerate through coppice in small groups of stems, generating gaps between regeneration groups that can ease the appearance of snow damage, for example due to avalanches [49]. In the case of spruce and wind damage, the higher susceptibility of the older stages of development was consistent with some studies [50] but disagreed with others [51]. It has to be mentioned that as the NFI data on wind damage focus on recording blown down trees, uprooting processes are overlooked in contrast to possible steam breakage, which is allocated as snow damage. These criteria in data recording can further explain our results, as larger specimens of spruce are more susceptible to uprooting [52] in contrast to smaller trees that often are more prone to break. In the case of biotic agents, it is important to mention that the Norwegian NFI focused on some specific agents, both in case of insects and pathogens. For instance, in the case of insects, priority was given to bark beetle (Ips typographus), European pine sawfly (Neodiprion sertifer), autumnal moth (Epirrita autumnata) and weevil (Curculionidae), each of them typically specialized in certain tree species. The joined analysis of different and specialized insect and fungus agents had an important effect on all the dimensions of the occurrence analysis, and was therefore further explored by species in the contingency analysis. For example, insect damage was especially visible in a region dominated by birch (region 6), as most of the insect related damage corresponds to autumnal moth attacks on birch. The temporal evolution of the attacks of insect frequency showed cyclic occurrence peaks and agrees with previous studies where autumnal moth typically present cyclical outbreaks every 10 years [53,54]. Our results also provided support to the hypothesis that autumnal moth on birch is more common in older stands [55,56], as mature birch stands offer more favorable places for oviposition and an enhancement of food resources [55]. Another interesting result is the lack of major bark beetle outbreaks on spruce forest, during the period 1995–2014. This result is probably associated to the limited storm related damage for the same period and the well-known interaction between storm damage and Ips outbreaks [57–59]. The observed trend of limited but highly variable presence of bark beetle damage, agrees with the trends identified for the period 1972–2002 by Økland and Bjønstad (2003) [60], when no important outbreak was detected the years following the major attack of the 70s [61,62]. Concerning fungal attacks, the records also corresponded with different and specialised agents: spruce needle rust fungus (Chrysomyxa abietis) on spruce and lophodermium needle cast pine (Lophodermium) on pine. The selection of agents had a clear effect on the allocation patterns of the recorded attacks, as the most affected regions corresponded to those were spruce was the dominant tree species. Regarding the temporal evolution of the fungi attacks, they correlated with humid springs [63], and in many cases warmer than average periods such as the 2002–2004 and specially 2013–2014. When the susceptibility of forest types to fungi was analysed, no apparent variations were identified but during the last years analysed, where an increase on attack susceptibility was observed on young or mature spruce stands (S1 Table). Browsing was one of the most relevant disturbance agents associated to Norwegian forests, both on occurrence and impact on the forest. Browsing frequency remained similar across time and the damage was allocated mainly on the most productive forests in the south of the country. Moose habitat selection is led by forage quality and shelter availability, both aspects changing in space and time [64]. Browsing affected mainly young stands, as typically moose prefer those because they offer more palatable forage at a reachable height [32]. The species composition of the stands also had an important effect on the allocation of browsing damage, due to specific food choices and preferences by ungulates [9,32]. Our study shows the preference by ungulates of mixed, birch and pine stands, except for the period 2005–2009 when spruce was also significantly damaged (S1 Table). Typically, moose prefer spruce stands the least [32], although it is also known that increased densities of moose population might also increase the risk of browsing damage (e.g. stands on migration routes or next to winter habitats). An important data limitation was that the 7th and 8th NFI (1995–2005) did not include damage on forests in regeneration stage. It would be reasonable to assume that the occurrence of browsing on those young stands would be even higher, as moose is known to browse in mixed stands with young pines (age<5 years) [65]. Despite the limitations, the Norwegian NFI has shown its potential as data source for assessing natural disturbances, their spatial and temporal distribution and the vulnerability of different forest types to them. We consider that the combination of four different NFI measurements provides a unique opportunity for this type of analysis. Our analysis suggests that a clear relation exists between wind and snow damage, which is supported by the recognized combined effect of wind intensity and snow load to induce storm related damage [28,51,66]. However, the instructions of the Norwegian NFI lead to consider stem breakage as snow derived damage (and never as a wind related damage) which could induce to misinterpretations. In this study we did not consider interactions among different disturbance agents due to the complexity of the analysis required and because of the absence of major abiotic disturbances that might trigger potential future damage by biotic agents [49,67–69]. The combination of different disturbance types within the same measurement was not considered, as the main disturbance was the only registered damage in over 85% of the damaged plots (S2 Table). Finally, not finding major disturbances threating Norwegian forests is by itself the main result of the study. After the catastrophic storms and bark beetle attacks of the 70s [62,70], we may say that the Norwegian forest has sustained a relatively long period of good health, either for the absence of extreme storms, or due to the effects that past storms and biotic attacks had on the temporal depletion of susceptible trees [71]. This assumption calls for further analysis such as predictions of the forest evolution to detect potential threats in the future due to the natural ageing of the forest, or the inclusion of structural or management dependent variables, aiming at identifying past management actions that may help mitigate the impact of each disturbance. For example, the inclusion of more detailed information about snow loads and stand management history would also help define the importance of variations in the stand and trees’ structure on storm related damage. The study of forest disturbances is an important source of information for developing a holistic management of forests and its associated ecosystem services [4]. The analysis of records of damage by different causing agents, with spatial approaches or with descriptors of forest stands can therefore help identify susceptible forest areas and at the same time, forest management practices that can reduce their probability. Supporting Information S1 Table Contingency tables. Tables presenting observed (numerator) and the calculated expected (denominator) stand damage by development class and disturbance agent and Norwegian National Forest Inventory (NFI). (DOCX) Click here for additional data file. S2 Table Disturbance agents simultaneously recorded in the same measurement. Percentage of the plots damaged by the main disturbance agent that were simultaneously damaged by a secondary and tertiary disturbance agent. (DOCX) Click here for additional data file. ==== Refs References 1 Oliver CD , Larson BC . Forest stand dynamics . 1st ed New York ; 1996 . 2 Bond WJ , Keeley JE . Fire as a global “herbivore”: the ecology and evolution of flammable ecosystems . Trends in Ecology & Evolution. Elsevier ; 2005 ;20 : 387 –394 . 10.1016/j.tree.2005.04.025 3 Schelhaas MJ , Nabuurs GJ , Schuck A . Natural disturbances in the European forests in the 19th and 20th centuries . Global Change Biology . 2003 ;9 : 1620 –1633 . 10.1046/j.1365-2486.2003.00684.x 4 Thom D , Seidl R . Natural disturbance impacts on ecosystem services and biodiversity in temperate and boreal forests . Biol Rev . 2015 10.1111/brv.12193 5 Bergeron Y , Gauthier S , Kafka V , Lefort P , Lesieur D . Natural fire frequency for the eastern Canadian boreal forest: consequences for sustainable forestry . Can J For Res . 2001 ;31 : 384 –391 . 10.1139/cjfr-31-3-384 6 Franklin JF , Spies TA , Van Pelt R , Carey AB , Thornburgh DA , Berg DR , et al Disturbances and structural development of natural forest ecosystems with silvicultural implications, using Douglas-fir forests as an example . Forest Ecology and Management . 2002 ;155 : 399 –423 . 10.1016/S0378-1127(01)00575-8 7 Gadow von K . Evaluating risk in forest planning models . Silva Fennica . 2000 ;34 : 181 –191 . 8 González-Olabarria JR , Pukkala T . Integrating fire risk considerations in landscape-level forest planning . Forest Ecology and Management . 2011 ;261 : 278 –287 . 10.1016/j.foreco.2010.10.017 9 Seidl R , Fernandes PM , Fonseca TF , Gillet F , Jönsson AM , Merganičová K , et al Modelling natural disturbances in forest ecosystems: a review . Ecological Modelling. Elsevier B.V ; 2011 ;222 : 903 –924 . 10.1016/j.ecolmodel.2010.09.040 10 Johnson EA , Gutsell SL . Fire frequency models, methods and interpretations . Advances in Ecological Research. Elsevier ; 1994 ;25 : 239 –287 . 10.1016/S0065-2504(08)60216-0 11 Runkle JR . Disturbance regimes in temperate forests In: Pickett STA , White PS , editors. The ecology of natural disturbance and patch dynamics . London ; 1985 pp. 17 –33 . 12 Fischer A , Marshall P , Camp A . Disturbances in deciduous temperate forest ecosystems of the northern hemisphere: their effects on both recent and future forest development Biodivers Conserv . Springer Netherlands ; 2013 ;22 : 1863 –1893 . 10.1007/s10531-013-0525-1 13 Bond WJ , Midgley JJ . Kill thy neighbour: an individualistic argument for the evolution of flammability . Oikos . 1995 ;73 : 79 10.2307/3545728 14 Hall FG , Botkin DB , Strebel DE , Woods KD , Goetz SJ . Large-scale patterns of forest succession as determined by remote sensing . Ecology . 1991 ;72 : 628 –640 . 15 Franklin J . Predictive vegetation mapping: geographic modelling of biospatial patterns in relation to environmental gradients Progress in Physical Geography . SAGE Publications ; 1995 ;19 : 474 –499 . 10.1177/030913339501900403 16 Chuvieco E . Remote sensing of large wildfires Chuvieco E , editor. Berlin, Heidelberg : Springer Science & Business Media ; 2012 10.1007/978-3-642-60164-4 17 Frolking S , Palace MW , Clark DB , Chambers JQ , Shugart HH , Hurtt GC . Forest disturbance and recovery: A general review in the context of spaceborne remote sensing of impacts on aboveground biomass and canopy structure . J Geophys Res . 2009 ;114 : n/a –n/a . 10.1029/2008JG000911 18 Allen E . Forest health assessment in Canada . Ecosystem Health . 2001 ;7 : 28 –34 . 19 Wulff S , Roberge C , Ringvall A , Holm S , Ståhl G . On the possibility to monitor and assess forest damage within large scale monitoring programmes—a simulation study . Silva Fennica . 2013 ;47 10.14214/sf.1000 20 Bergstedt J , Milberg P . The impact of logging intensity on field-layer vegetation in Swedish boreal forests . Forest Ecology and Management . 2001 ;154 : 105 –115 . 10.1016/S0378-1127(00)00642-3 21 Nevalainen S . Gremmeniella abietina in Finnish Pinus sylvestris stands in 1986–1992: A study based on the national forest inventory . Scandinavian Journal of Forest Research . 1999 ;14 : 111 –120 . 10.1080/02827589950152836 22 Nevalainen S , Yli-Kojola H . Extent of abiotic damage and its relation to defoliation of conifers in Finland . Forest Ecology and Management . 2000 ;135 : 229 –235 . 10.1016/S0378-1127(00)00313-3 23 Nevalainen S , Sirkiä S , Peltoniemi M , Neuvonen S . Vulnerability to pine sawfly damage decreases with site fertility but the opposite is true with Scleroderris canker damage; results from Finnish ICP Forests and NFI data . Ann For Sci . 2015 ;72 : 909 –917 . 10.1007/s13595-014-0435-8 24 González JR , Palahí M , Trasobares A , Pukkala T . A fire probability model for forest stands in Catalonia (north-east Spain) . Ann For Sci . 2006 ;63 : 169 –176 . 10.1051/forest:2005109 25 González JR , Trasobares A , Palahí M , Pukkala T . Predicting stand damage and tree survival in burned forests in Catalonia (North-East Spain) . Ann For Sci . 2007 ;64 : 733 –742 . 10.1051/forest:2007053 26 Garcia-Gonzalo J , Pukkala T , Borges JG . Integrating fire risk in stand management scheduling. An application to Maritime pine stands in Portugal . Ann Oper Res. Springer US ; 2011 ;219 : 379 –395 . 10.1007/s10479-011-0908-1 27 Marques S , Garcia-Gonzalo J , Botequim B , Ricardo A , Borges JG , Tome M , et al Asesing wild fire ocurence probability in Pinus pinaster Ait. stands in Portugal . Forest Syst . 2012 ;21 : 111 –120 . 28 Valinger E , Fridman J . Models to assess the risk of snow and wind damage in pine, spruce, and birch forests in sweden . Environmental Management . 1999 ;24 : 209 –217 . 10384030 29 Dobbertin M . Influence of stand structure and site factors on wind damage comparing the storms Vivian and Lothar . For Snow Landsc Res . 2002 ;77 : 187 –205 . 30 Hanewinkel M , Breidenbach J , Neeff T , Kublin E . Seventy-seven years of natural disturbances in a mountain forest area—the influence of storm, snow, and insect damage analysed with a long-term time series . Can J For Res . 2008 ;38 : 2249 –2261 . 10.1139/X08-070 31 Martín-Alcón S , González-Olabarria JR , Coll L . Wind and snow damage in the Pyrenees pine forests: effect of stand attributes and location . Silva Fennica . 2010 ;44 : 399 –410 . 32 Jalkanen A . The probability of moose damage at the stand level in southern Finland . Silva Fennica . 2001 ;35 : 159 –168 . 33 Vospernik S . Probability of bark stripping damage by red deer (Cervus elaphus) in Austria . Silva Fennica . 2006 ;40 : 589 –601 . 34 Landsskogtakseringens. Landsskogtakseringens feltinstruks 2008. 2008; 1–153. 35 Kartverket. N250, Norway; 2015. Available: www.kartverket.no. 36 Worton BJ . Kernel methods for estimating the utilization distribution in home-range studies . Ecology . 1989 ;70 : 164 10.2307/1938423 37 Seaman DE , Powell RA . An evaluation of the accuracy of kernel density estimators for home range analysis . Ecology . 1996 ;77 : 2075 10.2307/2265701 38 Abramson IS . On bandwidth variation in kernel estimates-a square root law . The Annals of Statistics . 1982 . 39 Koutsias N , Kalabokidis KD , Allgöwer B . Fire occurrence patterns at landscape level: Beyond positional accuracy of ignition points with kernel density estimation methods . Natural Resource Modeling . 2008 ;17 : 359 –375 . 10.1111/j.1939-7445.2004.tb00141.x 40 la Riva de J , Pérez-Cabello F , Lana-Renault N , Koutsias N . Mapping wildfire occurrence at regional scale . Remote Sensing of Environment . 2004 ;92 : 288 –294 . 10.1016/j.rse.2004.06.013 41 Terrell GR . The maximal smoothing principle in density estimation . Journal of the American Statistical Association . 1990 ;85 : 470 –477 . 10.1080/01621459.1990.10476223 42 Bowman AW , Azzalini A . Applied smoothing techniques for data analysis: The Kernel approach with S-Plus illustrations . Oxford ; 1997 . 43 Davies TM , Hazelton ML , Marshall JC . sparr: Analyzing Spatial Relative Risk Using Fixed and Adaptive Kernel Density Estimation in R . Journal of Statistical Software . 2011 ;39 : 1 –14 . 10.18637/jss.v039.i01 44 Wulff S , Hansson P , Witzell J . The applicability of national forest inventories for estimating forest damage outbreaks—Experiences from a Gremmeniella outbreak in Sweden . Can J For Res . 2006 ;36 : 2605 –2613 . 10.1139/x06-148 45 Nykänen ML , Peltola H , Quine C , Kellomäki S . Factors affecting snow damage of trees with particular reference to European conditions . Silva Fennica . 1997 ;31 : 192 –213 . 46 Valinger E , Pettersson N . Wind and snow damage in a thinning and fertilization experiment in Picea abies in southern Sweden . Forestry . 1996 ;69 : 25 –33 . 47 Peltola H , Nykänen ML , Kellomäki S . Model computations on the critical combination of snow loading and windspeed for snow damage of Scots pine, Norway spruce and birch sp. at stand edge . Forest Ecology and Management . 1997 10.1016/S0378-1127(97)00037-6 48 Jalkanen R , Konocpka B . Snow-packing as a potential harmful factor on Picea abies, Pinus sylvestris and Betula pubescens at high altitude in northern Finland . European Journal of Forest Pathology. Blackwell Publishing Ltd ; 1998 ;28 : 373 –382 . 10.1111/j.1439-0329.1998.tb01191.x 49 Bebi P , Kulakowski D , Rixen C . Snow avalanche disturbances in forest ecosystems—State of research and implications for management . Forest Ecology and Management. Elsevier ; 2009 ;257 : 1883 –1892 . 50 Jalkanen A , Mattila U . Logistic regression models for wind and snow damage in northern Finland based on the National Forest Inventory data . Forest Ecology and Management. Elsevier ; 2000 ;135 : 315 –330 . 51 Päätalo M-L . Risk of Snow Damage in Unmanaged and Managed Stands of Scots Pine, Norway Spruce and Birch . Scandinavian Journal of Forest Research . 2000 ;15 : 530 –541 . 10.1080/028275800750173474 52 Ilisson T , Metslaid M , Vodde F , Jõgiste K , Kurm M . Storm disturbance in forest ecosystems in Estonia . Scandinavian Journal of Forest Research . 2005 ;20 : 88 –93 . 10.1080/14004080510041020 53 Tenow O , Nilssen AC , Bylund H , Hogstad O . Waves and synchrony in Epirrita autumnata /Operophtera brumata outbreaks. I. Lagged synchrony: regionally, locally and among species . J Anim Ecology . 2007 ;76 : 258 –268 . 10.1111/j.1365-2656.2006.01204.x 54 Ruohomäki K , Tanhuanpaa M , Ayres MP , Kaitaniemi P , Tammaru T , Haukioja E . Causes of cyclicity of Epirrita autumnata (Lepidoptera, Geometridae): grandiose theory and tedious practice . Population Ecology . 2000 ;42 : 211 –223 . 55 Bylund H . Stand age-structure influence in a low population peak of Epirrita autumnata in a mountain birch forest . 1997 ;: 319 –326 . 56 Ruohomäki K , Virtanen T , Kaitaniemi P , Tammaru T . Old Mountain Birches at High Altitudes Are Prone to Outbreaks of Epirrita autumnata (Lepidoptera: Geometridae) . Population Ecology . 1997 ;26 : 1096 –1104 . 57 Økland B , Bjørnstad ON . A resource-depletion model of forest insect outbreaks . Ecology . 2006 ;87 : 283 –290 . 10.1890/05-0135 16637352 58 Stadelmann G , Bugmann H , Wermelinger B , Meier F , Bigler C . A predictive framework to assess spatio-temporal variability of infestations by the European spruce bark beetle . Ecography. Blackwell Publishing Ltd ; 2013 ;36 : 1208 –1217 . 10.1111/j.1600-0587.2013.00177.x 59 Stadelmann G , Bugmann H , Wermelinger B , Bigler C . Spatial interactions between storm damage and subsequent infestations by the European spruce bark beetle . Forest Ecology and Management. Elsevier B.V ; 2014 ;318 : 167 –174 . 10.1016/j.foreco.2014.01.022 60 Økland B , Bjørnstad ON . Synchrony and geographical variation of the spruce bark beetle (Ips typographus) during a non-epidemic period . Population Ecology . 2003 ;45 : 213 –219 . 10.1007/s10144-003-0157-5 61 Bakke A . Host tree and bark beetle interaction during a mass outbreak of Ips typographus in Norway . Zeitschrift für angewandte Entomologie . 1983 10.1111/j.1439-0418.1983.tb03651.x 62 Bakke A . The recent Ips typographus outbreak in Norway: Experiences from a control program . Holarctic ecology . 1989 . 63 Collins MA . Periodicity of spore liberation in Chrysomyxa abietis . Transactions of the British Mycological Society. British Mycological Society ; 1976 ;67 : 336 –339 . 10.1016/S0007-1536(76)80143-X 64 Bjørneraas K , Solberg EJ , Herfindal I , Moorter BV , Rolandsen CM , Tremblay J-P , et al Moose Alces alces habitat use at multiple temporal scales in a human-altered landscape . Wildlife Biology . 2011 ;17 : 44 –54 . 10.2981/10-073 65 Heikkilä R , Härkönen S . Moose browsing in young Scots pine stands in relation to forest management . Forest Ecology and Management . 1996 ;88 : 179 –186 . 10.1016/S0378-1127(96)03823-6 66 Fridman J , Valinger E . Modelling probability of snow and wind damage using tree, stand, and site characteristics from Pinus sylvestris sample plots . Scandinavian Journal of Forest Research . 1998 ;13 : 348 –356 . 10.1080/02827589809382994 67 Netherer S , Nopp-Mayr U . Predisposition assessment systems (PAS) as supportive tools in forest management—rating of site and stand-related hazards of bark beetle infestation in the High Tatra Mountains as an example for system application and verification . Forest Ecology and Management . 2005 ;207 : 99 –107 . 10.1016/j.foreco.2004.10.020 68 Christiansen E , Bakke A . Does drought really enhance Ips typographus epidemics? A Scandinavian perspective . In: Gregoire JC , Liebhold AM , Stephen FM , Day KR , Salom SM , editors. 1997 pp. 163 –171 . 69 Wermelinger B . Ecology and management of the spruce bark beetle Ips typographus—a review of recent research . Forest Ecology and Management . 2004 ;202 : 67 –82 . 10.1016/j.foreco.2004.07.018 70 Økland B , Berryman A . Resource dynamic plays a key role in regional fluctuations of the spruce bark beetles Ips typographus . 2004 ;6 : 141 –146 . 71 Wermelinger B , Obrist MK , Baur H , Jakoby O , Duelli P . Synchronous rise and fall of bark beetle and parasitoid populations in windthrow areas . Agr Forest Entomol. Blackwell Publishing Ltd ; 2013 ;15 : 301 –309 . 10.1111/afe.12018

PMC005xxxxxx/PMC5003384.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757127210.1371/journal.pone.0162148PONE-D-16-29712Research ArticleBiology and Life SciencesImmunologyImmune System ProteinsImmune ReceptorsToll-like ReceptorsMedicine and Health SciencesImmunologyImmune System ProteinsImmune ReceptorsToll-like ReceptorsBiology and Life SciencesBiochemistryProteinsImmune System ProteinsImmune ReceptorsToll-like ReceptorsBiology and Life SciencesCell BiologySignal TransductionImmune ReceptorsToll-like ReceptorsBiology and Life SciencesImmunologyVaccination and ImmunizationVaccinesMedicine and Health SciencesImmunologyVaccination and ImmunizationVaccinesMedicine and Health SciencesPublic and Occupational HealthPreventive MedicineVaccination and ImmunizationVaccinesBiology and Life SciencesDevelopmental BiologyNeonatesBiology and Life SciencesBiochemistryProteinsInterferonsBiology and life sciencesCell biologyCellular typesAnimal cellsBlood cellsWhite blood cellsNK cellsBiology and life sciencesCell biologyCellular typesAnimal cellsImmune cellsWhite blood cellsNK cellsBiology and life sciencesImmunologyImmune cellsWhite blood cellsNK cellsMedicine and health sciencesImmunologyImmune cellsWhite blood cellsNK cellsBiology and Life SciencesCell BiologyCellular TypesAnimal CellsBlood CellsWhite Blood CellsT CellsBiology and Life SciencesCell BiologyCellular TypesAnimal CellsImmune CellsWhite Blood CellsT CellsBiology and Life SciencesImmunologyImmune CellsWhite Blood CellsT CellsMedicine and Health SciencesImmunologyImmune CellsWhite Blood CellsT CellsBiology and Life SciencesGeneticsGene ExpressionBiology and Life SciencesCell BiologyCellular TypesAnimal CellsImmune CellsAntigen-Presenting CellsDendritic CellsBiology and Life SciencesImmunologyImmune CellsAntigen-Presenting CellsDendritic CellsMedicine and Health SciencesImmunologyImmune CellsAntigen-Presenting CellsDendritic CellsA Model to Explain How the Bacille Calmette Guérin (BCG) Vaccine Drives Interleukin-12 Production in Neonates BCG Vaccine Induction of IL-12 in NeonatesKativhu Chido Loveness Libraty Daniel H. *Division of Infectious Diseases and Immunology, Department of Medicine, University of Massachusetts Medical School, Worcester, Massachusetts, United States of AmericaKollmann Tobias R EditorUniversity of British Columbia, CANADACompeting Interests: The authors have declared that no competing interests exist. Conceptualization: DHL. Formal analysis: CLK DHL. Funding acquisition: DHL. Investigation: CLK. Methodology: CLK DHL. Validation: CLK. Writing – original draft: CLK. Writing – review & editing: CLK DHL. * E-mail: daniel.libraty@umassmed.edu29 8 2016 2016 11 8 e016214825 7 2016 17 8 2016 © 2016 Kativhu, Libraty2016Kativhu, LibratyThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.The Bacille Calmette Guérin (BCG) vaccine is the only routine vaccination at birth that effectively induces neonatal T-helper 1 (Th1)-polarized immune responses. The primary cytokine that drives CD4+ T-cell Th1 differentiation is interleukin (IL)-12 p70, a heterodimeric cytokine composed of the IL-12 p35 and IL-12 p40 subunits. We therefore examined the mechanisms involved in BCG vaccine stimulation of IL-12 p35 and p40 production from human umbilical cord (neonatal) cells. We found that BCG bacilli did not upregulate IL-12 p35 mRNA production, but upregulated IL-12 p40 mRNA production in a Toll-like receptor (TLR)2-dependent manner, in human neonatal monocyte-derived dendritic cells (mdDCs). The combination of TLR2 signaling, Type I interferon (IFN), and Type II IFN induced maximal levels of IL-12 p35 and p40 mRNA production in human neonatal mdDCs. The cell-free supernatants of reconstituted BCG vaccine vials contained extracellular mycobacterial (BCG) DNA which could induce IFN-α (Type I IFN) production in human neonatal plasmacytoid dendritic cells (pDCs). BCG bacilli also stimulated human neonatal CD16lo natural killer (NK) cells to produce IFN-γ (Type II IFN) in a TLR2-dependent manner. We have therefore proposed a model where BCG vaccine could stimulate the combination of neonatal conventional DCs (cDCs), pDCs, and CD16lo NK cells to produce optimal neonatal IL-12 p35 and p40 (IL-12 p70) production and subsequent CD4+ T-cell Th1 polarization. An adjuvant that emulates the mechanism by which the BCG vaccine stimulates neonatal IL-12 p35 and p40 production could improve vaccine strategies at birth for protection against intracellular pathogens and toxins. http://dx.doi.org/10.13039/100000060National Institute of Allergy and Infectious DiseasesR01AI091820Libraty Daniel H. This work was supported by a grant from the U.S. National Institutes of Health R01AI091820. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper.Data Availability All relevant data are within the paper. ==== Body Introduction The Bacille Calmette Guérin (BCG) vaccine is given to neonates in most countries to prevent infantile tuberculous meningitis and miliary disease. It is one of the most widely used vaccines worldwide. The efficacy of neonatal BCG administration has been linked to its ability to effectively induce anti-mycobacterial CD4+ T-cell T-helper 1 (Th1)-polarized neonatal immune responses [1–5]. Neonatal BCG vaccination has also been reported to reduce neonatal and infant mortality due to diseases other than tuberculosis [6–8]. The all-cause mortality benefit of neonatal BCG vaccination may be partially related to its ability to also induce heterologous Th1-polarizing immune responses during the neonatal period [9–12]. Th1 responses are characterized by CD4+ T-cell interferon (IFN)-γ production. The primary cytokine that drives CD4+ T-cell Th1 differentiation is interleukin (IL)-12 p70, a heterodimeric cytokine composed of the IL-12 p35 and IL-12 p40 subunits [13, 14]. Neonates and infants generally have reduced IL-12 p70 production and CD4+ T-cell Th1 responses to intracellular pathogens and toxins [15–18]. The fetal and early neonatal immune system is heavily Th2 and Th17-biased [15, 17, 19, 20]. Enhanced neonatal Th1-polarized immune responses would be beneficial for combating infections with intracellular pathogens and toxin-producing organisms. The BCG vaccine is the only routine vaccination that can be given at birth and induce Th1-polarized immune responses in neonates [1–4]. Previous reports have shown that BCG vaccination at birth results in neonatal IFN-γ production against mycobacterial antigens [1], and the levels of secreted IFN-γ are comparable to adult levels [2]. We and others have previously identified some of the heterologous immune effects of BCG vaccination, including Th1 polarization [5, 10, 11, 21]. IL-12 p70 is the prototypical Th1-polarizing cytokine [14]. We therefore examined the BCG vaccine stimulation of IL-12 p70 production from human umbilical cord (neonatal) cells by analyzing the induction of each subunit (p35 and p40). We found that BCG bacilli did not upregulate IL-12 p35 mRNA production, but upregulated IL-12 p40 mRNA production in a Toll-like receptor (TLR)2-dependent manner, in human neonatal monocyte-derived dendritic cells (mdDCs). IL-12 p35 gene transcription was previously shown to be repressed in human neonatal mdDCs [22, 23]. Goriely et al. [22] demonstrated that lipopolysaccharide (LPS) + IFN-γ (Type II IFN) stimulation of neonatal mdDCs was able to induce IL-12 p35 and p40 mRNA production that approached adult levels. LPS is a TLR4 ligand that stimulates both NF-κB activation and the Type I IFN pathway [24]. We found that the combination of a synthetic TLR2 agonist (NF-κB activation only), Type I IFN, and Type II IFN induced the maximal levels of IL-12 p35 and IL-12 p40 mRNA production in human neonatal mdDCs. We hypothesized that the BCG vaccine uses the same signaling combination, utilizing multiple innate immune cells, in order to produce adult-like levels of IL-12 p35 and p40 (IL-12 p70) in neonates. We found that the cell-free supernatants of reconstituted BCG vaccine vials contained extracellular mycobacterial (BCG) DNA which could induce IFN-α (Type I IFN) production in human neonatal plasmacytoid dendritic cells (pDCs). We also found that BCG bacilli stimulated human neonatal CD16lo natural killer (NK) cells to produce IFN-γ (Type II IFN) in a TLR2-dependent manner. We have therefore proposed a model where BCG vaccine could stimulate the combination of neonatal conventional DCs (cDCs), pDCs, and CD16lo NK cells thereby producing optimal neonatal IL-12 p35 and p40 production (IL-12 p70) and subsequent CD4+ T-cell Th1 polarization. Methods Ethics Statement Human umbilical cord blood was collected from the full-term placentas of healthy mothers with uncomplicated deliveries at the University of Massachusetts Memorial Medical Center Labor and Delivery Ward. The clinical study protocol was approved by the University of Massachusetts Medical School (UMMS) institutional review board (IRB). Umbilical cord blood was collected after verbal consent was obtained. Verbal consent was approved by the UMMS IRB, as obtaining written consent was not feasible on the Labor and Delivery Ward. Subject’s verbal consent was recorded on a datasheet. Drugs, reagents, and antibodies The recombinant (r) cytokines IL-2, IL-3, IL-4 and GM-CSF were obtained from Peprotech Inc., and used at final concentrations of 100 U/ml (rIL-2), 1000 U/ml (rIL-3), 100 U/ml (rIL-4) and 800 U/ml (rGM-CSF). The monoclonal antibodies (mAbs) against human TLR2 (clone #383936) and the IgG2B isotype control (clone #20116) were obtained from R&D Systems, and used at a final concentration of 20 μg/ml. Unless otherwise stated, all mAbs for FACS staining were obtained from BD Biosciences. For cell stimulations (mdDCs, pDCs, and NK cells), the Tice BCG vaccine (Merck) was reconstituted in sterile water, per the manufacturer’s instructions, and used at a final concentration of 8x104 colony forming units (cfu)/ml. For the pDC experiments, the synthetic TLR2 agonist Pam3CSK4 (Invivogen), was used at a concentration of 300 ng/ml; recombinant human IFN-β (PBL Assay Science) was used at a concentration of 1x104 U/ml; and recombinant human IFN-γ (PBL Assay Science) was used at a concentration of 1x103 U/ml. Cells Cord blood mononuclear cells (CBMC) were isolated by density gradient centrifugation (Histopaque™, Sigma-Aldrich) and cryopreserved within 24 hours post-delivery. To generate mdDCs, CBMC CD14+ monocytes were negatively selected by MACS® using the Monocyte Isolation kit II (Miltenyi Biotec). The monocytes were then plated in 24-well plates at 1.7x106 cells/well, and cultured in RPMI 1640/10% FCS/rIL-4/rGM-CSF for 7 days. Differentiation into immature mdDCs (lineage-CD1c+CD83lo) was ≥ 70%, as confirmed by flow cytometry. pDCs were isolated from CBMC by MACS® negative selection using the Plasmacytoid Dendritic Cell Isolation Kit II (Miltenyi Biotec), and cultured in RPMI 1640/10% FCS/rIL-3 (>95% purity). NK cells were negatively selected by MACS® using the NK Cell Isolation Kit Human (Miltenyi Biotec), and cultured in RPMI 1640/10% FCS/rIL-2 (>95% purity). BCG stimulation of mdDCs Immature mdDCs were left unstimulated, or stimulated with BCG x 18–24 hours ± pre-incubation for 1 hour with a TLR2 blocking mAb or an isotype control mAb. An approximate multiplicity of infection (MOI) = 5 was used. The cells were then lysed and cytosolic RNA was extracted using RNeasy Plus (Qiagen), according to the manufacturer’s instructions. Relative gene expression for IL-12 p35 and IL-12 p40 mRNA was determined by quantitative (q)RT-qPCR. GAPDH was used as the housekeeping gene. BCG DNA stimulation of pDCs BCG DNA was extracted from a BCG culture (kindly provided by Christopher Sassetti, UMMS). pDCs were stimulated with 100 μg/ml BCG DNA for 8 hours; cell lysates were collected for IFN-α1 and IFN-α2 gene expression analysis using qRT-PCR. β-actin was used as the housekeeping gene. In some experiments, pDCs were stimulated with intact BCG bacilli (Tice strain) in a similar fashion as described for mdDCs. All qRT-qPCR assays were performed using the TaqMan Gene Expression assay system (Applied Biosystems), following the manufacturer’s instructions. Analysis was performed using ABI 7500 Software v2.0 (Applied Biosystems). Identifying extracellular BCG DNA in BCG vaccine vials BCG vaccine vials were obtained from two different manufacturers- Tice BCG (Pasteur strain) from Merck, USA, and Tokyo BCG (Japan strain) from the Japan BCG Laboratory, Japan. The BCG vaccine vials were reconstituted in sterile water, per the manufacturer’s instructions, and spun down to collect the cell-free supernatants for PCR. PCR for 2 mycobacterial genes, rpoB and the RD8 portion in BCG, was performed as described in [25]. Flow cytometry Cord blood NK cells were isolated by MACS®, and seeded at a density of 1x106 cells/tube. The cells were stimulated with BCG overnight, incubated with Brefeldin-A (1 μg/ml, Qiagen) for the last 4 hours, stained with a vital dye (L/D Aqua), fixed and permeabilized (Cytofix/Cytoperm, BD Biosciences), and then stained with the mouse anti-human mAbs to CD3 (clone SK7), CD16 (clone 3G8), CD56 (clone NCAM16.2), and IFN-γ (clone 25723.11). A minimum of 50,000 events were acquired. To test if IFN-γ production was dependent on TLR2 signaling, isolated NK cells were pre-incubated for 1 hour with either anti-TLR2 antibody or IgG isotype control (R&D Systems) prior to stimulation with BCG. Events were acquired on an LSRII flow cytometer (Becton Dickinson), and analyzed using FlowJo™ v10.1 software (FlowJo LLC). Statistical analysis Statistical analysis was done using Prism 7 software (GraphPad Prism 7). Comparisons between paired groups were made using the non-parametric Wilcoxon's signed-rank test. P-values < 0.05 were considered significant. Results BCG does not induce IL-12 p35 mRNA production, but induces IL-12 p40 mRNA production in a TLR2-dependent manner, from human neonatal mdDCs It has been well established that the BCG vaccine is able to produce anti-mycobacterial neonatal CD4+ T-cell Th1 responses in vivo [1–5]. The prototypical Th1-polarizing cytokine is IL-12 p70 [14]. IL-12 p70 is a heterodimeric cytokine made up of the p35 and p40 subunits. We therefore examined BCG stimulation of IL-12 p35 and p40 mRNA production in human neonatal mdDCs. Human umbilical cord blood CD14+ monocytes were differentiated into mdDCs using IL-4 and GM-CSF. We found that BCG stimulation did not upregulate IL-12 p35 mRNA production in human neonatal mdDCs (Fig 1A). However, BCG upregulated IL-12 p40 mRNA production in a TLR2-dependent manner (Fig 1B). The addition of TLR4 blockade to TLR2 blocking did not further diminish IL-12 p40 mRNA production (data not shown). 10.1371/journal.pone.0162148.g001Fig 1 Bacille Calmette Guérin (BCG) does not induce IL-12 p35 mRNA production from human neonatal monocyte-derived dendritic cells (mdDCs), but induces IL-12 p40 mRNA production in a Toll-like receptor (TLR)2-dependent manner. Human umbilical cord blood CD14+ monocytes were differentiated into mdDCs using rIL-4 and rGM-CSF, stimulated with BCG x 18–24 hours, and then cellular mRNA was isolated for qRT-PCR. (a) relative expression of IL-12 p35 mRNA levels upon BCG stimulation compared to unstimulated control. Bars are median values, ‡ p = 0.3, n = 6 independent experiments; (b) relative expression of IL-12 p40 mRNA levels upon BCG stimulation compared to unstimulated control, in the presence or absence of a TLR2 blocking antibody. Bars are median values, * p<0.05 compared to BCG stimulation with isotype control antibody pre-incubation, n = 6 independent experiments. The combination of a synthetic TLR2 agonist (Pam3CSK4), rIFN-β (Type I IFN), and rIFN-γ (Type II IFN) stimulates maximal levels of IL-12 p35 and IL-12 p40 mRNA production in human neonatal mdDCs Human neonatal mdDCs have repressed IL-12 p35 gene expression compared to adult cells [22]. The transcriptional repression of IL-12 p35 in human neonatal mdDCs takes place at the chromatin level [23]. We interpreted data shown by Goriely et al. [22], that LPS + IFN-γ could increase IL-12 p35 gene expression in human neonatal mdDCs in similar fashion to adult mdDCs, to mean that a combination of NF-κB activation, Type I IFN signaling, and Type II IFN priming was needed for optimal neonatal IL-12 p35 production. We therefore primed neonatal mdDCs for 12 h with IFN-β and IFN-γ prior to stimulating them with a synthetic TLR2 agonist, Pam3CSK4. We observed that the combination of Pam3CSK4 + rIFN-β + rIFN-γ induced the maximal levels of IL-12 p35 mRNA (Fig 2A) and IL-12 p40 mRNA production (Fig 2B) from human neonatal mdDCs, and the mRNA relative expression levels were normally distributed (SPSS Statistics v24.0). 10.1371/journal.pone.0162148.g002Fig 2 Optimization of interleukin (IL)-12 p35 and p40 mRNA expression in human neonatal monocyte-derived dendritic cells (mdDCs). Neonatal mdDCs were primed with recombinant (r) interferon-β (IFN-β) (Type I IFN), and rIFN-γ (Type II IFN) for 12 hours prior to stimulating them with a synthetic TLR2 agonist, Pam3CSK4. The combination of Pam3CSK4, rIFN-β, and rIFN-γ induced the maximal levels of (a) interleukin (IL)-12 p35 mRNA and (b) IL-12 p40 mRNA expression in human neonatal monocyte-derived dendritic cells (mdDCs). * p<0.05 compared to unstimulated control. Data points represent individual donors, the number of individual donors for each condition is shown in the figure, and bars are mean values. The BCG vaccine contains extracellular mycobacterial (BCG) DNA which can induce IFN-α production in human neonatal pDCs pDCs are the predominant cellular source of secreted Type I IFN (IFN-α) [26]. We found that BCG bacilli did not stimulate IFN-α production from human neonatal pDCs (data not shown). However, using PCR for 2 genes used to identify BCG (rpoB and the RD8 portion in BCG) [25], we found that there was extracellular mycobacterial (BCG) DNA in the cell-free supernatants of reconstituted BCG vaccine vials (Fig 3A). BCG DNA stimulation induced modest upregulation of IFN-α1&2 mRNA production in isolated human umbilical cord pDCs at 8 hours (Fig 3B & 3C). Time points at 8, 16, and 24 hours were examined, and the greatest upregulation was seen at 8 hours. IFN-α is the secreted form of Type I IFN from pDCs [26], and IFN-α and IFN-β are both Type I IFNs that signal through the Type I IFN receptor [27]. 10.1371/journal.pone.0162148.g003Fig 3 The Bacille Calmette Guérin (BCG) vaccine contains extracellular mycobacterial (BCG) DNA which induces Type I interferon (IFN) production in human neonatal plasmacytoid dendritic cells (pDCs). (a) The cell-free supernatants of reconstituted BCG vaccine vials from two different manufacturers contain BCG DNA. PCR for 2 mycobacterial (BCG-specific) genes on the cell-free supernatants was performed as described in the Methods section, and the PCR products were run on an agarose gel and stained with ethidium bromide. Isolated human umbilical cord blood pDCs were stimulated with BCG DNA (100 μg/ml) for 8 hours and the relative expression levels of (b) IFN-α1 mRNA and (c) IFN-α2 mRNA were determined by qRT-PCR. †† p = 0.07, n = 3 independent experiments, bars are median values and error bars are S.D. BCG bacilli stimulate human neonatal CD16lo NK cells to produce IFN-γ in a TLR2-dependent manner NK cells are a source of early IFN-γ production, and they express TLR2 [28]. Isolated human umbilical cord NK cells were stimulated with BCG and intracellularly stained for IFN-γ production. BCG stimulation induced IFN-γ production from CD16lo NK cells. NK cells pre-incubated with an anti-TLR2 antibody showed diminished IFN-γ production compared to IgG isotype control (Fig 4). 10.1371/journal.pone.0162148.g004Fig 4 BCG stimulates CD16lo natural killer (NK) cells to produce interferon (IFN)-γ in a TLR2-dependent manner. Human neonatal NK cells were isolated from umbilical cord blood, stimulated with Bacille Calmette Guérin (BCG) overnight (in the presence of Brefeldin-A for the last 4 hours), stained with a vital dye (L/D Aqua), and intracellularly stained for CD3, CD16, CD56, and interferon (IFN)-γ for flow cytometry analysis. In Fig 4a-e, one representative experiment is shown. (a) gating strategy, (b) unstimulated control, (c) BCG stimulated, (d) BCG stimulated and pre-incubated with an isotype control antibody, (e) BCG stimulated and pre-incubated with a Toll-like receptor (TLR)2 blocking antibody. (f) Summary data for n = 5 independent experiments. Bars are mean values and error bars are S.D. † p = 0.06, BCG+TLR2 blocking IgG vs. BCG+isotype control IgG. Discussion The BCG vaccine is unique in its ability to induce Th1-polarized immune responses during the neonatal period. We propose that the BCG vaccine induces Th1 polarization in neonates by producing adult-like levels of IL-12 p35 and p40 (IL-12 p70) from cDCs, the predominant source of IL-12. BCG bacilli could activate TLR2 signaling in cDCs and CD16lo NK cells driving IL-12 p40 gene expression and IFN-γ secretion, respectively. Extracellular mycobacterial DNA in the BCG vaccine also could activate TLR9 signaling in pDCs driving IFN-α secretion. The combination of direct TLR2 activation on cDCs, TLR9-mediated IFN-α secretion by pDCs, and TLR2-mediated IFN-γ secretion by CD16lo NK cells, could act in concert to produce adult-like levels of IL-12 p35 and p40 mRNAs in the neonatal mdDCs. The adult-like levels of IL-12 p70 (IL-12 p35 and p40) from the cDCs then would drive Th1 differentiation of naïve CD4+ T-cells. An illustrative model is shown in Fig 5. 10.1371/journal.pone.0162148.g005Fig 5 Illustrative model demonstrating how the Bacille Calmette Guérin (BCG) vaccine might drive interleukin (IL)-12 p35 and p40 (IL-12 p70) production and CD4+ T-cell T-helper 1 (Th1) polarization in neonates. IFN = interferon; cDC = conventional dendritic cell; NK = natural killer; pDC = plasmacytoid dendritic cell; TLR = Toll-like receptor. The mycobacterial cell wall contains numerous lipoproteins and glycolipoproteins that act as potent TLR2/1 and TLR2/6 ligands [29–34]. The predominant role of TLR2 activation and signaling that we observed in BCG-induced IL-12 p40 production from human neonatal mdDCs is consistent with a previous publication utilizing TLR2 knockout mice [35]. Unlike another study using TLR4 knockout mice [36], we found that blocking TLR4 did not have an additive effect on the TLR2 blocking for BCG-stimulated IL-12 p40 production in human neonatal mdDCs. Given the predominant effect of TLR2 blockade, it would have been difficult to observe an additional effect of TLR4 blockade. Human neonatal mdDCs have repressed IL-12 p35 gene expression compared to adult cells [22]. Although TLR2 stimulation by itself (NF-κB activation only) was sufficient to upregulate IL-12 p40 mRNA a little, it was not sufficient to upregulate IL-12 p35 mRNA production in human neonatal mdDCs. TLR2 stimulation + Type II IFN could upregulate IL-12 p35 mRNA in neonatal mdDCs to a degree, but the combination of TLR2 stimulation + Type I IFN + Type II IFN achieved the maximal upregulation of IL-12 p35 and p40 mRNAs. In the cord blood (neonatal) mdDC preparations, the vast majority of the non-mdDCs were undifferentiated monocytes (approximately 25–30%). Although mdDCs are a primary source of IL-12, we cannot exclude the possibility that monocyte-derived IL-12 p35 and p40 production contributed to our findings. IL-12 p70 protein levels were also not examined in these experiments, as the upregulation solely of IL-12 p40 would manifest as increased IL-12 p70 levels. The absence of IL-12 p35 and p40 protein measurements is a potential limitation of this study. pDCs are the predominant cellular source of secreted Type I IFN (IFN-α) largely through TLR7/9 activation [26]. Mycobacterial DNA is a well-recognized potent TLR9 activating ligand [37]. We found that the cell-free supernatants of reconstituted BCG vaccine vials from 2 different manufacturers contained naked extracellular BCG DNA, and BCG DNA could stimulate IFN-α production from human neonatal pDCs (likely through TLR9 activation). The naked extracellular mycobacterial DNA in BCG vaccine vials is likely the result of dead and lysed BCG bacilli from the BCG vaccine manufacturing process. It should be noted that quality control standards established by the World Health Organization (WHO) for BCG vaccine products do not assess the presence of mycobacterial DNA in reconstituted vaccines [38]. The presence of extracellular mycobacterial DNA in reconstituted BCG vaccine vials likely serves as a TLR9 adjuvant, and an in vivo stimulus of local neonatal pDC Type I IFN production when the vaccine is given at birth. Mycobacterial DNA in the intact BCG bacilli might also contribute to TLR9-mediated neonatal pDC IFN-α production. However, we also observed that intact BCG bacilli did not stimulate IFN-α production from human neonatal pDCs. It is likely that the DNA in intact BCG bacilli cannot efficiently access TLR9 (in an endosomal location) in the pDCs. In addition to its effects on IL-12 p35 and p40 upregulation in neonates, Type I IFN can polarize neonatal CD4+ T-cells to a Th1 phenotype in an IL-12 independent manner [39]. Type I IFN can also induce IFN-γ production from neonatal NK cells [40]. Finally, NK cells are a source of early IFN-γ (Type II IFN) production and express TLR2 [41, 42]. CBMCs have a lower percentage of CD56+ NK cells compared to adult peripheral blood mononuclear cells (PBMCs) [43]. Confirming a previous study by Watkins et al.[44], we found that BCG bacilli could stimulate neonatal NK cells to produce IFN-γ. We further demonstrated that BCG bacilli induce IFN-γ production specifically from the CD16lo subset of human neonatal NK cells in a TLR2-dependent manner. The purity of the neonatal NK cell preparation (>95%) and the TLR2-dependence suggest that BCG bacilli are directly stimulating IFN-γ production in the CD16lo NK cells. However, we cannot exclude the possibility of some indirect NK cell activation. Combining the aforementioned data, we have proposed a model where the BCG vaccine could stimulate a combination of TLRs 2&9 signaling, and Types I and II IFN production, among neonatal cDCs, pDCs, and CD16lo NK cells. Intact BCG bacilli have TLR2 ligands in their cell wall, and extracellular BCG DNA in the vaccine vials could act as a pDC TLR9 ligand. The combination of the signals and cell subsets noted above could act in concert to optimize neonatal IL-12 p35 and p40 (IL-12 p70) production and subsequent CD4+ T-cell Th1 polarization in neonates (Fig 5). We postulate that the interaction between neonatal cDCs, pDCs, and CD16lo NK cells would take place within the dermis and regional draining lymph nodes of the BCG vaccination site in vivo. Our data demonstrates the potential mechanism by which the BCG vaccine could stimulate optimal neonatal IL-12 p35 and p40 (IL-12 p70) production by utilizing multiple innate immune cells. The production of adult-like levels of IL-12 p70 in neonates would promote the anti-mycobacterial CD4+ T-cell Th1 polarization that has been seen following neonatal BCG vaccination, and it could also contribute to beneficial heterologous Th1 immune effects of BCG vaccination [5, 9–11, 21]. Developing an effective Th1-inducing adjuvant for the neonatal period would be the first step towards the goal of improving vaccinations against toxins (e.g. diphtheria/pertussis/tetanus) and viruses (e.g. polioviruses, measles virus) so that they could begin at birth and require fewer booster doses. Our data suggests that an effective neonatal Th1-inducing adjuvant should be comprised of a TLR2 agonist, Type I IFN, and Type II IFN. The authors gratefully acknowledge the Molecular Biology section of the Research Institute for Tropical Medicine (RITM), Metro Manila, Philippines for performing the mycobacterial PCR on the cell-free supernatant of a reconstituted BCG vaccine vial (Tokyo strain) from the Japan BCG Laboratory. ==== Refs References 1 Marchant A , Goetghebuer T , Ota MO , Wolfe I , Ceesay SJ , De Groote D , et al Newborns develop a Th1-type immune response to Mycobacterium bovis bacillus Calmette-Guerin vaccination . J Immunol . 1999 ;163 (4 ):2249 –55 . .10438968 2 Vekemans J , Amedei A , Ota MO , D'Elios MM , Goetghebuer T , Ismaili J , et al Neonatal bacillus Calmette-Guerin vaccination induces adult-like IFN-gamma production by CD4+ T lymphocytes . Eur J Immunol . 2001 ;31 (5 ):1531 –5 . Epub 2001/07/24. 10.1002/1521-4141(200105)31:5<1531::AID-IMMU1531>3.0.CO;2-1 .11465110 3 Akkoc T , Aydogan M , Yildiz A , Karakoc-Aydiner E , Eifan A , Keles S , et al Neonatal BCG vaccination induces IL-10 production by CD4+ CD25+ T cells . Pediatr Allergy Immunol . 2010 ;21 (7 ):1059 –63 . 10.1111/j.1399-3038.2010.01051.x .20977501 4 Hussey GD , Watkins ML , Goddard EA , Gottschalk S , Hughes EJ , Iloni K , et al Neonatal mycobacterial specific cytotoxic T-lymphocyte and cytokine profiles in response to distinct BCG vaccination strategies . Immunology . 2002 ;105 (3 ):314 –24 . Epub 2002/03/29. 11918693 5 Jensen KJ , Larsen N , Biering-Sorensen S , Andersen A , Eriksen HB , Monteiro I , et al Heterologous immunological effects of early BCG vaccination in low-birth-weight infants in Guinea-Bissau: a randomized-controlled trial . J Infect Dis . 2015 ;211 (6 ):956 –67 . 10.1093/infdis/jiu508 25210141 6 Aaby P , Roth A , Ravn H , Napirna BM , Rodrigues A , Lisse IM , et al Randomized trial of BCG vaccination at birth to low-birth-weight children: beneficial nonspecific effects in the neonatal period? J Infect Dis . 2011 ;204 (2 ):245 –52 . 10.1093/infdis/jir240 .21673035 7 Shann F . Commentary: BCG vaccination halves neonatal mortality . Pediatr Infect Dis J . 2012 ;31 (3 ):308 –9 . Epub 2012/02/15. 10.1097/INF.0b013e3182465be8 00006454-201203000-00022 [pii]. .22330168 8 Higgins JPT, Soares-Weiser K, Reingold A. Systematic review of the non-specific effects of BCG, DPT and measles containing vaccines. Review. World Health Organization: 2014 13 March 2014. Report No. 9 Goodridge HS , Ahmed SS , Curtis N , Kollmann TR , Levy O , Netea MG , et al Harnessing the beneficial heterologous effects of vaccination . Nat Rev Immunol . 2016 ;16 (6 ):392 –400 . 10.1038/nri.2016.43 .27157064 10 Libraty DH , Zhang L , Woda M , Acosta LP , Obcena A , Brion JD , et al Neonatal BCG vaccination is associated with enhanced T-helper 1 immune responses to heterologous infant vaccines . Trials Vaccinol . 2014 ;3 :1 –5 . 10.1016/j.trivac.2013.11.004 24611083 11 Organization WH. Systematic Review of the Non-specific Immunological Effects of Selected Routine Childhood Immunizations. 2014 13.03.2014. Report No. 12 de Castro MJ , Pardo-Seco J , Martinon-Torres F . Nonspecific (Heterologous) Protection of Neonatal BCG Vaccination Against Hospitalization Due to Respiratory Infection and Sepsis . Clin Infect Dis . 2015 ;60 (11 ):1611 –9 . 10.1093/cid/civ144 .25725054 13 Trinchieri G . Interleukin-12 and its role in the generation of TH1 cells . Immunol Today . 1993 ;14 (7 ):335 –8 . 10.1016/0167-5699(93)90230-I .8103338 14 Wu CY , Demeure C , Kiniwa M , Gately M , Delespesse G . IL-12 induces the production of IFN-gamma by neonatal human CD4 T cells . J Immunol . 1993 ;151 (4 ):1938 –49 . .8102154 15 Barrios C , Brawand P , Berney M , Brandt C , Lambert PH , Siegrist CA . Neonatal and early life immune responses to various forms of vaccine antigens qualitatively differ from adult responses: predominance of a Th2-biased pattern which persists after adult boosting . Eur J Immunol . 1996 ;26 (7 ):1489 –96 . .8766551 16 Lee SM , Suen Y , Chang L , Bruner V , Qian J , Indes J , et al Decreased interleukin-12 (IL-12) from activated cord versus adult peripheral blood mononuclear cells and upregulation of interferon-gamma, natural killer, and lymphokine-activated killer activity by IL-12 in cord blood mononuclear cells . Blood . 1996 ;88 (3 ):945 –54 . .8704253 17 Kollmann TR , Crabtree J , Rein-Weston A , Blimkie D , Thommai F , Wang XY , et al Neonatal innate TLR-mediated responses are distinct from those of adults . J Immunol . 2009 ;183 (11 ):7150 –60 . 10.4049/jimmunol.0901481 19917677 18 Basha S , Surendran N , Pichichero M . Immune responses in neonates . Expert Rev Clin Immunol . 2014 ;10 (9 ):1171 –84 . 10.1586/1744666X.2014.942288 25088080 19 Debock I , Flamand V . Unbalanced Neonatal CD4(+) T-Cell Immunity . Front Immunol . 2014 ;5 :393 10.3389/fimmu.2014.00393 25221551 20 Morris MC , Surendran N . Neonatal Vaccination: Challenges and Intervention Strategies . Neonatology . 2016 ;109 (3 ):161 –9 . 10.1159/000442460 26757146 21 Ritz N , Mui M , Balloch A , Curtis N . Non-specific effect of Bacille Calmette-Guerin vaccine on the immune response to routine immunisations . Vaccine . 2013 ;31 (30 ):3098 –103 . Epub 2013/04/16. S0264-410X(13)00405-2 [pii] 10.1016/j.vaccine.2013.03.059 .23583897 22 Goriely S , Vincart B , Stordeur P , Vekemans J , Willems F , Goldman M , et al Deficient IL-12(p35) gene expression by dendritic cells derived from neonatal monocytes . J Immunol . 2001 ;166 (3 ):2141 –6 . Epub 2001/02/13. .11160266 23 Goriely S , Van Lint C , Dadkhah R , Libin M , De Wit D , Demonte D , et al A defect in nucleosome remodeling prevents IL-12(p35) gene transcription in neonatal dendritic cells . J Exp Med . 2004 ;199 (7 ):1011 –6 . Epub 2004/03/31. 10.1084/jem.20031272 [pii]. 15051764 24 Lu YC , Yeh WC , Ohashi PS . LPS/TLR4 signal transduction pathway . Cytokine . 2008 ;42 (2 ):145 –51 . 10.1016/j.cyto.2008.01.006 .18304834 25 Kim Y , Choi Y , Jeon BY , Jin H , Cho SN , Lee H . A simple and efficient multiplex PCR assay for the identification of Mycobacterium genus and Mycobacterium tuberculosis complex to the species level . Yonsei Med J . 2013 ;54 (5 ):1220 –6 . 10.3349/ymj.2013.54.5.1220 23918573 26 Reizis B , Bunin A , Ghosh HS , Lewis KL , Sisirak V . Plasmacytoid dendritic cells: recent progress and open questions . Annu Rev Immunol . 2011 ;29 :163 –83 . 10.1146/annurev-immunol-031210-101345 21219184 27 Uze G , Schreiber G , Piehler J , Pellegrini S . The receptor of the type I interferon family . Curr Top Microbiol Immunol . 2007 ;316 :71 –95 . .17969444 28 Esin S , Counoupas C , Aulicino A , Brancatisano FL , Maisetta G , Bottai D , et al Interaction of Mycobacterium tuberculosis cell wall components with the human natural killer cell receptors NKp44 and Toll-like receptor 2 . Scand J Immunol . 2013 ;77 (6 ):460 –9 . 10.1111/sji.12052 .23578092 29 Brightbill HD , Libraty DH , Krutzik SR , Yang RB , Belisle JT , Bleharski JR , et al Host defense mechanisms triggered by microbial lipoproteins through toll-like receptors . Science . 1999 ;285 (5428 ):732 –6 . .10426995 30 Gilleron M , Quesniaux VF , Puzo G . Acylation state of the phosphatidylinositol hexamannosides from Mycobacterium bovis bacillus Calmette Guerin and mycobacterium tuberculosis H37Rv and its implication in Toll-like receptor response . J Biol Chem . 2003 ;278 (32 ):29880 –9 . 10.1074/jbc.M303446200 .12775723 31 Noss EH , Pai RK , Sellati TJ , Radolf JD , Belisle J , Golenbock DT , et al Toll-like receptor 2-dependent inhibition of macrophage class II MHC expression and antigen processing by 19-kDa lipoprotein of Mycobacterium tuberculosis . J Immunol . 2001 ;167 (2 ):910 –8 . .11441098 32 Pecora ND , Gehring AJ , Canaday DH , Boom WH , Harding CV . Mycobacterium tuberculosis LprA is a lipoprotein agonist of TLR2 that regulates innate immunity and APC function . J Immunol . 2006 ;177 (1 ):422 –9 . .16785538 33 Quesniaux VJ , Nicolle DM , Torres D , Kremer L , Guerardel Y , Nigou J , et al Toll-like receptor 2 (TLR2)-dependent-positive and TLR2-independent-negative regulation of proinflammatory cytokines by mycobacterial lipomannans . J Immunol . 2004 ;172 (7 ):4425 –34 . .15034058 34 Gehring AJ , Dobos KM , Belisle JT , Harding CV , Boom WH . Mycobacterium tuberculosis LprG (Rv1411c): a novel TLR-2 ligand that inhibits human macrophage class II MHC antigen processing . J Immunol . 2004 ;173 (4 ):2660 –8 . .15294983 35 von Meyenn F , Schaefer M , Weighardt H , Bauer S , Kirschning CJ , Wagner H , et al Toll-like receptor 9 contributes to recognition of Mycobacterium bovis Bacillus Calmette-Guerin by Flt3-ligand generated dendritic cells . Immunobiology . 2006 ;211 (6–8 ):557 –65 . Epub 2006/08/22. S0171-2985(06)00062-3 [pii] 10.1016/j.imbio.2006.05.004 .16920494 36 Uehori J , Matsumoto M , Tsuji S , Akazawa T , Takeuchi O , Akira S , et al Simultaneous blocking of human Toll-like receptors 2 and 4 suppresses myeloid dendritic cell activation induced by Mycobacterium bovis bacillus Calmette-Guerin peptidoglycan . Infect Immun . 2003 ;71 (8 ):4238 –49 . 12874299 37 Akira S , Uematsu S , Takeuchi O . Pathogen recognition and innate immunity . Cell . 2006 ;124 (4 ):783 –801 . 10.1016/j.cell.2006.02.015. .16497588 38 Milstien JB , Gibson JJ . Quality control of BCG vaccine by WHO: a review of factors that may influence vaccine effectiveness and safety . Bull World Health Organ . 1990 ;68 (1 ):93 –108 . 2189588 39 Zhang X , Casartelli N , Lemoine S , Mozeleski B , Azria E , Le Ray C , et al Plasmacytoid dendritic cells engagement by influenza vaccine as a surrogate strategy for driving T-helper type 1 responses in human neonatal settings . J Infect Dis . 2014 ;210 (3 ):424 –34 . 10.1093/infdis/jiu103 .24558121 40 Eljaafari FM , Takada H , Tanaka T , Doi T , Ohga S , Hara T . Potent induction of IFN-gamma production from cord blood NK cells by the stimulation with single-stranded RNA . Journal of clinical immunology . 2011 ;31 (4 ):728 –35 . 10.1007/s10875-011-9528-4 .21487893 41 Lauzon NM , Mian F , MacKenzie R , Ashkar AA . The direct effects of Toll-like receptor ligands on human NK cell cytokine production and cytotoxicity . Cell Immunol . 2006 ;241 (2 ):102 –12 . 10.1016/j.cellimm.2006.08.004 .17049504 42 Souza-Fonseca-Guimaraes F , Adib-Conquy M , Cavaillon JM . Natural killer (NK) cells in antibacterial innate immunity: angels or devils? Mol Med . 2012 ;18 :270 –85 . 10.2119/molmed.2011.00201 22105606 43 Osugi Y , Hara J , Kurahashi H , Sakata N , Inoue M , Yumura-Yagi K , et al Age-related changes in surface antigens on peripheral lymphocytes of healthy children . Clinical and experimental immunology . 1995 ;100 (3 ):543 –8 . 7774067 44 Watkins ML , Semple PL , Abel B , Hanekom WA , Kaplan G , Ress SR . Exposure of cord blood to Mycobacterium bovis BCG induces an innate response but not a T-cell cytokine response . Clin Vaccine Immunol . 2008 ;15 (11 ):1666 –73 . Epub 2008/09/26. CVI.00202-08 [pii] 10.1128/CVI.00202-08 18815231

PMC005xxxxxx/PMC5003385.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757147410.1371/journal.pone.0161907PONE-D-16-20588Research ArticleMedicine and Health SciencesVascular MedicineBlood PressureBiology and Life SciencesAnatomyBody FluidsBloodBlood FlowMedicine and Health SciencesAnatomyBody FluidsBloodBlood FlowBiology and Life SciencesPhysiologyBody FluidsBloodBlood FlowMedicine and Health SciencesPhysiologyBody FluidsBloodBlood FlowMedicine and Health SciencesHematologyBloodBlood FlowEngineering and TechnologyEquipmentMeasurement EquipmentPhysical SciencesPhysicsClassical MechanicsContinuum MechanicsFluid MechanicsFluid DynamicsFluid FlowResearch and Analysis MethodsBioassays and Physiological AnalysisCardiovascular AnalysisCerebral Blood Flow AssayBiology and Life SciencesAgricultureLivestockSwineBiology and Life SciencesOrganismsAnimalsVertebratesAmniotesMammalsSwineMedicine and Health SciencesNeurologyBrain DamageMedicine and Health SciencesCritical Care and Emergency MedicineTrauma MedicineBrain DamagePhysical SciencesPhysicsClassical MechanicsPressureHydrostatic PressureCerebral Autoregulation Real-Time Monitoring Cerebral Autoregulation Real-Time MonitoringTsalach Adi *Ratner Eliahu Lokshin Stas Silman Zmira Breskin Ilan Budin Nahum Kamar Moshe Ornim Medical Ltd, Kfar Saba, IsraelBaud Olivier EditorHopital Robert Debre, FRANCECompeting Interests: The study was sponsored by Ornim Medical LTD. Adi Tsalach, Eliahu Ratner, Stas Lokshin, Zmira Silman, Ilan Breskin, Nahum Budin and Moshe Kamar are eomplyed by Ornim Medical LTD. Ornim Medical LTD. holds the patent of the UT-NIRS technology (US 20150327779). There are no additional patents, products in development or marketed products to declare. This does not alter the authors' adherence to all the PLOS ONE policies on sharing data and materials, as detailed online in the guide for authors. Conceptualization: AT NB SL MK. Formal analysis: AT ER ZS. Funding acquisition: NB MK. Investigation: AT MK. Methodology: AT NB SL IB MK. Supervision: NB MK. Validation: AT. Writing – original draft: AT. Writing – review & editing: IB ZS MK. * E-mail: adi@ornim.com29 8 2016 2016 11 8 e016190722 5 2016 12 8 2016 © 2016 Tsalach et al2016Tsalach et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Cerebral autoregulation is a mechanism which maintains constant cerebral blood flow (CBF) despite changes in mean arterial pressure (MAP). Assessing whether this mechanism is intact or impaired and determining its boundaries is important in many clinical settings, where primary or secondary injuries to the brain may occur. Herein we describe the development of a new ultrasound tagged near infra red light monitor which tracks CBF trends, in parallel, it continuously measures blood pressure and correlates them to produce a real time autoregulation index. Its performance is validated in both in-vitro experiment and a pre-clinical case study. Results suggest that using such a tool, autoregulation boundaries as well as its impairment or functioning can be identified and assessed. It may therefore assist in individualized MAP management to ensure adequate organ perfusion and reduce the risk of postoperative complications, and might play an important role in patient care. Ornim Medical LTDTsalach Adi The study was sponsored by Ornim Medical LTD. Ornim Medical LTD provided support in the form of salaries for authors [AT, ER, SL, ZS, IB, NB, MK], but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper.Data Availability All relevant data are within the paper. ==== Body Introduction Cerebral autoregulation (AR) refers to the intrinsic ability of the brain’s vasculature to react to changes in arterial blood pressure (BP), in order to maintain stable cerebral blood flow (CBF)[1–3]. Lassen et al.[1] have reported an AR curve which was set as standard, comprising of a CBF plateau with lower and upper limits in mean arterial pressure (MAP) values of 50 to 150mmHg. However, these limits are subject to inter-patient variations[4], and alterations controlled by factors that change CBF and its vasorectivity, such as the sympathetic activity or the vascular rennin-angiotensin system[5]. In chronic hypertension for example, the limits of AR are shifted toward higher blood pressures[6]. AR may be further compromised by certain disease states of the brain, ranging from impairment to non-function, leaving the brain unprotected against potentially harmful effects of BP changes. In severe head injury or acute ischemic stroke, autoregulation may be impaired or even lost[7, 8]. Thus, assessing whether this mechanism is intact or impaired, and determining the boundaries of the aforementioned plateau, is important in many clinical settings, where primary or secondary injuries to the brain may occur due to hypo or hyperperfusion. However, until today, no real time monitor for AR functioning exists. Over the years several methods were introduced to examine AR, all using surrogates indices, such as PRx (using intra cranial pressure (ICP))[9–11], Mx (using CBF Velocity (CBFV) measured by Transcranial Doppler (TCD))[12–14], COx (using cerebral oximetry measured by Near Infrared Spectroscopy (NIRS))[14, 15], and others. Though many additional parameters may be associated with cerebral autoregulation, its basic mechanism is usually described by either the static or dynamic relationship between MAP and CBF[1, 2, 16, 17]. Various analysis techniques were developed to assess and quantify the association between these two parameters in either time or frequency domains[18–20]. Few examples include: assessment of the rate of recovery of CBF in response to a MAP challenge[2], calculation of an autoregulation index based on mathematical models[21], and computation of a transfer function with evaluation of its properties[22, 23]. A common analysis method incorporates calculation of the correlation between CBF to cerebral perfusion pressure[24] or MAP[12]. In this method, a moving linear correlation coefficient is calculated for a certain time window and evaluated over time. Correlation coefficients approaching 1 are associated with loss of autoregulation as CBF is dependent on MAP changes (i.e. “pressure-passive”), while lower values, approaching zero, are related to cases in which autoregulation is intact. This method was used to continuously monitor autoregulation in numerous studies and various disease\clinical states. For these studies, a correlation coefficient threshold value of 0.4–0.45 was established to distinguish between intact and impaired autoregulation[13, 14, 25]. The value is dependent upon the correlated signals (CBF surrogate), calculation procedure, etc. This correlation technique, however, was mostly applied in post processing rather than providing a real time AR indication, as no device incorporating both measurements is present. Ornim’s acousto-optic based device (c-FLOW and its predecessor the CerOx) is a noninvasive, continuous CBF monitor[26–29]. The ability to perform offline calculations of the AR state using its cerebral flow index (CFI) and an independent MAP measurement was demonstrated in previous studies[28, 29]. Herein we describe the development of a new AR monitor which correlates a direct CBF signal with concurrent MAP measurement to produce a real time autoregulation index (ARI). Methods The c-FLOW is an ultrasound tagged light based device, which utilizes near infrared laser light (808nm) modulated by low power 1MHz ultrasound, to perform continuous real time blood flow monitoring in the tissue microcirculation. Being a non-invasive device, its sensors are usually placed on the patient’s forehead and provide local microcirculation CBF monitoring in approximately 1cm3 volume, located 2cm deep underneath them. Due to the use of ultrasound, placement of the sensors on every other location but the forehead requires pre-shaving of the region of interest. The c-FLOW provides a Cerebral Flow Index (CFI) which describes changes in cerebral blood flow in arbitrary units between 0–100, where 0 represents no flow. It is also capable of setting a baseline flow value and presenting the percent flow change from this baseline. Each CFI value represents a moving average of the last 30 seconds of ultrasound tagged near infra red (UT-NIR) signal and is updated every 2–3 seconds. The specifics on how it operates and eliminates the effect of superficial flow are detailed elsewhere[30]. Due to the use of ultrasound and light, the c-FLOW has several inherent limitations. As every other NIR based device, its measurement depth is confined to the cortex only because of light absorption in tissues. In addition, since it includes ultrasound, good coupling between the sensors to the skin is obtained by US gel which has to be renewed every few hours. To enable autoregulation monitoring, a modified version of the c-FLOW that includes a BP unit was developed, termed c-FLOW-AR. This unit connects to a standard invasive blood pressure (IBP) sensor, and enables tracking MAP concurrently with CFI and display both trends synchronically. The real time combination of these two parameters enables the calculation of a correlation index (termed ARI), which reflects the interrelationship between changes of MAP to those of CFI. ARI values range between 0 to 100, where 0 represents a condition of no correlation between MAP and CFI changes, while 100 represents a perfect one. Accordingly, when cerebral autoregulation is intact, a change in MAP is not followed by a corresponding change of CBF, as the brain autoregulates. In this case, ARI will get lower values (approaching zero). If autoregulation is impaired, it is expected that changes in MAP will cause corresponding changes in CBF and ARI should get higher values. ARI is calculated only for time intervals in which marked MAP changes exist. Identification of notable MAP changes is obtained by a Relevance Vector Machine (RVM) linear classifier based on the trend slope and derivatives. For all other time intervals, in which MAP is approximately constant, no ARI is calculated. The ARI is therefore the reflection of the autoregulation status, once changes in marked MAP occur. The ARI value is obtained by calculating the cross correlation maxima in a predetermined time interval (5 minutes). In this interval, the correlation values between CFI and MAP are calculated for all time shifts. They are then multiplied by a time-shift dependent weight function, in which the weight decays as time shift increases, and the maximum obtained value is selected. In this way, the ARI compensates for the intrinsic physiological time delays between changes in systemic blood flow and pressure to those of the brain. In-vitro experiments CBF was modeled using a previously described acousto-optic phantom which mimics blood flow in tissues[30, 31]. The phantom is made of Dermasol (CA medical innovations) which is a synthetic polymer matrix soaked with oil. Titanium Dioxide (TiO2) particles (0.1% by weight) were added as light scattering agents. The optical and acoustic properties of the phantom are similar to those of tissue, as detailed in Table 1. 10.1371/journal.pone.0161907.t001Table 1 optical and acoustic properties of the phantom and the tissue[32, 33]. Property Tissue–muscle/brain Phantom Light Effective decay coefficient 2.17/2.12 cm-1 2.2 ±0.2cm-1 Sound Velocity 1.5 ·105 cm/s 1.43 ·105 cm/s Acoustic impedance 150–170 Kg/cm2∙s 149 Kg/cm2∙s Simulating tissue microcirculation, we designed a Dermasol slab containing 20 hollow parallel channels with an outer diameter of 1mm through which scattering fluid (similar to blood) can flow. A schematic diagram of the experimental set-up is presented in Fig 1. 10.1371/journal.pone.0161907.g001Fig 1 Schematic illustration of the flow modeling experimental setup. The C-FLOW’s measuring sensor is placed above the tissue phantom. Flow within the phantom’s tubes is generated using the syringe pump. Flow within the phantom’s tubes was generated using a calibrated computer-controlled syringe pump (Chemix, Model Infusion 200) and measured using the c-FLOW-AR’s (Ornim Medical, Israel) sensor which was placed above. Though flow within cerebral microcirculation is multidirectional, flow within the phantom’s channels was linear. As the c-FLOW is insensitive to flow direction, this simplified model was sufficient to demonstrate the device’s capabilities. BP was modeled using a hydraulic pressure system that was able to generate periodical hydrostatic pressure, similarly to blood pressure. The following experimental setup was utilized (Fig 2A). 10.1371/journal.pone.0161907.g002Fig 2 (A) Schematic illustration of the pressure modeling experimental setup comprising of a water column connected to a computer controlled hydraulic pump and a peristaltic pump. Pressure was measured using a standard disposable pressure transducer (DPT) connected directly to the C-FLOW-AR. (B) An example of a pressure wave generated by the hydraulic pressure system designed to mimic blood pressure. All pressure properties, such as average pressure (MAP), oscillations magnitude (Systolic-Diastolic pressures), and frequency (HR), are controlled and can be predetrmined. Constant static pressure was created by a water column of a predetermined height (h), as illustrated on the left of Fig 2(A). The height (h) relative to the pressure sensor was digitally changed using a computer controlled hydraulic pump and a pressure controller, thus creating different static pressure levels, corresponding to different MAP values. To create pressure oscillations around the static pressure (pulsatile pressure), the water column was hydraulically connected to a ring-like tube filled with water. A peristaltic pump (MasterFlex L/S easy-load II, model 77200–62) connected to that tube was used to create pressure variations over time. The water column was hydraulically connected to a standard disposable pressure transducer (DPT), similar to the one used in arterial lines (Art-LineTM Biometrix, Israel), and directly connected to the c-FLOW-AR. Example for a pressure wave generated by the described hydraulic pressure system, measured by the c-FLOW-AR is illustrated in Fig 2(B). A designated LabView® program was used to control both the flow and the pressure systems synchronically. To examine the ARI performance in different AR conditions, the following protocol was applied. Each of the two cFLOW-AR’s sensors was placed on a different acousto-optic flow phantom (depicted in Fig 1) to enable the implementation of different flow protocols to each of the sensors. Average pressure (MAP) was raised and lowered with a notable amplitude change (from 100mmHg to 180mmHg). Flow in phantom number 1 (measured by sensor number 1) was changed in accordance with MAP changes (modeling pressure passive condition), while flow in phantom number 2 (measured by sensor number 2) was kept constant (modeling intact AR). In such a way, both cases of impaired and intact AR were simulated simultaneously (with sensors 1 & 2 respectively). Mean pressure, flow index and the correlation (ARI) between them were real-time displayed on the C-FLOW-AR’s screen and automatically saved to the device hard drive for post processing purposes. The coefficient of variance was used to estimate the homogeneity of CFI as measured by the two c-FLOW-AR’s sensors and validate the applied flow protocols. ARI values measured for the two different flow protocols were compared using independent t-test. Significance level was defined as α = 0.05. Preclinical case study A case study of real-time assessment of a swine autoregulatory state is presented. The procedures were approved by “Asaf Harofe” medical center Institutional Animal Care and Use Committee (IACUC) and conducted in strict accordance with the guideline for animal care and use established by the IACUC. Animal preparation A female piglet (Sus domestica, 2–3 months old), weighing 25.6Kg was anaesthetized with an initial bolus of IM Ketamine, 1.5mg/Kg and Xylazine 2mg/Kg. After induction, the animal was intubated and mechanically ventilated, keeping SaO2 above 93% and ETCO2 at 35–40mmHg. Anesthesia and analgesia was maintained using IV Propofol 0.02–0.1 mg/Kg/min and Fentanyl 0.015 mcg/Kg/min. An arterial line was inserted to the carotid artery for BP monitoring. To avoid hair and enable both sensors adhesion to the skin and good US coupling, the animal’s head was shaved (~10x10cm area) and cleaned with alcohol solution. The skin remained intact with no visible scratches or wounds. Monitoring MAP was measured using c-FLOW-AR monitor (Ornim Medical, Israel) via arterial line introduced to the carotid artery. CFI was monitored with a non-invasive sensor (5x2.5x1.5cm) placed on the skin surface of animal’s forehead connected to the c-FLOW-AR. Correlation Index between MAP and CFI (ARI) was calculated in real time and presented on the monitor. Heart rate and ventilation parameters (respiratory rate, end tidal CO2, arterial saturation) were continuously monitored using non invasive pulse oximetry and capnography (Novametrix, USA). Experimental procedure Cardiac preload, and therefore cardiac output, was optimized using fluid boluses of 5–10 ml/kg, while monitoring BP and PaO2. After MAP stabilization, experiment began. Baseline CFI was recorded for a period of 15–30 minutes prior to starting each manipulation. To create MAP variations, BP was pharmacologically manipulated. Intravenous (IV) Phenylephrine (50 mcg/ml) was used to increase MAP. Incremental dosages were infused, starting at 2.5 ml/hr and increasing in 2.5 ml/hr every 7–10 minutes, until MAP doubled from baseline. Once targeted MAP was reached, Phenylephrine injection was stopped for 30–60 minutes and a new baseline was acquired. IV Nitropruside (2 mg/ml) was used to decrease BP. Incremental dosages, starting at 2 ml/hr and increasing in 2 ml/hr were infused every 7–10 minutes, until MAP dropped by 50% or reached 40 mmHg. Once targeted MAP was reached, Nitroprusside stopped for 30–60 minutes allowing the animal to stabilize in a new baseline. Data Collection MAP, CFI, and the calculated correlation index (ARI) were digitally saved to the c-FLOW-AR device. Other physiologic parameters were sampled using a designated LabView program and saved to an excel worksheet for analysis. Data Analysis Changes in CFI were correlated with changes in MAP throughout the monitoring period. To demonstrate the autoregulation assessment ability, CFI was plotted as a function of MAP, to illustrate the AR curve and detect its boundaries (upper and lower limits). The slope and 95% confidence interval (CI) between MAP and CFI was calculated for the two MAP regions. ARI values were binned into groups according to their corresponding MAP values (10mmHg segments). Averaged values for each bin were presented using a columns diagram. Receiver Operating Curve (ROC) analysis was used to estimate the ARI performance in classifying MAP as under or over the limits of autoregulation. Analyses were carried out using SPSS 23.0.01 and Matlab R2014a (8.3.0.532). Results In-vitro Experiments Laboratorial experimental setup was used to increase and decrease MAP between 100mmHg to 180mmHg (Fig 3 -Panel C). Flow measured by sensor number 1 (panel B) experienced corresponding concurrent changes (CFI = 44.93±18.18, coefficient of variance (CV) = 0.4), while flow measured by sensor 2 (Panel D) was kept constant throughout the experiment (CFI = 22.89±4.6, CV = 0.2). CV obtained for sensor 1 was twice higher than the one of sensor 2, indicating on fluctuating versus steady flow, respectively. To further validate the difference between CFI values measured by the two different sensors during the experiment, measurements were divided to periods before and after MAP increase, and the difference between groups was evaluated using t-test. Averaged CFI values obtained for the two groups with their corresponding p values are summarized in Table 2. 10.1371/journal.pone.0161907.g003Fig 3 MAP (C), CFI (B,D) and ARI (A,E) data obtained in the in-vitro experiment. MAP (panel C) was significantly increased and decreased, followed by a similar flow protocol which was applied to sensor 1 (panel B) versus constant flow which was measured by sensor 2 (panel D). Sensor 1 which experienced a pressure passive flow protocol was used to model No AutoRegulation (NAR) state, while sensor 2 which measured constant flow was utilized to model intact AutoRegulation (AR) state. Corresponding calculated autoregulation indexes are depicted in panels A and E respectively. Delayed ARI values relative to MAP changes are due to data collection and buffering required for ARI calculation. 10.1371/journal.pone.0161907.t002Table 2 Averaged CFI values obtained in in-vitro experiment. CFI Before MAP Increase CFI After MAP Increase P value Sensor 1 39.80 59.61 6.63E-58 Sensor 2 23.35 23.84 0.21 Results show that significant difference was obtained before and after MAP increase in sensor 1 (P<0.001), while sensor 2 measured stable values (P>0.05), validating the applied manipulation. Obtained autoregulation indexes for the two disparate sensors were significantly different (ARI for sensor 1 = 61.74±21.36, ARI for sensor 2 = 37.66±17.99, Independent T-test, P<0.001). Fig 3 illustrates data over time (MAP (C), 2 CFI channels (B,D) and corresponding ARI values(A,E))) obtained in the in-vitro experiment. To better illustrate the distribution of the ARI values, Fig 4 demonstrates a boxplot of the ARI calculated for the two sensors, exhibiting again a distinct separation between the two groups. 10.1371/journal.pone.0161907.g004Fig 4 Boxplot for autoregulation index values calculated for the two to cFLOW-AR sensors. Pink astrics represent averaged ARI for each condition. A distinct separation between the conditions is apparent. Preclinical case study MAP and CFI data were obtained, using c-FLOW-AR, for a duration of over 4 hours. Anesthesia with IV Propofol is known not to compromise AR as opposed to volatile anesthesia in high concentrations[34], therefore intact AR curve was expected to be observed. MAP was lowered to approximately 40mmHg and raised to 140mmHg using stepwise increases of drug dose. Data over time is presented in Fig 5 (left), where blue points represent all MAP values and red points are associated with periods in which the algorithm identified a significant MAP change and therefore the ARI was calculated. 10.1371/journal.pone.0161907.g005Fig 5 Left—MAP and CFI data over time throughout the study. MAP was increased to 140mmHg followed by a return to baseline and a decrease to 40mmHg. Dashed green lines represent initial injections of Phnylephrine and Nitroprusside respectively. Blue points represent all MAP values. Red points are associated with periods in which the algorithm identified a significant MAP change and a correlation index (ARI) can be calculated. Right—Scatter plot of CFI versus MAP revealing two distinct slopes obtained for values under of over 100mmHg. This point was defined as the upper limit of autoregulation (ULA). Mean baseline MAP value was 67.6±0.36mmHg. A scatter plot of synchronic CFI over MAP data was outlined to demonstrate the AR curve (Fig 5 (right)), with a cutoff MAP distinguishing between a plateau and a linear behavior obtained in two distinct MAP ranges (only the periods of significant MAP change are plotted). From this figure it is evident that cutoff MAP for this animal was observed at 100mmHg. To validate the MAP cutoff, the slope of CFI change in response to MAP change was calculated for both MAP regions, demonstrating significantly different slope values for MAP under 100 mmHg (slope = 0.108, 95% confidence interval [0.104, 0.112]) and over 100 mmHg (slope = 0.556, 95% confidence interval [0.544, 0.568]). We therefore defined this point as the Upper Limit of Autoregulation (ULA). The lower limit of autoregulation (LLA) could not be determined within the obtained MAP range, as no additional slope was identified. To further exemplify the ARI ability to differ between MAP values over and under the ULA, ARI bar graph and corresponding ROC analysis are presented in Fig 6. 10.1371/journal.pone.0161907.g006Fig 6 Left–Bar Diagram of averaged autoregulation index for each 10mmHg MAP segment. Error bars represent the standard error of the mean. Threshold value is 52. Right–ROC analysis for MAP classification as over or under the ULA. Area Under the Curve (AUC) is 0.848 (95% confidence interval [0.783, 0.912]). The bar height chart represents the mean ARI obtained in each MAP segment (error bars stand for the standard error of the mean in each segment). ARI values obtained for MAP over the ULA were significantly higher than those obtained for MAP under the ULA (unpaired t-test, p<0.001). ARI threshold used was 52 (Fig 6 Left). The ROC analysis for classifying MAP as over or under the ULA according to its matching ARI yielded an Area Under the Curve (AUC) of 0.848 (95% confidence interval [0.783, 0.912]), indicating a good accuracy in MAP discrimination. Discussion The development of a novel device incorporating MAP and CFI measurements to enable real time autoregulation monitoring is presented. Its ability to correlate the two parameters in real time and provide an index indicating on autoregulation is introduced and validated in both in-vitro experiments and a preclinical case study. Results suggest that using such a tool, autoregulation boundaries as well as its impairment or functioning can be identified and assessed. The combination of CBF monitoring along with synchronized systemic MAP was extensively researched and was suggested to be beneficial in providing an indication of the autoregulatory state[35]. In those studies, CBF was not directly measured, instead, surrogate parameters, such as TCD cerebral blood flow velocity (CBFV), ICP and NIRS oximetry, were used. The use of these parameters has substantial limitations. For TCD, which insonates large vessels, the flow measurement requires the assumption of constant vessel diameter, an inherent problem when assessing the vasorectivity based mechanism of AR. In addition, TCD is difficult to use continuously, especially in an environment of electrical noise, such as the operating room. Because of the difficulties in acquiring and maintaining a stable TCD signal, its mean velocity index (Mx) is not the most widely used monitor for assessing autoregulation, although it was arguably the first[10]. Another widely studied technique involves the correlation of MAP with intracranial pressure (ICP) yielding a pressure reactivity index (PRx), which is also an indirect indication for the patient’s autoregulatory state[9–11]. This method is based on pressure AR theory rather than on CBF AR. But more important, it is invasive and therefore can be applied to a limited patient population, namely severe traumatic brain injury who are subject to intensive intracranial monitoring. Regional NIRS oximetry (rStO2), which measures tissue mixed venous blood oxygen saturation, assumes that changes in blood flow have direct implication on rStO2. Yet this parameter is proportional to hemoglobin concentration in the tissue, which does not necessarily correspond to blood flow. An additional disadvantage of this technology is the influence of superficial extracerebral layers on the signal which has been shown to be significant in a number of studies[36, 37]. This may introduce misleading conclusions while assessing AR and pursuing the difference between systemic and cerebral behaviors. Ornim’s Ultrasound Tagged Near-Infra Red (UT-NIR) based device (the c-FLOW and its predecessor the CerOx) directly monitors the CBF trend continuously and non-invasively. Preliminary studies which utilize its cerebral flow index (CFI) for autoregulation assessment were conducted[28, 29] and showed good agreement between CBF autoregulation monitored by CerOx (correlation flow index (CFx)) compared with the TCD-based monitoring (mean velocity index (Mx))[28]. Therefore, the development of a new monitor using the linear correlation construct is a logical choice. This device which measures both MAP and CFI data and correlates them in real-time to obtain an indication for the autoregulatory state presented here, opens a new opportunity for an integrated monitor. Autoregulation assessment in this study is based on the relationship between CBF (as expressed by CFI) to MAP changes. By definition, AR refers to CBF response to MAP changes and the ability of the brain to maintain constant CBF in a certain range of MAP values. Hence, assessment of the relationship between synchronized MAP and CBF signals should provide a solid indication for the patient’s autoregulatory state. It is well established that there are additional factors which may contribute to the estimation of AR condition, such as oxygen saturation[15] and blood volume[38]. Future work should investigate the possibility to combine these three parameters, using the UT-NIR technology of the c-FLOW, as suggested by Brady[10], to assist in drawing a bigger picture as to the patient’s autoregulation condition. Yet, even the narrower observation on MAP and CFI only described here, might be sufficient, at first, and provide valuable information with straightforward clinical virtues. Experiment in laboratorial environment was designed to validate the device ability to measure both signals simultaneously and correlate them in real-time. The pressure wave produced by the experimental setup (illustrated in Fig 2B) although not being identical to a typical BP waveform, modeled it adequately. BP parameters could be continuously calculated by both the c-FLOW and other vital sign monitors (such as Phillips intellivue MP50), indicating that it can serve as a reliable BP simulator for multiple purposes. Proper acousto-optic modeling of CBF using the described phantom was previously proved and validated[30, 31]. Thus, both MAP and CFI trends were reliably modeled. Though modeling CBF adequately, it is important to notice that the acousto-optic flow phantom utilized in the experiment is not ideal. Small instantaneous changes in flow within its channels may occur due to the syringe pump accuracy in flow supply or the uneven dispersion of scatterer centers within the flowing fluid. This factors may contribute to the relatively high CFI coefficient of variance obtained for the stable flow in sensor 2 (CV = 0.2). Obtained results confirmed the algorithm ability to distinguish cases in which both MAP and CFI trends are changing concurrently, and differ them from cases in which CFI does not track MAP changes using the cross correlation based ARI (Fig 5). Theoretically, ARI values range between 0–100, with 0 representing no correlation and 100 representing a perfect one. In our in-vitro experiment, ARI values of 37.66±17.99 and 61.74±21.36 were associated with AR and no AR (NAR) conditions, respectively. These absolute obtained values are not compatible with the hypothetical values of 0 versus 100, however, the two groups were significantly different (P<0.001). Further work should be dedicated to noise reduction and expansion of the obtained ARI values range to improve and enhance its ability to differ between the AR conditions. Based on previous studies to track AR utilizing correlation between MAP and CBF[13, 14, 25, 28, 29, 39], the above results support the use of the calculated ARIs for autoregulation monitoring. The experimental setup described here, combining both BP and flow modeling, is modularly built enabling simultaneous control on each of its parameters independently. Thus, in future studies it may be used to simulate more complicated protocols, different disease states, and various relationships between CBF and MAP. It may be utilized to further investigate the algorithm and predict its response to different conditions such as CBF change without changes of MAP (as can happen in severe head injury[40]) or delayed response of CBF to a MAP change (as in hypothermia cases when CBF becomes uncoupled from the metabolic rate[41] or in case of delay in autoregulatory action depending on PaCO2 [2]), etc. Additional validation for the c-FLOW-AR ability to detect AR-related trends was provided in the swine case study. The upper limit of autoregulation was clearly identified by both slope analysis of CFI versus MAP scatter plot (Fig 5(right)), and using the calculated ARI which quantitatively differed MAP segments under and over the ULA (Fig 6). For this piglet, ULA was identified at a MAP of 100mmHg. This value may seem rather low comparing to the traditional autoregulation curve described in humans, ranging between 50mmHg to 150mmHg, however in previous studies performed with pigs similar values were reported[42]. Though ULA was distinctly evident in the presented case study, the LLA was not identified within the obtained MAP range. According to Fig 5, the lowest measured MAP was approximately 40mmHg. For this swine, lower MAP values weren’t attainable even while increasing Nitroprusside to a dose of 62ml/hr. In a study designed to describe changes of LLA according to temperature[43], LLA in normothermia was achieved at 38±8mmHg with LLA values of even lower than 35mmHg. We therefore suggest that the LLA might be lower than the lowest obtained BP in our study, hence it was not apparent. Further studies, using higher drug doses or different protocols, should be carried to identify LLA as well. Beyond visual examination and slope analysis for the CFI versus MAP scatter plot given in Fig 5(right), a quantitative evaluation was performed using the calculated autoregulation indexes. Observation on the binned averaged ARI in Fig 6, unambiguously and quantitatively differ the two mentioned MAP segments (under and over the ULA). ARI threshold value was selected to be 52, which is slightly higher than the values reported in previous studies (0.4–0.45, which correspond to 40–45 in our ARI units) [13, 14, 28]. Most studies were carried using CBF velocity as a surrogate to CBF rather than the measurement of CFI, thus are not necessarily comparable in terms of absolute correlation values and thresholds. Yet, needless to say that in order to establish a reliable and repetitive threshold value for CFI and MAP correlaion index, more statistics should be obtained. The preclinical case study presented here provides a good evidence for the ability of the c-FLOW-AR to provide real-time autoregulation index in-vivo. This is only a preliminary validation of this device performance and recurrent studies should provide more statistics. As several other studies[28, 29, 39] were already conducted to evaluate autoregulation using UT-NIR technology for CBF monitoring with the difference of having an independent MAP tracking, similar results are expected using the integrated device. The c-FLOW-AR has a great potential in outlining autoregulation and becoming useful in clinical practice for individualized management of BP, and improving patients outcome. Studies in cardio vascular surgeries suggested that it might be able to predict postoperative complications, such as acute kidney injury[39] or delirium[44]. They further indicated that individualizing MAP management during Cardio Pulmonary Bypass surgery based on CBF autoregulation monitoring holds promise as an innovative approach to ensure organ perfusion and reduce the risk of postoperative complications after cardiac surgery. ==== Refs References 1 Lassen NA . Cerebral blood flow and oxygen consumption in man: Am Physiological Soc ; 1959 . 2 Aaslid R , Lindegaard K-F , Sorteberg W , Nornes H . Cerebral autoregulation dynamics in humans . Stroke . 1989 ;20 (1 ):45 –52 . 2492126 3 Paulson O , Strandgaard S , Edvinsson L . Cerebral autoregulation . Cerebrovascular and brain metabolism reviews . 1989 ;2 (2 ):161 –92 . 4 Strandgaard S , Sengupta D , Mackenzie E , Rowan J , Olesen J , Skinhøj E , et al The lower and upper limits for autoregulation of cerebral blood flow Cerebral circulation and metabolism : Springer ; 1975 p. 3 –6 . 5 Paulson O , Waldemar G . Role of the local renin-angiotensin system in the autoregulation of the cerebral circulation . Journal of Vascular Research . 1991 ;28 (1–3 ):231 –5 . 6 Strandgaard S , Olesen J , Skinhøj E , Lassen N . Autoregulation of brain circulation in severe arterial hypertension . BMJ . 1973 ;1 (5852 ):507 –10 . 4692673 7 Immink RV , van Montfrans GA , Stam J , Karemaker JM , Diamant M , van Lieshout JJ . Dynamic cerebral autoregulation in acute lacunar and middle cerebral artery territory ischemic stroke . Stroke . 2005 ;36 (12 ):2595 –600 . 16254228 8 Czosnyka M , Smielewski P , Piechnik S , Steiner LA , Pickard JD . Cerebral autoregulation following head injury . Journal of neurosurgery . 2001 ;95 (5 ):756 –63 . 11702864 9 Czosnyka M , Smielewski P , Kirkpatrick P , Laing RJ , Menon D , Pickard JD . Continuous assessment of the cerebral vasomotor reactivity in head injury . Neurosurgery . 1997 ;41 (1 ):11 –9 . 9218290 10 Brady K , Andropoulos DB , Kibler K , Easley RB . A New Monitor of Pressure Autoregulation: What Does It Add? Anesthesia & Analgesia . 2015 ;121 (5 ):1121 –3 .26484451 11 Steiner LA , Czosnyka M , Piechnik SK , Smielewski P , Chatfield D , Menon DK , et al Continuous monitoring of cerebrovascular pressure reactivity allows determination of optimal cerebral perfusion pressure in patients with traumatic brain injury . Critical care medicine . 2002 ;30 (4 ):733 –8 . 11940737 12 Lang EW , Lagopoulos J , Griffith J , Yip K , Mudaliar Y , Mehdorn HM , et al Noninvasive cerebrovascular autoregulation assessment in traumatic brain injury: validation and utility . Journal of neurotrauma . 2003 ;20 (1 ):69 –75 . 12614589 13 Joshi B , Brady K , Lee J , Easley B , Panigrahi R , Smielewski P , et al Impaired autoregulation of cerebral blood flow during rewarming from hypothermic cardiopulmonary bypass and its potential association with stroke . Anesthesia & Analgesia . 2010 ;110 (2 ):321 –8 .20008083 14 Brady K , Joshi B , Zweifel C , Smielewski P , Czosnyka M , Easley RB , et al Real-time continuous monitoring of cerebral blood flow autoregulation using near-infrared spectroscopy in patients undergoing cardiopulmonary bypass . Stroke . 2010 ;41 (9 ):1951 –6 . 10.1161/STROKEAHA.109.575159 20651274 15 Brady KM , Lee JK , Kibler KK , Smielewski P , Czosnyka M , Easley RB , et al Continuous time-domain analysis of cerebrovascular autoregulation using near-infrared spectroscopy . Stroke . 2007 ;38 (10 ):2818 –25 . 17761921 16 Panerai RB . Assessment of cerebral pressure autoregulation in humans-a review of measurement methods . Physiological measurement . 1998 ;19 (3 ):305 9735883 17 Panerai RB . Cerebral autoregulation: from models to clinical applications . Cardiovascular Engineering . 2008 ;8 (1 ):42 –59 . 18041584 18 Czosnyka M , Piechnik S , Richards HK , Kirkpatrick P , Smielewski P , Pickard JD . Contribution of mathematical modelling to the interpretation of bedside tests of cerebrovascular autoregulation . Journal of Neurology, Neurosurgery & Psychiatry . 1997 ;63 (6 ):721 –31 . 19 Kirkham S , Craine R , Birch AA . A new mathematical model of dynamic cerebral autoregulation based on a flow dependent feedback mechanism . Physiological measurement . 2001 ;22 (3 ):461 11556667 20 Van Beek AH , Claassen JA , Rikkert MGO , Jansen RW . Cerebral autoregulation: an overview of current concepts and methodology with special focus on the elderly . Journal of Cerebral Blood Flow & Metabolism . 2008 ;28 (6 ):1071 –85 .18349877 21 Tiecks FP , Lam AM , Aaslid R , Newell DW . Comparison of static and dynamic cerebral autoregulation measurements . Stroke . 1995 ;26 (6 ):1014 –9 . 7762016 22 Giller CA . The frequency-dependent behavior of cerebral autoregulation . Neurosurgery . 1990 ;27 (3 ):362 –8 . 2234328 23 Diehl RR , Linden D , Lücke D , Berlit P . Spontaneous blood pressure oscillations and cerebral autoregulation . Clinical Autonomic Research . 1998 ;8 (1 ):7 –12 . 9532415 24 Czosnyka M , Smielewski P , Kirk-patrick P , Menon D , Pickard JD . Monitoring of cerebral autoregulation in head-injured patients . Journal of neurosurgical anesthesiology . 1997 ;9 (2 ):200 . 25 Ono M , Joshi B , Brady K , Easley RB , Kibler K , Conte J , et al Cerebral blood flow autoregulation is preserved after continuous-flow left ventricular assist device implantation . Journal of Cardiothoracic and Vascular Anesthesia . 2012 ;26 (6 ):1022 –8 . 10.1053/j.jvca.2012.07.014 23122299 26 Schytz HW , Guo S , Jensen LT , Kamar M , Nini A , Gress DR , et al A new technology for detecting cerebral blood flow: a comparative study of ultrasound tagged NIRS and 133Xe-SPECT . Neurocritical care . 2012 ;17 (1 ):139 –45 . 10.1007/s12028-012-9720-2 22610823 27 Schwarz M , Rivera G , Hammond M , Silman Z , Jackson K , Kofke WA . Acousto-Optic Cerebral Blood Flow Monitoring During Induction of Anesthesia in Humans . Neurocritical care . 2015 :1 –6 . 28 Hori D , Hogue CW Jr, Shah A , Brown C , Neufeld KJ , Conte JV , et al Cerebral autoregulation monitoring with ultrasound-tagged near-infrared spectroscopy in cardiac surgery patients . Anesthesia & Analgesia . 2015 ;121 (5 ):1187 –93 .26334746 29 Murkin JM , Kamar M , Silman Z , Balberg M , Adams SJ . Intraoperative Cerebral Autoregulation Assessment Using Ultrasound Tagged Near Infrared-Based Cerebral Blood Flow in Comparison to Transcranial Doppler Cerebral Flow Velocity: A Pilot Study . Journal of Cardiothoracic and Vascular Anesthesia . 2015 . 30 Tsalach A , Schiffer Z , Ratner E , Breskin I , Zeitak R , Shechter R , et al Depth selective acousto-optic flow measurement . Biomedical optics express . 2015 ;6 (12 ):4871 –86 . 10.1364/BOE.6.004871 26713201 31 Racheli N, Ron A, Metzger Y, Breskin I, Enden G, Balberg M, et al., editors. Non-invasive blood flow measurements using ultrasound modulated diffused light. SPIE BiOS; 2012: International Society for Optics and Photonics. 32 Cheong W-F , Prahl SA , Welch AJ . A review of the optical properties of biological tissues . IEEE journal of quantum electronics . 1990 ;26 (12 ):2166 –85 . 33 Christensen D . Ultrasonic bioinstrumentation: wiley ; 1988 . 34 Lagerkranser M , Stånge K , Sollevi A . Effects of Propofol on Cerebral Blood Flow, Metabolism and Cerebral Autoregulation in the Anesthetized Pig . Journal of neurosurgical anesthesiology . 1997 ;9 (2 ):188 –93 . 9100192 35 Czosnyka M , Brady K , Reinhard M , Smielewski P , Steiner LA . Monitoring of cerebrovascular autoregulation: facts, myths, and missing links . Neurocritical care . 2009 ;10 (3 ):373 –86 . 10.1007/s12028-008-9175-7 19127448 36 Germon T , Kane N , Manara A , Nelson R . Near-infrared spectroscopy in adults: effects of extracranial ischaemia and intracranial hypoxia on estimation of cerebral oxygenation . British journal of anaesthesia . 1994 ;73 (4 ):503 –6 . 7999492 37 Murkin J , Arango M . Near-infrared spectroscopy as an index of brain and tissue oxygenation . British journal of anaesthesia . 2009 ;103 (suppl 1 ):i3 –i13 . 10.1093/bja/aep299 20007987 38 Lee JK , Kibler KK , Benni PB , Easley RB , Czosnyka M , Smielewski P , et al Cerebrovascular reactivity measured by near-infrared spectroscopy . Stroke . 2009 ;40 (5 ):1820 –6 . 10.1161/STROKEAHA.108.536094 19286593 39 Hori D , Hogue C , Adachi H , Max L , Price J , Sciortino C , et al Perioperative optimal blood pressure as determined by ultrasound tagged near infrared spectroscopy and its association with postoperative acute kidney injury in cardiac surgery patients . Interactive cardiovascular and thoracic surgery . 2016 :ivv371. 40 Gobiet W , Grote W , Bock W . The relation between intracranial pressure, mean arterial pressure and cerebral blood flow in patients with severe head injury . Acta neurochirurgica . 1975 ;32 (1–2 ):13 –24 . 1163313 41 Gravlee GP , Davis RF , Utley JR . Cardiopulmonary bypass: principles and practice: Lippincott Williams & Wilkins Philadelphia ; 2000 . 42 Pesek M , Kibler K , Easley RB , Mytar J , Rhee C , Andropoulos D , et al The upper limit of cerebral blood flow autoregulation is decreased with elevations in intracranial pressure . Neurosurgery . 2014 ;75 (2 ):163 –70 . 10.1227/NEU.0000000000000367 24739364 43 Lee JK , Brady KM , Mytar JO , Kibler KK , Carter EL , Hirsch KG , et al Cerebral blood flow and cerebrovascular autoregulation in a swine model of pediatric cardiac arrest and hypothermia . Critical care medicine . 2011 ;39 (10 ):2337 10.1097/CCM.0b013e318223b910 21705904 44 Hori D , Max L , Laflam A , Brown C , Neufeld K , Adachi H , et al Blood Pressure Deviations from Optimal Mean Arterial Pressure during Cardiac Surgery Measured with a Novel Monitor of Cerebral Blood Flow and Risk for Perioperative Delirium: A Pilot Study . Journal of Cardiothoracic and Vascular Anesthesia . 2016 .

PMC005xxxxxx/PMC5003386.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2757141510.1371/journal.pone.0161852PONE-D-16-15498Research ArticleBiology and Life SciencesDevelopmental BiologyDiapauseBiology and Life SciencesPhysiologyPhysiological ProcessesDiapauseMedicine and Health SciencesPhysiologyPhysiological ProcessesDiapauseBiology and Life SciencesAnatomyReproductive SystemOvariesMedicine and Health SciencesAnatomyReproductive SystemOvariesBiology and Life SciencesGeneticsGene ExpressionResearch and analysis methodsExtraction techniquesRNA extractionResearch and Analysis MethodsModel OrganismsAnimal ModelsDrosophila MelanogasterBiology and Life SciencesOrganismsAnimalsInvertebratesArthropodaInsectsDrosophilaDrosophila MelanogasterBiology and life sciencesMolecular biologyMolecular biology techniquesSequencing techniquesRNA sequencingResearch and analysis methodsMolecular biology techniquesSequencing techniquesRNA sequencingBiology and Life SciencesComputational BiologyGenome AnalysisTranscriptome AnalysisBiology and Life SciencesGeneticsGenomicsGenome AnalysisTranscriptome AnalysisBiology and Life SciencesBiochemistryMetabolismCarbohydrate MetabolismGlucose MetabolismTranscriptional Differences between Diapausing and Non-Diapausing D. montana Females Reared under the Same Photoperiod and Temperature Transcriptional Differences in D. montana Diapause in Critical PhotoperiodKankare Maaria 1*Parker Darren J. 12Merisalo Mikko 1Salminen Tiina S. 3Hoikkala Anneli 11 Department of Biological and Environmental Science, University of Jyväskylä, P.O. Box 35, Jyväskylä, Finland2 Centre for Biological Diversity, School of Biology, University of St Andrews, Fife, KY16 9TH, St Andrews, United Kingdom3 BioMediTech, Biokatu 6, F1-33014, University of Tampere, Tampere, FinlandEtges William J. EditorUniversity of Arkansas, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Data curation: DP. Formal analysis: DP MM MK. Funding acquisition: AH MK. Investigation: TS MM MK. Methodology: DP MK. Project administration: MK. Resources: AH MK. Supervision: MK AH. Validation: MK. Visualization: DP. Writing – original draft: MK MM. Writing – review & editing: MK DP TS AH. * E-mail: maaria.kankare@jyu.fi29 8 2016 2016 11 8 e016185217 4 2016 13 8 2016 © 2016 Kankare et al2016Kankare et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Background A wide range of insects living at higher latitudes enter diapause at the end of the warm season, which increases their chances of survival through harsh winter conditions. In this study we used RNA sequencing to identify genes involved in adult reproductive diapause in a northern fly species, Drosophila montana. Both diapausing and non-diapausing flies were reared under a critical day length and temperature, where about half of the emerging females enter diapause enabling us to eliminate the effects of varying environmental conditions on gene expression patterns of the two types of female flies. Results RNA sequencing revealed large differences between gene expression patterns of diapausing and non-diapausing females, especially in genes involved with metabolism, fatty acid biosynthesis, and metal and nucleotide binding. Differently expressed genes included several gene groups, including myosin, actin and cytochromeP450 genes, which have been previously associated with diapause. This study also identified new candidate genes, including some involved in cuticular hydrocarbon synthesis or regulation (desat1 and desat2), and acyl-CoA Δ11-desaturase activity (CG9747), and few odorant-binding protein genes (e.g. Obp44A). Also, several transposable elements (TEs) showed differential expression between the two female groups motivating future research on their roles in diapause. Conclusions Our results demonstrate that the adult reproductive diapause in D. montana involves changes in the expression level of a variety of genes involved in key processes (e.g. metabolism and fatty acid biosynthesis) which help diapausing females to cope with overwintering. This is consistent with the view that diapause is a complex adaptive phenotype where not only sexual maturation is arrested, but also changes in adult physiology are required in order to survive over the winter. Academy of Finland268214Kankare Maaria Academy of Finland272927Kankare Maaria Academy of Finland132619Hoikkala Anneli http://dx.doi.org/10.13039/501100000270Natural Environment Research CouncilStudentshipParker Darren J This work was supported by projects 268214 and 272927 to Maaria Kankare and project 132619 to Anneli Hoikkala, http://www.aka.fi/en, and NERC studentship to Darren J. Parker, http://www.nerc.ac.uk/. Data AvailabilityRNAseq data has been submitted to NCBI's Gene Expression Omnibus. The transcriptome assembly has been deposited at DDBJ/EMBL/GenBank under the accession GECM00000000. The version described in this paper is the first version, GECM01000000. Raw reads have been deposited in GEO under accession codes: SRR2910695, SRR2910698, SRR2910701, SRR2910702, SRR2910703, and SRR2910704.Data Availability RNAseq data has been submitted to NCBI's Gene Expression Omnibus. The transcriptome assembly has been deposited at DDBJ/EMBL/GenBank under the accession GECM00000000. The version described in this paper is the first version, GECM01000000. Raw reads have been deposited in GEO under accession codes: SRR2910695, SRR2910698, SRR2910701, SRR2910702, SRR2910703, and SRR2910704. ==== Body Introduction Seasonally changing environmental conditions pose great challenges for organisms inhabiting northern latitudes, and many species enter some level of dormancy to survive over the cold period. One of the most common types of dormancy in insects is diapause, where growth and/or reproduction are halted over winter and synchronized to resume in a more favorable season [1]. At northern latitudes, seasonal changes in photoperiod are the most reliable signal for forthcoming environmental changes so many insects use this signal to coordinate their diapause [2]. Diapause is considered a dynamic pathway induced by specific environmental cues and it involves various phases [3] and modules including photoperiodism, hormonal events as well as diapause itself [4]. Photoperiodic signals are effective in evoking diapause only during the sensitive period of an insect’s life cycle which is species-specific and may occur well before the actual diapause stage [1]. Seasonal timing of diapause is often defined by measuring the critical photoperiod/day length (CDL): the photoperiod where a diapause response measured by the size of the ovaries is observed in approximately half of the females of a given population, while the other half continues to sexual maturation [5]. CDL is typically a narrow time period because even small changes in day length in either direction can increase or decrease diapause incidence [6]. Consequently, local selection for the optimal timing of diapause at different latitudes can lead to population-specific CDLs and give rise to clines in this trait despite high gene flow between populations [7,8,9]. A longer growing season induced by global warming has already been shown to favour shorter southern CDLs over longer northern ones in several species [10,11]. The sensitive period for diapause induction is followed by a preparative phase during which diapause-destined individuals change their behavioral [12] and feeding patterns [13] in preparation for the adverse season. In adult reproductive diapause, nutrient reserves are accumulated in fat bodies, mainly as lipids (triacylglycerides) [14] but also as glycogen [15] and storage proteins [1,16], at the expense of ovarian development (e.g. [17]). It is critical that diapausing individuals gather enough energy reserves in advance, since feeding is often greatly reduced, if not arrested, during the actual diapause phase, and insufficient reserves can affect both diapause entry and termination [18]. In addition to nutrient reserves, molecular chaperones and cryoprotectants, including heat shock proteins and glycerol, are often synthesized to protect proteins and tissue from stressful conditions such as freezing or desiccation [19,20]. During the actual diapause phase, metabolic activity is suppressed [21] and energy usage is transferred from costly tissues such as flight muscle [22] to more critical systems like the brain [23]. As a result, diapausing individuals are more stress-tolerant and live longer than non-diapausing ones [24,25]. One factor impeding an understanding of insect diapause is variation of the phenotype which may at least be partly because diapause has evolved independently in different insect species [26,27]. Much research on diapause has been directed towards economically important species, such as the silkworm (B. mori), Colorado potato beetles (Leptinotarsa decemlineata), Rice stem borer (Chilo suppressalis) and mosquito species that are disease vectors (e.g. Aedes aegypti, Culex pipiens), with emphasis on the ecological and physiological aspects of diapause. In contrast, the underlying molecular and genetic causes of diapause are less well known [1,28]. The discovery of adult reproductive diapause in Drosophila melanogaster [29] has enabled in-depth research on the genetics of diapause [30–34]. However, the diapause response in D. melanogaster is shallow, recent in origin, observed only under certain temperatures, and it never reaches a full 100 percent response, i.e. not all females enter into diapause [35,36]. Therefore, wild Drosophila species that evolved a robust diapause response in temperate and northern latitudes [37] may be better suited for studies on diapause genetics than D. melanogaster [1] and bring a new insights and dimensions to these studies. Here we show gene expression changes linked with adult reproductive diapause at the whole transcriptome level in a northern fly species, Drosophila montana, a member of the Drosophila virilis group. D. montana diverged from D. virilis approximately 9 million years ago [38] and from D. melanogaster approximately 63 million years ago [39]. D. montana females overwinter in an adult reproductive diapause, where ovarian development is halted at the pre-vitellogenic stage [40]. The CDL of D. montana varies along a latitudinal cline and decreases by ca 1.6 h from northern to southern populations in Finland (67–61°N), documenting a strong photoperiodic response and local adaptation even in the presence of high gene flow [9,41]. The flies of this species are also very cold-tolerant [42] and show adaptive changes in their daily and annual locomotor activity rhythms [43]. We performed RNA sequencing (RNAseq; transcriptome) analyses for diapausing (D) and non-diapausing (ND) D. montana females reared at the same CDL and temperature, where about half of the emerging females will enter diapause, to identify genes that are differentially expressed during reproductive diapause. This experimental design enabled us to eliminate the effects of different environmental conditions on gene expression patterns of D and ND females, a comparison that has been mostly overlooked in previous studies. D and ND females differ dramatically in the relative sizes of their ovaries. Since we used whole-body extractions and this ‘tissue heterogeneity’ could bias our RNAseq results, we performed additional analysis to estimate the extent of bias generated by the size of the ovaries. First, we examined whether the genes that showed differential expression between the two female types in our study were enriched for ovary genes. Second, we conducted qPCR analyses for D and ND females whose ovaries had been removed using a set of genes that had shown differential expression in the RNAseq analysis. Material and Methods Sample preparation Our study material consisted of D. montana females from three isofemale strains (3OL8, 175OJ8 and 265OJ8), each of which was founded from the offspring of a single fertilized female fly collected from Oulanka, Finland (66°N) in 2008. No permission is needed for collecting Drosophila flies in Oulanka (Finland). Since their founding, these strains have been maintained in the laboratory with overlapping generations under diapause preventing conditions (constant light, 19°C and 60% humidity) in plastic bottles containing malt medium [44]. Females of these strains were collected for RNA sequencing in 2010, i.e. after about 16 generations of laboratory maintenance. Diapause response of D. montana females is genetically determined and does not change under constant light in the laboratory [40]. Virgin, female flies were collected from the malt bottles within one day after eclosion using light CO2 anaesthesia and transferred in malt vials into a climate chamber (Sanyo MLR-351H). Flies were reared in this chamber in a light:dark cycle of 18.5:5.5, which represents the critical day length for diapause induction (CDL) for females [9] and corresponds to the beginning of August in the Oulanka population. In our study, ca 10% to 20% of about 50 females dissected in each isofemale strain reared in this CDL showed an intermediate phenotype [9]; all of these females were discarded and we only used D and ND flies that were easiest to identify. The reproductive state of females was determined by submerging frozen females into RNAlaterICE (Ambion) and dissecting them under a light microscope. The females with pre-vitellogenic, small and transparent ovaries with no yolk accumulation or visible segments were classified as diapausing (D) and the females with large vitellogenic ovaries with visible eggs as non-diapausing (ND) [45]. Females that had intermediate ovaries with some yolk accumulation and visible segments, but no eggs, were discarded. Ten D and ten ND females from each of the 3 strains were pooled to create 3 biological samples of both female types for RNAseq (each strain made up one sample). For the qPCR analysis, we used the RNA extracted from single D and ND females whose ovaries had been removed before RNA extraction. All samples were collected from the environmental chamber at the same time. RNA extraction and RNAseq RNA extraction was performed using a Tri Reagent (Sigma-Aldrich) extraction kit followed by a RNeasy Mini (Qiagen) kit with DNase treatment. RNA concentration and purity was checked using NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies) and its integrity using a 2100 Bioanalyzer (Agilent Technologies). A MicroPoly(A) Purist kit (Ambion AM1919) was used to enrich mRNA from total RNA sample. Sequencing libraries were constructed from each sample using a SOLiD total RNAseq Kit with unique barcodes (SOLiD Transcriptome Multiplexing Kit) and sequenced in the Finnish Microarray and Sequencing Centre (Turku, Finland) using a SOLiD 5500XL Genetic Analyzer (Applied Biosciences) to generate 75 base pair forward and 35 base pair paired-end reads. Raw sequence reads were trimmed using SOLiD TRIM (with run options: -p 3 -q 22 -y y -e 2 -d 10) to remove polyclonal errors from the data [46] and the reads that passed this filter were corrected using SOLiD Accuracy Enhancer Tools (SAET) to reduce the amount of colour calling errors/erroneous bases in the sequence. Finally, any remaining low quality bases at the end of the reads were trimmed using CLC Genomics Workbench 5.0.1 (CLC Bio http://www.clcbio.com/) (quality score: 0.02). De novo assembly Since D. montana does not currently have a reference genome available, a de novo assembly of the transcriptome was produced using CLC Bio (available from http://www.clcbio.com/) with default settings. To produce a reference assembly, we used the reads described above, along with the reads produced in two other D. montana transcriptome projects on the effects of cold acclimation [47] and different photoperiods [48] on gene expression. Annotation of the contigs Contigs were annotated using Blast2GO [49]. Specifically, all contigs were blasted (blastX) to the non-redundant protein sequence (nr) database. Contigs without a significant blast hit (E-value >0.001) were then blasted (blastN) to non-redundant nucleotide collection (nt). Contigs that still had no hits were blasted (blastN) against the Reference Sequence (RefSeq) genomic database. Contigs that blasted to non-arthropods or rRNA genes were discarded prior to mapping, while contigs that did not get any blast hits were kept in the reference assembly as they could still represent functional genes. Read mapping and gene expression analysis Reads for each sample were mapped individually to the de novo reference assembly using CLCBio. HTSeq [50] was used to quantify the number of uniquely mapping reads to each contig. Differential expression (DE) of contigs between D and ND females was calculated using DESeq package (version 1.9.4, [51] in R (v. 2.14.2; R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/). Variation in library size caused by differences in sequence depths between the samples was taken into account by performing between-library normalization. After this, a generalized linear model (GLM) with a negative binomial distribution was fitted in DESeq with diapause state as a factor. The P values from the GLM were corrected for multiple testing using Benjamini and Hochberg's algorithm [52] to control for false discovery rate (FDR) at P < 0.05. Functional classification DAVID (v. 6.7) [53] was used to carry out functional clustering for DE contigs at P < 0.0001. Functional categories of the annotation terms over-represented in the study list were ranked based on their level of enrichment defined as the geometric mean of P values for each annotation term within the group [53]. All the contigs with identified orthologs were used as the ‘background’ for comparisons. Estimation of the effect of tissue heterogeneity through enrichment analysis Substantial differences in the tissue composition of the study samples may cause mis-estimation of gene expression differences between samples [54]. In order to determine whether tissue heterogeneity in the ovary size of D and ND females could explain gene expression differences between females, we examined whether our differentially expressed (DE) genes were enriched for ovary genes. To do this, we used the modENCODE high-throughput RNAseq data from D. melanogaster [55] to classify D. montana orthologs as having low, medium, high, or very high ovary expression. We then compared the proportion of these genes in our DE genes (contigs) to the proportion in all genes, using a one-sided Fisher’s exact test implemented in R (R Core Team, 2013). Estimation of the effect of tissue heterogeneity with qPCR To examine the effects of tissue heterogeneity on the RNAseq data, we also performed qPCR runs with a set of upregulated genes. We performed qPCR runs for D and ND females collected at the same time as the original samples whose ovaries had been removed. Total RNA was extracted from single females without ovaries using a ZR Tissue & Insect RNA MicroPrep Kit with DNAse treatment (Zymo Research), using six females (replicates) from each of the three strains. Purity and integrity of RNA was measured with NanoDrop and TapeStation (Agilent) and cDNA was generated using equal concentrations of RNA (400 ng/μl) and iScript Reverse Transcription Supermix (Bio-Rad Laboratories). The five genes for the qPCR analyses were selected among new potential candidate genes or on the basis of their earlier detected function in diapause or in cold tolerance (see below), and most of them were picked from the two gene clusters showing the highest number of significantly DE genes in our transcriptome analysis (see results). Desaturase 1, desat1, and cyp12a5, a cytochrome P450 gene were among the largest annotated gene cluster of 285 DE metal ion binding genes. From these, desat1 has been connected to cold tolerance (e.g. [56]) and couch potato (cpo), from the nucleotide binding (242 DE genes) cluster, to diapause by Schmidt et al [57] including our earlier studies [58]. Odorant binding protein 44A, Obp44A, and period, per were not present in any of the functional clusters identified by DAVID, but were upregulated in diapausing females. Earlier studies have also shown per to be upregulated in the initiation, maintenance and overwintering stages in diapausing D. montana females [59]. We tested several control genes in the samples and Ribosomal protein L32 (RpL32) and 18SrRNA (18S) were selected, as they showed least variation in their expression level in qPCR. Primers for the control and experimental genes were designed using D. montana sequences and NetPrimer (http://www.premierbiosoft.com/netprimer/). Amplification efficiency of the primers was analysed using 2-fold serial dilutions of pooled cDNA from all the samples (S2 Table). The 20 μl reaction mixes for qPCR contained 10 μl of 2x Power SYBR Green PCR Master Mix (Bio-Rad Laboratories), 0.3 μM of each gene-specific primer and 1 μl of cDNA solution. Cycling conditions in Bio-Rad CFX96 instrument were: 3 min. 95°C, 10 s. 95°C, 10 s. 55°C and 30 s. 72°C (40x), followed by melting curve analysis (65°C-95°C) for amplification specificity checking. Gene expression values for all comparisons were calculated with normalized expression method (ΔΔ(Ct) [60]) using two control genes and real efficiency values. Statistical significance of the expression differences between different comparisons was tested using an ANOVA in R. Results Transcriptome assembly and gene expression analysis RNA sequencing produced approximately 37 million paired-end reads of 75 and 35 bases in length. Trimmed reads were assembled into 31880 contigs with a N50 of 527 and the mean contig length of 471 (minimum contig length was 200 bases). We obtained blast results for 99% of the contigs assembled from the RNA sequencing data, and out of them close to 82% blasted to known genes and over 14% to genomic scaffolds in the RefSeq database. As expected, most of the blast hits (>25 000 contigs) were to sequences from D. virilis, a close relative of D. montana with a well annotated genome available. Almost all of the remaining hits were to other Drosophila species and less than 2% of contigs (647) were discarded as possible contaminants, due to their hit non-arthropod sequences. 321 contigs did not get a significant blast hit, and were retained. Of the 37 million paired-end reads obtained approximately 41% mapped uniquely to the reference transcriptome. We found around a third of the contigs tested (11178) were significantly DE between the two female types. Of these, around half of the contigs were upregulated (5353) and half downregulated (5825) between ND and D females (Fig 1, S1 Table). 10.1371/journal.pone.0161852.g001Fig 1 Log2 fold change versus FKPM (Fragments Per Kilobase Million) for each contig. Significantly DE contigs (FDR < 0.05) are colored in red. Cluster groups created with the GO Term enrichment analysis Using DAVID we identified a total of 15 significantly enriched gene clusters (Enrichment score > 1.3, which corresponds to P < 0.05) between D and ND females (Table 1, significant orthologs in each of the clusters are given in Tables A-O in S1 File). Many of these clusters (1, 3, 6, 11, 12 and 14) were connected to DNA replication and are likely to be related to different oogenesis processes in D and ND females. Moreover, cluster 9 included down-regulated genes annotated with tudor domain protein features, including the tudor gene itself. Tudor genes are known to be involved in germ cell development and oogenesis, reflecting the differences in oocyte maturation between D and ND females. The rest of the 8 clusters were more interesting, and they are likely to include several genes with a direct link to diapause (see discussion). Two of these clusters were connected to metabolism, cluster 10 to glycolysis and glucose metabolism and cluster 13 to lipid and fatty acid biosynthesis, and one, cluster 5, to protein transporter activity. These clusters most probably represent metabolic changes occurring in D flies when they are preparing for the cold season. The remaining five clusters included two metal binding clusters (clusters 8 and 15), an actin binding cluster (cluster 2), a nucleotide binding cluster (cluster 4) and an oxidative stress cluster (cluster 7) (Table 1). 10.1371/journal.pone.0161852.t001Table 1 Functional clusters of the DE genes. Cluster No. No. of contigs E score P-value GO terms Up/downregulated genes 1 104 6,82 1,51E-07 DNA replication, nucleic acid binding Nf1/DNApol-α180 2 147 2,89 0,001 Actin binding, cytoskeletal protein binding Mch, Act/CG18190 3 98 2,73 0,002 ATP-dependent helicase activity futch/CG11403 4 606 2,22 0,006 Nucleotide, ATP and nucleoside binding Hsp70Cb, cpo, Act/greatwall 5 172 2,1 0,008 Protein transporter activity Mhc/CG3509 6 14 1,89 0,013 Nuclear chromosome, sallimus/DNApol-α50 7 12 1,71 0,019 Oxidoreductase activity, response to oxidatative stress Irc/Pxt, Duox 8 22 1,71 0,019 Metal and iron-sulfur binding CG8102/Acon 9 28 1,7 0,02 Maternal tudor protein CG17454/CG15930 10 42 1,58 0,026 Glycolysis, glucose metabolic process Zw/CG4747 11 33 1,5 0,032 Replication fork, DNA replication factor C complex comt/CG11403 12 42 1,47 0,034 DNA polymerase and nucleotidyltransferase activity mRNA-cap/DNApol-α180 13 34 1,43 0,037 Fatty acid and lipid biosynthesis processes desat1, 2, Inos, CG9747/CG33116 14 69 1,4 0,04 DNA packaging, chromatin and nucleosome organization sallimus/CG3509 15 648 1,33 0,047 Metal and zinc ion binding, zinc finger desat1,2, cyp12a5, CytP450/Phf7, Pxt Functional clusters of the DE genes (numbers given in contigs) showing differential expression between diapausing and non-diapausing females from DAVID [53], P < 0.0001. Examples of up- and down-regulated genes are given in the last column and genes marked with bold are discussed in more details in the discussion. All significantly DE genes in each of the clusters are given in Tables A-O in S1 File. Tissue heterogeneity: enrichment of the ovary genes We estimated the effects of tissue heterogeneity due to ovary size on the gene expression differences between D and ND females by investigating whether DE genes were enriched for ovary expression. We found that genes categorized as having either medium, high, or very high levels of ovary expression were overrepresented in gene lists identified as DE between D and ND females (Fig 2). This finding suggests that a proportion of the genes identified as being DE were likely to due to the heterogeneous nature of the tissues compared, rather than DE. 10.1371/journal.pone.0161852.g002Fig 2 Percentage of contigs with a low, medium, high or very high expression level in female ovaries in our transcriptome data (background, black) compared to the ones found to be DE between diapausing and non-diapausing females (white). P values are from a one-sided Fisher’s exact test, see text for details: NS = P > 0.05, ***P < 0.001. Validation of the RNA sequencing data and estimation of the effect of tissue heterogeneity with qPCR To determine whether tissue heterogeneity between D and ND females influenced the expression patterns of the selected candidate genes, we performed qPCR runs for individual females whose ovaries had been removed. qPCR produced similar upregulation patterns as observed in the RNAseq for three out of five studied candidate genes (Fig 3), with significant expression level differences between the two female types for cyp12a5 (P = 0.00001), desat1 (P = 0.00026) and per genes (P = 0.0078) and suggestive differences in Obp44 (P = 0.07737). The same pattern was also seen in cpo, but the difference remained non-significant (P = 0.15457) due to high variance between the samples (Fig 3). 10.1371/journal.pone.0161852.g003Fig 3 Estimating the effect of the tissue heterogenity. Normalized gene expressions levels comparing diapausing and non-diapausing D. montana females A) from the RNAseq analysis (for comparison), and B) from the qPCR using the females whose ovaries had been removed before RNA extractions. Both data sets present the combined data for all the three isofemales strains (see text for details). CPM = counts per million. Significance levels for qPCR: •0.1 > P > 0.05, **0.01 > P > 0.001, ***P < 0.001. Transposable elements We found 15 unique TE groups to be significantly DE between D and ND D. montana females. The majority of the groups (e.g. polyproteins and retrotransposons) showed upregulation and only three (e.g. Xanthias, Hobo) were downregulated in D females while different contigs annotated as Gypsy protein showed both up and down regulation (see S3 Table for more details). Discussion Adult photoperiodic reproductive diapause is a common over-wintering strategy in temperate insect species. Here, we identified functional gene clusters and genes involved in diapause in D. montana by comparing gene expression differences between diapausing and non-diapausing females reared under the same environmental conditions. We found that around a third of the contigs in our transcriptome were DE, showing that diapause has a profound effect on female D. montana gene expression. This was somewhat expected as diapause is a complex trait, known to cause large shifts in reproductive state, cold tolerance, and behavior. Since D and ND females differed dramatically in the relative sizes of their ovaries, some of the contigs we identified as being DE were a result of comparing heterogeneous tissue types. Consequently, as we expected that genes over expressed in ovaries would be more likely to be affected by this issue, we determined whether DE contigs were enriched for genes expressed in the ovary. We found that our DE contigs were enriched for genes expressed at least at moderate levels in the ovary suggesting that some of these contigs were likely to be false positives due to the influence of tissue heterogeneity. Despite this, we note that the magnitude of the overrepresentation was relatively small (Fig 1), and in fact it could also be expected because contigs involved in reproduction would be also involved in controlling reproductive diapause. In addition, we examined changes in expression in ND and D females when their ovaries had been removed for a few key genes and found that the differences were similar to those obtained from RNAseq. Taken together, this suggests that tissue heterogeneity may be responsible for a small but significant proportion of the contigs identified as DE, and hence when evaluating the results of the present study, we focused on the gene clusters/genes with no direct connection to ovaries or/and oogenesis. There were 15 significantly enriched gene clusters, where several genes showed differential expression patterns between D and ND females. Three of these clusters were involved with glycolysis and glucose metabolism, fatty acid and lipid biosynthesis and protein transporter activity, all of which are suggested to represent metabolic changes occurring in diapausing flies when they are preparing for the cold season. In addition, five of the clusters were connected to metal, nucleotide and actin binding and/or to oxidative stress, which may also play a role in diapause. All clusters contained new candidate genes for diapause in addition to the previously identified genes. The last seven clusters were connected to DNA replication and are thus more likely to be linked with oogenesis rather than the diapause per se. The lists of all genes in each of the clusters are given in Tables A-O in S1 File. Genes connected to metabolic changes when diapausing flies are preparing for the cold season In the current study, DE orthologs involved with glycolysis and glucose metabolic processes are most probably linked to metabolic changes which occur when the flies begin to go into diapause. The most common cryoprotectant is glycerol, but many other polyols and sugars, such as glucose, can also function in this way [61]. Moreover, the function of glucose-6-phosphate dehydrogenase (G6PDH) coded by Zw in Drosophila affects the synthesis of cryoprotectant polyols in insect larvae [62] and this gene was found to be upregulated in D females in our study. Overexpression of Zw has been also associated with increased life span in D. melanogaster [63] and G6PDH is a key enzyme in NADPH synthesis which has been suggested to directly function as an antioxidant [64]. Diapausing D. montana females are known to be more cold tolerant than the non-diapausing ones, and decreases in photoperiod and/or temperature has also been found to increase cold tolerance [65]. Glucose is known to be stored in the diapausing flies mainly as glycogen. In our earlier metabolic analyses on D. montana, glucose levels of D females were found to increase almost 2-fold in autumn [65]. However, glucose may not to be the major metabolite in diapausing D. montana, as myo-inositol increased up to 144-fold during the winter in both sexes of this species [65]. We have also found that Inos, which encodes the enzyme myo-inositol-1-phosphate synthase, is connected to cold acclimation in D. montana and D. virilis [47]. Because Inos was up-regulated in D females compared to the ND ones in a constant environment, we suggest that the expression of this gene can increase during diapause without a decrease in temperature. Inos has also been connected to increased cold tolerance in some Coleoptera and Lepidoptera species [66,67], to egg diapause in migratory locusts [68] and to pupal diapause termination in the flesh fly Sarcophaga crassipalpis at the protein level [69]. Desaturase genes (desat1 and desat2) function in fatty acid metabolism and lipid metabolic processes, and were found to be upregulated in diapausing D. montana females. The finding is in accordance with our earlier microarray studies, where desat1 and desat2 were upregulated in D. montana females at the initiation stage, and desat2 in the termination phase of diapause [56]. The same genes were found to be down-regulated in cold acclimated non-diapausing flies, desat1 in D. montana and desat2 in D. virilis [56]. Desat2 is also known to play a role in desiccation [70] and cold and starvation tolerance [71]. Reynolds and Hand [72] showed upregulation of Δ9-desaturase during diapause in the cricket Allonemobius socius, and Sim and Denlinger [73] detected similar increases in Culex pipiens. Desaturases have also been found to play a role in pheromone biosynthesis in insects (e.g. [74]) which may explain why diapausing D. montana females are not attractive to males [75]. We also found another gene with acyl-CoA Δ11-desaturase activity, CG9747, to be upregulated in diapausing females. This gene has been found to show high expression levels in the initiation and early maintenance phases of D. montana diapause [59] and recently Kučerová et al. [76] have linked it to diapause in D. melanogaster. Desaturases have also been shown to increase membrane lipid unsaturation and maintenance of membrane fluidity in starvation-induced autophagy in Drosophila [77]. Finally, in addition to desaturases, we found that six odorant binding protein genes showed differential expression between D and ND females (S4 Table), although none of them were found to belong in any of the significantly enriched gene clusters in this study. Connection of odorant genes to diapause clearly demands future studies. Cytochrome P450 genes form another important group of genes involved in the metabolism of steroids and fatty acids [78] and some of these genes have been connected to larval diapause in the silkmoth, Antheraea yamamai [79]. We found 13 cytochrome P450 orthologs (e.g. cyp12a5) from the metal ion binding cluster (cluster 15, Table O in S1 File) that showed different expression levels in D and ND females, suggesting that these genes may function in D. montana diapause. Interestingly, cytP450 genes have been associated with insecticide resistance [80,81], stress resistance in aging flies [82] and immunity (e.g. [83]). Insecticide and stress resistance, as well as immunity, may share many metabolic pathways and pleiotropic genes, but more in-depth studies are needed to clarify the connection between these traits and diapause. Myosin heavy chain (Mhc) and several actin genes (from protein transport, nucleotide and actin binding clusters) were upregulated in D females in our study. Myosin is associated with muscles, where it drives muscle contraction by a cyclical interaction with actin, and both above-mentioned genes also play a role in the induction of structural changes in the cytoskeleton. Cytoskeleton structure changes at low temperatures and this has been found to increase cold tolerance in plants [84,85]. In insects, low temperatures have been found to induce changes in the distribution and structure of actin [86]. In diapausing individuals these changes can be even more pronounced than in non-diapausing ones, and are thus expected to be accompanied/caused by the upregulation of actin genes [86,87]. Interestingly, in D. americana, ActinD1 expression levels have been also shown to reflect observed differences in life span of both diapausing and non-diapausing flies [88]. Molecular chaperones and cryoprotectants, like the heat shock proteins, are often synthesized to protect proteins and tissues in stressful conditions, such as freezing or desiccation [19,89], and different HSPs have been found to be active in diapausing individuals (e.g. [19,42,59,90]. However, we did not detect any significantly enriched clusters containing stress-related genes even though several heat shock protein (Hsp) genes were significantly DE in diapausing flies (S5 Table). Only two Hsp genes (Hsp22 and Hsp67Bc) were significantly upregulated in diapausing D. montana females. Interestingly, Kučerová et al. [77] also found Hsp67Bc to be upregulated while Hsp22 was downregulated together with other small heat shock proteins (Hsp23, Hsp26 and Hsp27) in diapausing D. melanogaster. In our study, two Hsps (Hsp23 and Hsp26) were downregulated in diapausing females. Furthermore, several heat shock cognate genes (e.g. Hsc70 group genes) showed altered levels of expression in the present study (S5 Table) compared to our previous microarray study [59] where only Hsc70-2 showed upregulation during different phases of diapause. Clearly, heat shock genes do not seem to constitute a homogeneous group of genes, but rather to show both up- and down-regulation depending on the taxonomic group and environmental conditions in the species and population home site. DE genes showing multiple transcripts Expression patterns were determined for the original contigs instead of genes, because there were more than 4000 contigs that matched only to general genomic sequences in our study. It is likely that many of them are actually transcripts from a known gene, which fall outside the current gene model boundaries. In fact, further examination found that several contigs in our assembly contained regions that aligned to regions annotated as introns in other species and many of them were from genes already identified in our transcriptome. Thus, it is likely that several of our contigs represent alternative transcripts of a single gene. For example, trehalose-6-phosphatase synthase 1 (Tps1), CG1648 (molecular function not known) and CG6426 (having lysozyme activity), showed expression differences in some contigs but not in others. Among these genes Tps1 has earlier been connected to diapause of Onion maggots [91] and trehalose is one of the metabolites connected to the seasonal improvement of cold tolerance in D. montana [65]. On the other hand, to our knowledge CG1648 and CG6426 have not been connected before to diapause, but Zang et al. [92] have suggested that expression of CG1648 is sensitive to the cold exposure in D. melanogaster. Such genes are interesting, not only as candidate genes for diapause processes per se, but also because they are candidates for investigating how genes in diapause may be isoform-specific. Transposable elements in diapause The possible role and/or the importance of the transposable elements (TEs) in diapause is currently unknown, but several studies have connected TEs to this trait. For example, one retrotransposon has been shown to be highly expressed in the early pupal diapause of S. crassipalpis [88]. Moreover, Yokum et al. [93] showed the presence of miniature subterminal inverted repeat-like elements (MSITE) within the promoter regions and introns of diapause regulated genes DAT-2 and DAT-3 (diapause-associated transcripts) in Leptinotarsa decemlineata. These authors suggested that this may indicate a possible role of TEs in the evolution and regulation of diapause on Colorado potato beetle [93]. In our study, we found 15 unique TEs to be significantly DE between D and ND D. montana females; the majority of these TEs showed upregulation and only three of them were downregulated in diapausing females (S3 Table). Among these there were TE classes such as polyproteins, retrotransposons and non-long terminal repeats but also well characterized TEs like Ulysses, Gypsy, and Penelope, known from D. virilis group species [94]. Evidently, the clarification of the overall role of TEs in diapause and/or in the function of specific diapause connected genes demands more detailed research. Conclusions We used RNA sequencing to study genes expression changes associated with reproductive diapause in the northern fly species D. montana. Our results demonstrated large gene expression differences between the diapause phenotype, capable of surviving the long winters, and the non-diapausing summer phenotype. This is in accordance with the view of diapause being an alternative dynamic adaptation and not just an arrest in the normal active summer development. We also found several functionally enriched gene clusters associated with metabolism, fatty acid biosynthesis and metal binding, including genes with connection to cuticular hydrocarbons (desat genes) or having acyl-CoA Δ11-desaturase activity (CG9747 and several odorant binding genes). In addition, we found several transposable elements (TE) to be differentially expressed between the two female types suggesting that TEs may play a role in, or be influenced by diapause. Finally, we showed that comparing transcriptional changes in the diapausing and non-diapausing females collected from the same environmental conditions produced comparable results, even though tissue heterogeneity has to be taken into account. RNAseq data has been submitted to NCBI's Gene Expression Omnibus. The transcriptome assembly has been deposited at DDBJ/EMBL/GenBank under the accession GECM00000000. The version described in this paper is the first version, GECM01000000. Raw reads have been deposited in GEO under accession codes: SRR2910695, SRR2910698, SRR2910701, SRR2910702, SRR2910703, and SRR2910704. Supporting Information S1 File Compressed zip file of Tables A-O including significant gene clusters from DAVID analysis. Table A. Cluster 1 DNA replication. Table B. Cluster 2 Actin binding. Table C. Cluster 3 Helicase activity. Table D. Cluster 4 Nucleotide binding. Table E. Cluster 5 Protein transporter activity. Table F. Cluster 6 Nuclear chromosome. Table G. Cluster 7 Oxidative stress. Table H. Cluster 8 Metal binding. Table I. Cluster 9 Maternal tudor protein. Table J. Cluster 10 Glycolysis. Table K. Cluster 11 DNA replication. Table L. Cluster 12 DNA polymerase. Table M. Cluster 13 Fatty acid biosynthesis. Table N. Cluster 14 DNA packing. Table O. Cluster 15 Metal ion binding. (ZIP) Click here for additional data file. S1 Table Differentially expressed contigs. (XLSX) Click here for additional data file. S2 Table Primers for quantitative PCR with the efficiency (E%) and R2 values. (DOCX) Click here for additional data file. S3 Table Transposable elements. (XLSX) Click here for additional data file. S4 Table Olfactory genes. (XLSX) Click here for additional data file. S5 Table Hsp genes. (XLSX) Click here for additional data file. Tapio Envall is acknowledged for conducting the qPCR runs, Kalevi Trontti for the preliminary analysis of the different splicing forms of the DE genes and Laura Vesala for helping in the fly maintenance and dissections. We sincerely thank four anonymous reviewers for helpful comments and professor William J. Etges for suggestions and language review that clearly improved the manuscript. ==== Refs References 1 Denlinger DL (2002 ) Regulation of diapause . Annu Rev Entomol 47 : 93 –122 . 11729070 2 Koštál V (2011 ) Insect photoperiodic calendar and circadian clock: Independence, cooperation, or unity? J Insect Physiol 57 : 538 –556 . 10.1016/j.jinsphys.2010.10.006 21029738 3 Koštál V (2006 ) Eco-physiological phases of insect diapause . J Insect Physiol 52 : 113 –127 . 16332347 4 Emerson KJ , Bradshaw WE , Holzapfel CM (2009 ) Complications of complexity: integrating environmental, genetic and hormonal control of insect diapause . Trends Genet 25 : 217 –225 . 10.1016/j.tig.2009.03.009 19375812 5 Saunders DS , In: Steel CGH , Vafopoulou X , Lewis RD , editors. Insect Clocks , 3rd ed. Elsevier Science , Amsterdam , 2002 . 6 Xiao H-J , Mou F-C , Zhu X-F , Xue F-S (2010 ) Diapause induction, maintenance and termination in the rice stem borer Chilo suppressalis (Walker) . J Insect Physiol 56 : 1558 –1564 . 10.1016/j.jinsphys.2010.05.012 20546744 7 Bradshaw WE (1976 ) Geography of photoperiodic response in a diapausing mosquito . Nature 262 : 384 –386 . 8 Lankinen P (1986 ) Geographical variation in circadian eclosion rhythm and photoperiodic adult diapause in Drosophila littoralis . J Comp Physiol A 159 : 123 –142 . 9 Tyukmaeva VI , Salminen TS , Kankare M , Knott KE , Hoikkala A (2011 ) Adaptation to a seasonally varying environment: a strong latitudinal cline in reproductive diapause combined with high gene flow in Drosophila montana . Ecol Evol 1 : 160 –168 . 10.1002/ece3.14 22393492 10 Menzel A , Fabian P (1999 ) Growing season extended in Europe . Nature 397 : 659 . 11 Bradshaw WE , Holzapfel CM (2001 ) Genetic shift in photoperiodic response correlated with global warming . Proc Natl Acad Sci USA 98 : 14509 –14511 11698659 12 Calvert WH , Brower LP (1986 ) The location of monarch butterfly (Danaus plexippus L.) overwintering colonies in Mexico in relation to topography and climate . J Lepid Soc 40 : 164 –187 . 13 Bowen MF (1992 ) Patterns of sugar feeding in diapausing and nondiapausing Culex pipiens (Diptera:Culicidae) females . J Med Entomol 29 : 843 –849 . 1404264 14 Arrese EL , Soulages JL (2013 ) Insect fat body: energy, metabolism and regulation . Annu Rev Entomol 55 : 207 –225 . 15 Zhou G , Miesfeld RL (2009 ) Energy metabolism during diapause in Culex pipiens mosquitoes . J Insect Physiol 55 : 40 –46 . 10.1016/j.jinsphys.2008.10.002 18992753 16 Burmester T (1999 ) Evolution and function of the insect hexamerins . Eur J Entomol 96 : 213 –225 . 17 Adams AJ , (1985 ) The critical field photoperiod inducing ovarian diapause in the cabbage whitefly, Aleyrodes proletella (Homoptera: Aleyrodidae) . Physiol Entomol 10 : 243 –249 . 18 Hahn DA , Denlinger DL (2007 ) Meeting the energetic demands of insect diapause: nutrient storage and utilization . J Insect Physiol 53 : 760 –773 . 17532002 19 Rinehart JP , Li A , Yocum GD , Robich RM , Hayward SAL , Denlinger DL (2007 ) Up-regulation of heat shock proteins is essential for cold survival during insect diapause . Proc Natl Acad Sci USA 104 : 11130 –11137 . 17522254 20 Ishiguro S , Li Y , Nakano K , Tsumuki H , Goto M (2007 ). Seasonal changes in glycerol content and cold hardiness in two ecotypes of the rice stem borer, Chilo suppressalis, exposed to the environment in the Shonai district, Japan . J Insect Physiol 53 : 392 –397 . 17336324 21 Guppy M , Withers P (1999 ) Metabolic depression in animals: physiological perspectives and biochemical generalizations . Biol Rev Camb Philos Soc 74 : 1 –40 . 10396183 22 Kim M , Denlinger DL (2009 ) Decrease in expression of beta-tubulin and microtubule abundance in flight muscles during diapause in adults of Culex pipiens . Insect Mol Biol 18 : 295 –302 . 10.1111/j.1365-2583.2009.00870.x 19523062 23 Hahn DA , Denlinger DL (2011 ) Energetics of Insect Diapause . Annu Rev Entomol 56 : 103 –121 . 10.1146/annurev-ento-112408-085436 20690828 24 Herman WS , Tatar M (2001 ) Juvenile hormone regulation of longevity in the migratory monarch butterfly . Proc Roy Soc B Biol Sci 268 : 2509 –2514 . 25 Tatar M , Chien SA , Priest NK (2001 ) Negligible senescence during reproductive dormancy in Drosophila melanogaster . Am Nat 158 : 248 –258 . 10.1086/321320 18707322 26 Nishizuka M , Azuma A , Masaki S (1998 ) Diapause response to photoperiod and temperature in Lepisma saccharina linnaeus (Thysanura: Lepismatidae) . Entomol Sci 1 : 7 –14 . 27 MacRae TH (2010 ) Gene expression, metabolic regulation and stress tolerance during diapause . Cell Mol Life Sci 67 : 2405 –2424 . 10.1007/s00018-010-0311-0 20213274 28 Bradshaw WE , Holzapfel CM (2001 ) Phenotypic evolution and the genetic architecture underlying photoperiodic time measurement . J Insect Physiol 47 : 809 –820 . 29 Saunders DS , Henrich VC , Gilbert LI (1989 ) Induction of diapause in Drosophila melanogaster: photoperiodic regulation and the impact of arrhythmic clock mutations on time measurement . Proc Natl Acad Sci USA 86 : 3748 –3752 . 2498875 30 Saunders DS , Gilbert LI (1990 ) Regulation of ovarian diapause in Drosophila melanogaster by photoperiod and moderately low temperature . J Insect Physiol 36 : 195 –200 31 Schmidt PS , Paaby AB , Heschel MS (2005 ) Genetic Variance for Diapause Expression and Associated Life Histories in Drosophila melanogaster . Evolution 59 : 2616 –2625 . 16526509 32 Baker DA , Russell S (2009 ) Gene expression during Drosophila melanogaster egg development before and after reproductive diapause . BMC Genomics 10 : 242 10.1186/1471-2164-10-242 19463195 33 Paaby AB , Schmidt PS (2009 ) Dissecting the genetics of longevity in Drosophila melanogaster . Fly 3 : 29 –38 . 19182541 34 Kubrak OI , Kučerová L , Theopold U , Nässel DR (2014 ) The sleeping beauty: how reproductive diapause affects hormone signaling, metabolism, immune response and somatic maintenance in Drosophila melanogaster . PlosOne 9 (11 ). 35 Saunders DS , Gilbert LI (1990 ) Regulation of ovarian diapause in Drosophila melanogaster by photoperiod and moderately low temperature . J Insect Physiol 36 : 195 –200 36 Schmidt PS , Paaby AB (2008 ) Reproductive diapause and life-history clines in North American populations of Drosophila melanogaster . Evolution 62 : 1204 –1215 10.1111/j.1558-5646.2008.00351.x 18298646 37 Carson HL , Stalker HD (1948 ) Reproductive diapause in Drosophila robusta . Proc Natl Acad Sci USA 34 : 124 –129 . 16578279 38 Morales-Hojas R , Reis M , Vieira CP , Vieira J (2011 ) Resolving the phylogenetic relationships and evolutionary history of the Drosophila virilis group using multilocus data . Mol Phylogenet Evol 60 : 249 –258 . 10.1016/j.ympev.2011.04.022 21571080 39 Tamura K ., Subramanian S , Kumar S (2004 ) Temporal Patterns of Fruit Fly (Drosophila) Evolution Revealed by Mutation Clocks . Mol Biol Evol 21 : 36 –44 . 12949132 40 Lumme J (1978 ) Phenology and photoperiodic diapause in northern populations of Drosophila In: Dingle H , editor. Evolution of Insect Migration and Diapause . New York : Springer-Verlag pp. 145 –170 . 41 Lankinen P , Tyukmaeva VI , Hoikkala A (2013 ) Northern Drosophila montana flies show variation both within and between cline populations in the critical day length evoking reproductive diapause . J Insect Physiol 59 : 745 –751 . 10.1016/j.jinsphys.2013.05.006 23702203 42 Vesala L , Hoikkala A (2011 ) Effects of photoperiodically induced reproductive diapause and cold hardening on the cold tolerance of Drosophila montana . J Insect Physiol 57 : 46 –51 . 10.1016/j.jinsphys.2010.09.007 20932841 43 Kauranen H , Menegazzi P , Costa R , Helfrich-Förster C , Kankainen A , Hoikkala A (2012 ) Flies in the north: Locomotor behavior and clock neuron organization of Drosophila montana . J Biol Rhythm 27 : 377 –387 . 44 Lakovaara S (1969 ) Malt as a culture medium for Drosophila species . Dros Info Serv 44 : 128 . 45 Salminen TS , Hoikkala A (2013 ) Effect of temperature on the duration of sensitive period and on the number of photoperiodic cycles required for the induction of reproductive diapause in Drosophila montana . J Insect Physiol 59 : 450 –457 . 10.1016/j.jinsphys.2013.02.005 23428942 46 Sasson A , Michael TP (2010 ) Filtering error from SOLiD output . Bioinformatics 26 : 849 –850 . 10.1093/bioinformatics/btq045 20207696 47 Parker DJ , Vesala L , Ritchie MG , Laiho A , Hoikkala A , Kankare M (2015 ) Do closely related species use the same genes to adapt to fluctuating environmental conditions? Heredity 1 –9 . 48 Parker DJ , Ritchie MG , Kankare M (2016 ). Preparing for winter: the transcriptomic response associated with different day lengths in Drosophila montana . G3 (Bethesda) 3;6 (5 ): 1373 –1381 .26976440 49 Conesa A , Götz S , Garcia-Gomez JM , Terol J , Talon M , Robles M (2005 ) Blast2GO: a universal tool for annotation, visualization and analysis in functional genomics research . Bioinformatics 21 : 3674 –3676 . 16081474 50 Anders S , Pyl PT , Huber W (2015 ) HTSeq—A Python framework to work with high-throughput sequencing data . Bioinformatics 15; 31 (2 ): 166 –9 . 10.1093/bioinformatics/btu638 25260700 51 Anders S , Huber W (2010 ) Differential expression analysis for sequence count data . Genome Biol 11 : R106 10.1186/gb-2010-11-10-r106 20979621 52 Benjamini Y , Hochberg Y (1995 ) Controlling the false discovery rate: a practical and powerful approach to multiple testing . J R Stat Soc Series B Stat Methodol 57 : 289 –300 . 53 Huang da W , Sherman BT , Lempicki RA (2009 ) Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources . Nat Protoc 4 : 44 –57 . 10.1038/nprot.2008.211 19131956 54 Neville M , Goodwin SF (2012 ) Genome-wide approaches to understanding behavior in Drosophila melanogaster . Brief Funct Genomics 11 : 395 –404 . 10.1093/bfgp/els031 22843979 55 Graveley BR, Brooks AN, Carlson JW, Cherbas L, Choi, J, Davis CA et al. (2010). Database: The D. melanogaster transcriptome: modENCODE RNA-Seq data. http://www.modencode.org/celniker/. 56 Vesala L , Salminen TS , Laiho A , Hoikkala A , Kankare M (2012 ) Cold tolerance and cold-induced modulation of gene expression in two Drosophila virilis group species with different distributions . Insect Mol Biol 21 : 107 –118 . 10.1111/j.1365-2583.2011.01119.x 22122733 57 Schmidt PS , Zhu CT , Das J , Batavia M , Yang L , Eanes MF (2008 ) An amino acid polymorphism in the couch potato gene forms the basis for climatic adaptation in Drosophila melanogaster . Proc Natl Acad Sci USA 105 : 16207 –16211 . 10.1073/pnas.0805485105 18852464 58 Kankare M , Salminen TS , Laiho A , Vesala L , Hoikkala A (2010 ). Changes in gene expression linked with adult reproductive diapause in a northern malt fly species: a candidate gene microarray study . BMC Ecol 10 , 3 10.1186/1472-6785-10-3 20122138 59 Salminen TS , Vesala L , Laiho A , Merisalo M , Hoikkala A , Kankare M (2015 ) Seasonal gene expression kinetics between diapause phases in Drosophila virilis group species and overwintering differences between diapausing and non-diapausing females . Sci Rep 5 : 11197 10.1038/srep11197 26063442 60 Livak KJ , Schmittgen TD (2001 ) Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method . Methods 25 : 402 –408 . 11846609 61 Lee RE . Principles of insect low temperature tolerance In: Insects at Low Temperature . Lee E , Denlinger FL , editors. Chapman and Hall , New York 1991 pp. 17 –46 . 62 Storey KB , Keefe D , Kourtz L , Storey JM (1991 ) Glucose-6-phosphate dehydrogenase in cold hardy insects: Kinetic properties, freezing stabilization, and control of hexose monophosphate shunt activity . Insect Biochem 21 : 157 –164 . 63 Legan SK , Rebrin I , Mockett RJ , Radyuk SN , Klichko VI , Sohal RS et al (2008 ) Overexpression of glucose-6-phosphate dehydrogenase extends the life span of Drosophila melanogaster . J Biol Chem 283 : 32492 –32499 . 10.1074/jbc.M805832200 18809674 64 Kirsch M , De Groot H (2001 ) NAD(P)H, a directly operating antioxidant? FASEB J . 15 : 1569 –1574 . 11427489 65 Vesala L , Salminen TS , Koštál V , Zahradníčková H , Hoikkala A (2012 ) Myo-inositol as a main metabolite in overwintering flies: seasonal metabolomic profiles and cold stress tolerance in a northern drosophilid fly . J Exp Biol 215 (16 ): 2891 –2817 .22837463 66 Watanabe M , Tanaka K (1999 ) Seasonal change of the thermal response in relation to myo-inositol metabolism in adults of Aulacophora nigripennis (Coleoptera Chrysomelidae) . J Insect Physiol 45 :167 –172 . 12770385 67 Watanabe M (2002 ) Cold tolerance and myo-inositol accumulation in overwintering adults of a lady beetle Harmonia axyridis (Coleoptera: Coccinellidae) . Eur J Entomol 99 : 5 –9 . 68 Tu X , Wang J , Hao K , Whitman DW , Fan Y , Cao G et al (2015 ) Transcriptomic and proteomic analysis of pre-diapause and non-diapause eggs of migratory locust, Locusta migratoria L. (Orthoptera: Acridoidea) . Sci Rep 5 : 11402 10.1038/srep11402 26091374 69 Li H-M , Buczkowsk G , Mittapalli O , Xie J , Wu J , Westerman R et al (2008 ) Transcriptomic profiles of Drosophila melanogaster third instar larval midgut and responses to oxidative stress . Insect Mol Biol 17 (4 ): 325 –339 . 10.1111/j.1365-2583.2008.00808.x 18651915 70 Rouault J-D , Marican C , Wicker-Thomas C , Jallon J-M (2004 ) Relations between cuticular hydrocarbon (HC) polymorphism, resistance against desiccation and breeding temperature; a model for HC evolution in D. melanogaster and D. simulans . Genetica 120 : 195 –212 . 15088658 71 Greenberg AJ , Moran JR , Coyne JA , Wu C-I (2003 ) Ecological adaptation during incipient speciation revealed by precise gene replacement . Science 302 : 1754 –1757 . 14657496 72 Reynolds JA , Hand SC (2009 ) Embryonic diapause highlighted by differential expression of mRNAs for ecdysteroidogenesis, transcription and lipid sparing in the cricket Allonemobius socius . J Exp Biol 212 (13 ): 2075 –2084 .19525434 73 Sim C , Denlinger DL (2009 ) Transcription profiling and regulation of fat metabolism genes in diapausing adults of the mosquito Culex pipiens . Physiol Genomics 39 (3 ): 202 –9 . 10.1152/physiolgenomics.00095.2009 19706691 74 Tillman JA , Steven J , Seybold SJ , Jurenka RA , Blomquist GJ (1999 ) Insect pheromones—an overview of biosynthesis and endocrine regulation . Insect Biochem Mol Biol 29 : 481 –514 . 10406089 75 Aspi J , Lumme J , Hoikkala A , Heikkinen E (1993 ) Reproductive ecology of the boreal riparian guild of Drosophila . Ecography 16 : 65 –72 . 76 Kučerová L , Kubrak OI , Bengtsson JM , Strnad H , Nylin S , Theopold U et al (2016 ) Slowed aging during reproductive dormancy is reflected in genome-wide transcriptome changes in Drosophila melanogaster . BMC Genomics 17 : 50 10.1186/s12864-016-2383-1 26758761 77 Köhler K , Brunner E , Guan XL , Boucke K , Greber UF ,Mohanty S et al, (2009 ) A combined proteomic and genetic analysis identifies a role for the lipid desaturase Desat1 in starvation-induced autophagy in Drosophila . Autophagy 5 :7 , 980 –990 .19587536 78 Brun A , Cuany A , Le Mouel T , Berge J , Amichot M (1996 ) Inducibility of the Drosophila melanogaster Cytochrome P450 gene, CYP6A2, by phenobarbital in insecticide susceptible or resistant strains . Insect Biochem Mol Biol 26 : 697 –703 . 8995791 79 Yang P ., Tanaka H , Kuwano E , Suzuki K (2008 ) A novel cytochrome P450 gene (CYP4G25) of the silkmoth Antheraea yamamai: Cloning and expression pattern in pharate first instar larvae in relation to diapause . J Insect Physiol 54 : 636 –643 . 10.1016/j.jinsphys.2008.01.001 18308331 80 Liu N , Scott JG (1995 ) Genetics of resistance to pyrethroidiInsecticides in the house fly, Musca domestica . Pestic Biochem Physiol 52 : 116 –124 . 81 Le Goff G , Boundy S , Daborn PJ , Yen JL , Sofer L , Lind R et al (2003 ) Microarray analysis of cytochrome P450 mediated insecticide resistance in Drosophila . Insect Biochem Molec 33 : 701 –708 . 82 Pletcher SD , Macdonald SJ , Marguerie R , Certa U , Stearns SC , Goldstein DB , Partridge L (2002 ) Genome-wide transcript profiles in aging and calorically restricted Drosophila melanogaster . Curr Biol 12 : 712 –723 . 12007414 83 Riddell CE , Garces JDL , Adams S , Barribeau SM , Twell D , Mallon BE (2015 ) Differential gene expression and alternative splicing in insect immune specificity . BMC Genomics 15 : 1031 . 84 Abdrakhamanova A , Wang QY , Khokhlova L , Nick P (2003 ) Is microtubule disassembly a trigger for cold acclimation? Plant Cell Physiol 44 : 676 –686 . 12881495 85 Pokorna J , Schwarzerova K , Zelenkova S , Petrasek J , Janotova I , Capkova V , Opatrny Z (2004 ) Sites of actin filament initiation and reorganization in cold treated tobacco cells . Plant Cell Environ 27 : 641 –653 . 86 Kim M , Robich RM , Rinehart JP , Denlinger DL (2006 ) Upregulation of two actin genes and redistribution of actin during diapause and cold stress in the northern house mosquito, Culex pipiens . J Insect Physiol 52 : 1226 –1233 . 17078965 87 Robich RM , Rinehart JP , Kitchen LJ , Denlinger DL (2007 ) Diapause-specific gene expression in the northern house mosquito, Culex pipiens L. identified by suppressive subtractive hybridization . J Insect Physiol 53 : 235 –245 . 17098250 88 Reis M , Valer FB , Vieira CP , Vieira J (2015 ) Drosophila americana diapausing females show features typical of young flies . PlosOne (10 ): 9 . 89 Rinehart JP , Yocum GD , Denlinger DL (2000 ) Developmental upregulation of inducible hsp70 transcripts, but not the cognate form, during pupal diapause in the flesh fly, Sarcophaga crassipalpis . Insect Biochem Mol Biol 30 : 515 –521 . 10802243 90 Yocum GD , Joplin KH , Denlinger DL (1998 ) Upregulation of a 23 kDa small heat shock protein transcript during pupal diapause in the flesh fly, Sarcophaga crassipalpis . Insect Biochem Mol Biol 28 : 677 –682 . 9755478 91 Guo Q , Hao YJ , Li Y , Zhang YJ , Ren S , Si FL , Chen B (2015 ) Gene cloning, characterization and expression and enzymatic activities related to trehalose metabolism during diapause of the onion maggot Delia antiqua (Diptera: Anthomyiidae) . Gene 565 (1 ): 106 –15 . 10.1016/j.gene.2015.03.070 25841989 92 Zhang J , Marshall KE , Westwood JT , Clark MS , Sinclair BJ (2011 ) Divergent transcriptomic responses to repeated and single cold exposures in Drosophila melanogaster . J Exp Biol 214 (23 ): 4021 –9 .22071194 93 Yocum GD , Toutges MJ , Roehrdanz R.L , Dihle PJ (2011 ) Insertion of miniature subterminal inverted repeat-like elements in diapause-regulated genes in the Colorado potato beetle, Leptinotarsa decemlineata (Coleoptera: Chrysomelidae) . Eur J Entomol 108 (2 ): 197 –203 . 94 Evgen’ev M , Zelentsova H , Shostak N , Kozitsina M , Barskyi V , Henner VB , et al (1997 ) Penelope, a new family of transposable elements and its possible role in hybrid dysgenesis in Drosophila virilis . Proc Natl Acad Sci USA 94 (1 ): 196 –201 . 8990185

PMC005xxxxxx/PMC5003387.txt

==== Front PLoS Negl Trop DisPLoS Negl Trop DisplosplosntdsPLoS Neglected Tropical Diseases1935-27271935-2735Public Library of Science San Francisco, CA USA 2757103510.1371/journal.pntd.0004792PNTD-D-16-00099Research ArticleBiology and Life SciencesOrganismsProtozoansParasitic ProtozoansTrypanosomaTrypanosoma CruziMedicine and Health SciencesEpidemiologyDisease VectorsTriatomaBiology and Life SciencesOrganismsAnimalsVertebratesAmniotesMammalsMedicine and Health SciencesPathology and Laboratory MedicinePathogenesisHost-Pathogen InteractionsBiology and Life SciencesBiogeographyPhylogeographyEcology and Environmental SciencesBiogeographyPhylogeographyEarth SciencesGeographyBiogeographyPhylogeographyBiology and Life SciencesEvolutionary BiologyPopulation GeneticsPhylogeographyBiology and Life SciencesGeneticsPopulation GeneticsPhylogeographyBiology and Life SciencesPopulation BiologyPopulation GeneticsPhylogeographyPeople and placesGeographical locationsSouth AmericaBrazilBiology and Life SciencesPopulation BiologyPopulation DynamicsGeographic DistributionPeople and placesGeographical locationsSouth AmericaOver Six Thousand Trypanosoma cruzi Strains Classified into Discrete Typing Units (DTUs): Attempt at an Inventory Inventory of Trypanosoma cruzi Discrete Typing UnitsBrenière Simone Frédérique 12*Waleckx Etienne 3Barnabé Christian 11 IRD-CIRAD, INTERTRYP (Interactions hôtes-vecteurs-parasites-environnement dans les maladies tropicales négligées dues aux Trypanosomatidés), IRD Center, Montpellier, France2 Pontificia Universidad Católica del Ecuador, Centro de Investigación para la Salud en América Latina (CISeAL), Quito, Ecuador3 Centro de Investigaciones Regionales “Hideyo Noguchi”, Universidad Autónoma de Yucatán, Mérida, Yucatán, MéxicoDebrabant Alain EditorUS Food and Drug Administration, UNITED STATESThe authors have declared that no completing interest exist. Conceived and designed the experiments: SFB. Performed the experiments: SFB CB. Analyzed the data: SFB CB. Wrote the paper: SFB EW CB. * E-mail: Frederique.Breniere@ird.fr29 8 2016 8 2016 10 8 e000479220 1 2016 31 5 2016 © 2016 Brenière et al2016Brenière et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Trypanosoma cruzi, the causative agent of Chagas disease, presents wide genetic diversity. Currently, six discrete typing units (DTUs), named TcI to TcVI, and a seventh one called TcBat are used for strain typing. Beyond the debate concerning this classification, this systematic review has attempted to provide an inventory by compiling the results of 137 articles that have used it. A total of 6,343 DTU identifications were analyzed according to the geographical and host origins. Ninety-one percent of the data available is linked to South America. This sample, although not free of potential bias, nevertheless provides today’s picture of T. cruzi genetic diversity that is closest to reality. DTUs were genotyped from 158 species, including 42 vector species. Remarkably, TcI predominated in the overall sample (around 60%), in both sylvatic and domestic cycles. This DTU known to present a high genetic diversity, is very widely distributed geographically, compatible with a long-term evolution. The marsupial is thought to be its most ancestral host and the Gran Chaco region the place of its putative origin. TcII was rarely sampled (9.6%), absent, or extremely rare in North and Central America, and more frequently identified in domestic cycles than in sylvatic cycles. It has a low genetic diversity and has probably found refuge in some mammal species. It is thought to originate in the south-Amazon area. TcIII and TcIV were also rarely sampled. They showed substantial genetic diversity and are thought to be composed of possible polyphyletic subgroups. Even if they are mostly associated with sylvatic transmission cycles, a total of 150 human infections with these DTUs have been reported. TcV and TcVI are clearly associated with domestic transmission cycles. Less than 10% of these DTUs were identified together in sylvatic hosts. They are thought to originate in the Gran Chaco region, where they are predominant and where putative parents exist (TcII and TcIII). Trends in host-DTU specificities exist, but generally it seems that the complexity of the cycles and the participation of numerous vectors and mammal hosts in a shared area, maintains DTU diversity. Author Summary Trypanosoma cruzi, the causative agent of Chagas disease, has been classified into six genetic groups (discrete typing units, DTUs) named TcI-TcVI and a seventh one called TcBat. Currently, several genetic molecular markers are used to classify the strains after their isolation in culture or directly from biological samples. The current inventory compiling the published works aiming to identify the DTUs of T. cruzi strains accumulated a total of 6,343 identifications. Although this inventory is not free of sampling bias, like all samples, it is the largest sampling to date and hence likely represents the closest picture of the current diversity of T. cruzi strains (i) circulating throughout the endemic area from the southern United States to Argentina and (ii) circulating in vectors as well as in wild and domestic mammals, and humans. Data analysis helps identify trends and provides a basis for further comparisons of new data, in a context where human factors (migration, vector control, urbanization, deforestation, agricultural expansion, resource exploitation) influence the epidemiological patterns of Chagas disease. The authors received no specific funding for this work. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Trypanosoma cruzi is a pathogenic microorganism, the causative agent of Chagas disease, characterized by high genetic and phenotypic intraspecific diversity. Population genetics suggests that clonality is an important mode of propagation of the natural populations of T. cruzi [1], although, likely sexual reproduction [2, 3] and recombination events occur to some extent and are important mechanisms that generate genetic diversity within the taxon, as discussed in a recent review [4]. The consensual nomenclature recognizes six discrete typing units (DTUs) named TcI to TcVI and a recently proposed seventh, Tcbat [5–7]. This classification is widely used as a reference in epidemiological studies. However, there is not consensus on the best method to identify the different DTUs. Similarly, the evolutionary relationships between the DTUs and therefore the evolutionary history of T. cruzi continue to be researched [8]. Several mechanisms of evolution have been recognized such as clonality, hybridization, and conventional and nonconventional genetic exchanges. In addition, several studies have demonstrated the extraordinary plasticity of the T. cruzi genome. The evolutive relationships among these DTUs has not been fully elucidated, but two of them (TcV and TcVI) clearly have a hybrid origin with TcII and TcIII as putative parents [9] according to the authors, TcIII and TcIV could also originate from a hybrid between TcI and TcII [10, 11] but some claim that is not the case [12, 13]. TcI and TcII remain two pure lines that are evolving separately from a common ancestor dating from approximately 1–3 million years ago [11, 13]. The main properties of the different DTUs have been reported previously [3, 5, 14, 15]. Briefly, (i) TcI has a wide distribution, from the southern United States to northern Argentina and Chile; this DTU is the most frequently sampled in sylvatic cycles, but it is also frequent in domestic cycles and it is the dominant DTU responsible for the transmission of Chagas disease in endemic countries located north of the Amazon basin; (ii) studies show that TcII, V and VI are more likely to be associated with domestic cycles and patients with chronic Chagas disease in the Southern Cone countries and Bolivia; (iii) TcIII and IV are mainly sampled in rainforest sylvatic cycles; (iv) Tcbat previously identified in bats, has recently been found in humans [7, 16–18]. It is well known that various DTUs can coexist in the same vector and in a single host [19–21]. The different DTUs present substantial genetic diversity. Various reports have shown that the parasite’s genetic diversity has a profound impact on its epidemiological, biological and medical characteristics [22]. Consequently, it is indispensable to characterize the genotypes that are circulating in space and in hosts. Moreover, the tracking of the different genotypes is of great interest in eco-epidemiology, providing a better understanding of epidemiological systems. After the biogeographic overview of T. cruzi DTUs by Miles and his colleagues [23], no other exhaustive review has been done, while very numerous new genotyping studies using new genetic markers and additional parasite strains have been conducted. Although we are conducting studies on the limits of DTUs classifications of T. cruzi strains and their actual existence as genetically separated units, it seemed important to take all existing data that refer to the current classification and to examine the geographic properties and host specificities of the different DTUs. Methods Data were obtained from a total of 137 articles (including our own published results) selected after searching PubMed (http://www.ncbi.nlm.nih.gov/pubmed) with “DTU”, “genetic characterization”, “lineage”, “genotype”, “isozyme”, “isoenzyme”, and “Trypanosoma cruzi” as key words. This research, as exhaustive as possible, was updated to April 27, 2016. Research has also been conducted by authors having worked on the genetic characterization of T. cruzi strains. For our published data, additional data, not present in the publications, was included in the current inventory because this information was available from our own records. For example, the names and data concerning the strain origins analyzed in Barnabé et al. [24] were added here. The publications included in the inventory used genetic markers that allowed DTU typing according to the consensual nomenclature in 6–7 DTUs [5, 6]. Moreover, in some cases correspondences between typing methods with different markers were used for the data interpretation [6, 25]. The data are shown in an Excel spreadsheet (S1 Table) where each line corresponds to a single determination from an isolate, a strain, a laboratory clone, mammal blood or tissue samples, and different vector digestive tract samples (“sample type” column in S1 Table). Several lines were recorded when different DTUs were detected in a strain and its laboratory clones. When more than 1 DTU was detected in one vector or mammal host (mixed infection), several lines corresponding to each DTU were recorded in the file. A total of 6,343 determinations were compiled. Each of them has a code corresponding to the strain/sample name reported in the publications, except for the records not identified with a name but only counted in publications, which we have labeled “anonymous”. In a few publications, undistinguished DTUs were reported for part of the identifications; consequently, additional categories were created for them: TcI/TcII (three cases), TcII/TcV/TcVI (26 cases), TcII/TcVI (two cases), TcIII/TcIV (31 cases), and TcV/VI (47 cases). These undistinguished DTUs accounted for 1.7% of the total inventory. The geographical origin was informed by the country name (no missing data), the upper continental subdivision of North, Central, and South America, the upper administrative divisions such as state, province, department or region, and the lower administrative divisions such as municipality, province, or community according to the information existing in the publications. The collection dates of the strain or biological samples were not always documented (52.5% of missing year data). Host origin was generally informed by the species (31 missing data), and columns were added indicating the order, genus and tribe for the triatomines. Also, the cycles to which the different hosts belonged were classified as “domestic” when the hosts were living and/or were captured in the intra- and peridomicile areas, and “sylvatic” when the hosts were captured in the field outside domestic areas. When the location of the capture site was missing, the wild mammals where classified in the sylvatic cycle except for the synanthropic species such as opossums and rodents for which the information was considered as unknown (uk). The information on the methods used for the characterization of the DTUs is also included in S1 Table. The first column indicates if the DTU was characterized at nuclear or mitochondrial level or both, the second one indicates the method(s) used, and the third one on the markers, the names of the genes, or the number of loci for MLMT (multilocus microsatellite typing) and MLEE (multilocus enzyme electrophoresis). Results General overview of available data The 6,343 samples of T. cruzi DTUs compiled in this review were identified in vectors and mammalian hosts from 19 different countries, covering an area from the southern United States to Argentina (S2 Table). No data is available from Belize in Central America, and Uruguay and Guyana in South America. The vast majority of data relate to South America (90.7%). The DTUs were identified in 86 genera (32 missing cases), 158 different species of which 42 are vectors belonging to 7 genera (Dipetalogaster, Eratyrus, Meccus, Mepraia, Panstrongylus, Rhodnius, and Triatoma). Approximately of the identifications in South America 49.3% were from vector species; however, in North and Central America most of the identifications were from vectors (69.3% and 65.8% respectively). The mammal species belong to nine orders of which the most represented is the Primate order (61.5%), because 59.4% of the identifications in mammals were made in samples from humans (n = 1902). One-third of the DTU identifications (31.0%) corresponded to parasites from hosts (vectors and mammals) captured in sylvatic ecotopes, 57.6% from intra- and peridomestic hosts, and the others were undetermined (n = 719, 11.3%) because in several studies the origin of the vectors was not specified. Overall distribution of the DTUs (TcI-TcVI and Tcbat) In 1.7% of the samples, the DTU (n = 109) was reported as a group of DTUs: (i) in one dog, 15 coati from Brazil, and ten triatomines from Argentina, TcII, TcV, and TcVI were not distinguished; (ii) TcII or TcVI was reported in two T. infestans from Paraguay; (iii) 47 infections with TcV or TcVI in dogs, humans, T. infestans from Chile and Bolivia and P. megistus in Brazil were reported; (iv) in 31 vectors and mammal hosts from Brazil and Mexico TcIII/TcIV were not discriminated; and (v) in three cases TcI and TcII were not discriminated in T. pallidipennis. In the 6234 other records, TcI was found in approximatively 60.0% of the overall identifications; TcII, TcV and TcVI were identified in around 10% each; and TcIII, TcIV and Tcbat were rarer with percentages ≤ 3.6%. Fig 1 presents the proportions of DTUs observed, excluding from the calculation the ambiguous DTU determinations over the entire endemic area, and in North, Central and South America (see below). 10.1371/journal.pntd.0004792.g001Fig 1 T. cruzi DTUs distribution (TcI-TcVI and Tcbat) over 6234 determinations in vector and mammalian hosts from 19 endemic countries in the overall endemic area: In North America (n = 459), Central America (n = 120) and South America (n = 5655). The ambiguous determinations of DTUs were deleted from the samples. Geographical distribution of the DTUs According to the current available records, the DTU distribution was different between North, Central, and South America (Fig 1). In Central America only two DTUs (TcI and TcIV) were identified while all DTUs were detected in South America. In North America the latest studies have identified TcII, TcV and TcIII in addition to TcI and TcIV, which remain the major strains, in Central America. In South America the DTU distribution was highly variable depending on the country, and the current trend is a predominance of TcI north of the Amazon and the presence of all DTUs south of the Amazon with abundance of TcV and TcVI (Fig 2). 10.1371/journal.pntd.0004792.g002Fig 2 T. cruzi DTU distribution per country in South America out of a total of 5655 identifications. The ambiguous determinations of DTUs were deleted from the samples. Tcbat is a recently proposed DTU that is genetically more closely related to TcI than to any other DTU. Therefore this DTU is probably underestimated because it is not recognized by the markers used in many publications, and consequently it may have been erroneously equated with TcI. This DTU was identified in 59 bats belonging to 12 different species in Brazil, Colombia, and Ecuador [16, 17, 26, 27], in one specimen of T. sordida from the State of Mato Grosso do Sul State in Brazil [28], and in a Colombian patient infected with a mixture of TcI and TcBat [18]. As mentioned above, TcI was the most frequently identified DTU in the overall sample, with a lower percentage in South America (58.2%) than in North America (79.5%) and Central America (93.3%). It was identified in all the countries included in the study. In South America, the low frequencies of TcI in Argentina (19.9% of 589 determinations) and Paraguay (2.8% of 181) contrasted with the proportions of this DTU in the other South American countries (at least > 47.0%) (Fig 2). TcII was much more rarely identified (9.6% of overall DTUs identified). It was not identified in Central America out of 120 identifications, and only 13 identifications were reported from North America out of 459 (2.8%). Eight of these 13 TcII were found in Mexico, four in T. dimidiata captured in domestic cycles in the state of Veracruz [29] and four in Meccus pallidipennis collected in Michoacan [30]. The five other identifications were from mice and rats captured in the immediate surroundings of the dwelling of the first described autochthonous case of T. cruzi transmission in Louisiana, near New Orleans [31, 32]. In South America, TcII presents a higher proportion, reaching 10.4% and was reported in Colombia, Surinam, Peru, Bolivia, Brazil, Argentina, Paraguay and Chile. TcIII and TcIV, which are thought to result from ancestral hybridization between TcI and TcII, reached 3.4% and 3.6% of the identifications, respectively. In North America, both of these DTUs were reported in Mexico in several publications [29, 30, 33, 34], but for the moment only TcIV has been identified in the US [24, 31, 35, 36]. In Central America, only TcIV has been identified in Guatemala in humans and vectors [37, 38]. In other Central American countries, neither TcIII nor TcIV has been reported. In South America, TcIII could be more cosmopolitan (Argentina, Bolivia, Brazil, Chile, Colombia, Paraguay, Peru and Venezuela) than TcIV, which has not yet been identified yet in Argentina, Chile and Paraguay. The last two DTUs, TcV and TcVI, were the recent hybrids, derived from hybridizations between TcII and TcIII. These DTUs showed the most differential geographical distribution. Indeed, TcV was identified in North America in exceptional cases in Mexico (Veracruz) in T. dimidiata as well as above-mentioned TcII [29]. TcV and TcVI have never been identified in US in 148 determinations, nor in Central America in 120 cases. In contrast, in South America, these DTUs together have frequently been identified in several countries, Argentina (76.9%), Bolivia (44.6%), Chile (28.6%) and Paraguay (55.2%)—but very rarely in others such as in Colombia (1.1%) [24, 39–41], Ecuador (3.3%) [42], and Brazil (1.5%) [24]. In Peru they were identified in 13.0% [24, 43, 44]. Moreover, when the two DTUs coexist, different proportions can be observed in the different countries. The most remarkable case was the identification of TcV and TcVI in Bolivia with 43.1% and 1.0% respectively, while in Argentina TcVI was more common (50.0%) and TcV less frequently detected (26.5%). Eco-epidemiology of the DTUs Domestic versus sylvatic cycles T. cruzi circulates in nature in different environments and two categories are usually distinguished: (i) the domestic cycles where T. cruzi evolves between domestic vectors, domestic and synanthropic mammals, and humans that are living in dwellings or/and around dwellings in the peridomestic areas; see Walter et al. for a comprehensive definition of peridomicile [45]; and (ii) the sylvatic cycles where T. cruzi evolves between wild mammals and vectors living outside domestic areas. The current results (Fig 3) show that all DTUs, including Tcbat and taking into account the two cases described in the domestic cycle [18, 28], participate in domestic and sylvatic cycles in some places. According to the current inventory, TcBat, TcI and TcIII are significantly more frequently identified in sylvatic cycles than in domestic cycles (X2 test, p < 10−4) and inversely for TcII, TcIV TcV and TcVI (X2, p < 10−4). Nevertheless, these tests are only indicative because they correspond to a very gross approach that ignores sampling bias, which obviously exists. 10.1371/journal.pntd.0004792.g003Fig 3 Differential T. cruzi DTUs distributions in wild and domestic cycles. The ambiguous determinations of DTUs were deleted from the samples. Vector infections To date, TcI is the major DTU identified in vectors (78.1%). TcI was also the only DTU identified in the genus Eratyrus (n = 6) or was very predominant (> 91%) in Meccus (n = 176), Panstrongylus (n = 715), and Rhodnius (n = 525) (Table 1). 10.1371/journal.pntd.0004792.t001Table 1 DTUs of T. cruzi currently detected in seven genera of T. cruzi vectors. Vector genus DTU of T. cruzi             TcBat TcI TcII TcIII TcIV TcV TcVI Total Dipetalogaster   1           1 Eratyrus   6           6 Meccus   161 4 2 9     176 Mepraia   113 52     29 26 220 Panstrongylus   689 5 20 1     715 Rhodnius   499 3 1 21   1 525 Triatoma 1 952 48 43 12 180 200 1436 Total 1 2421 112 66 43 209 227 3079 For these genera where TcI was found highly prevalent, it is useful to detail which are the other DTUs identified: (i) in the genus Meccus TcII, TcIII, and TcIV were identified in one report in the species M. pallidipennis collected in municipalities of the State of Michoacan de Ocampo in Mexico [30]. In this study, of 26 specimens of this species, TcI only reached 42.3%; TcIII or TcIV have also been detected in T. longipennis in Jalisco state [34]. (ii) In the genus Panstrongylus, besides TcI, the dominant DTU (96.7%), TcII was identified in Brazil, TcIII in sylvatic cycles in three countries (Brazil, Colombia and Venezuela) and in domestic cycles in Bolivia [46], and TcIV [47] and TcV or TcVI were identified once in Venezuela and Brazil respectively [13, 47]. The hybrid strains (TcV or TcVI) were identified in P. megistus collected in Minas Gerais (Brazil) in a domestic environment. These results suggest a remarkably high diversity of DTUs in this genus. (iii) In the genus Rhodnius, besides TcI, the other DTUs were very scarce. Among them, TcIV was the most common (4.0%). It was identified in three species: R. brethesi and R. robustus in the Brazilian Amazon and R. prolixus in Colombia [24, 39], Venezuela [47] and Guatemala [37]. TcII was identified in only three R. neglectus and R. pictipes bugs in the state of Para in Brazil [48, 49]. Finally, TcIII and TcVI were reported in a single individual each (R. brethesi and R. prolixus respectively) [39, 50]. In contrast, in the genera Mepraia and Triatoma although TcI remains a major strain (51.4% and 66.3% respectively), the other DTUs were found more frequently. In Mepraia, the identifications were made in two species (M. gajardoi and M. spinolai) captured in a sylvatic environment for which, in addition to TcI, remarkably high percentages of TcII (23.6%), TcV (13.2%) and TcVI (11.8%) were identified [51–53]. In the genus Triatoma, the data were available for 18 species (Table 2), but the results concerned principally T. infestans (1081 identifications, 73.8%). Lower numbers of DTU identifications were available for T. dimidiata (170), T. sordida (50), T. barberi (46), T. rubida (24) T. maculata (19), T. eratyrusiformis (14) and T. protracta and T. braziliensis (12). For the remaining species, there were fewer than ten identifications. 10.1371/journal.pntd.0004792.t002Table 2 DTUs of T. cruzi currently detected in the genus Triatoma. Species DTU of T. cruzi           Tcbat TcI TcII TcIII TcIV TcV TcVI Total Triatoma barberi   46           46 Triatoma brasiliensis   6 3         9 Triatoma carrioni   3           3 Triatoma dimidiata   143 4 5 9 9   170 Triatoma eratyrusiformis   6         1 7 Triatoma gerstaeckeri   7           7 Triatoma infestans   627 37 31 1 170 194 1060 Triatoma maculata   19           19 Triatoma matogrossensis     1         1 Triatoma nigromaculata   3           3 Triatoma nitida   1           1 Triatoma protracta   11     1     12 Triatoma pseudomaculata 4           4 Triatoma rubida   24           24 Triatoma rubrovaria       7       7 Triatoma sanguisuga   8     1     9 Triatoma sordida 1 40 3     1 5 50 Triatoma venosa   2           2 Total 1 950 48 43 12 180 200 1434 In this set of species, TcI was very dominant (>80%) except in T. infestans and T. braziliensis where it was less abundant (59.1% and 66.6%respectively). In T. infestans all DTUs were identified. TcI, TcV and TcVI dominated (93.5%), while TcII and TcIII accounted for about 6.4% together and TcIV was only detected once in an endemic valley in southern Peru. Although T. brasiliensis has an epidemiological importance in Brazil, few strains were identified in this vector, all from states located in the Northeast region in Brazil (N = 12) mostly from domestic cycles: six TcI, three TcII, and three TcIII or TcIV. In the other species where TcI predominated, other DTUs were identified. In T. dimidiata, TcI was the only DTU identified except in one study in which of 33 specimens of the Mexican state of Veracruz, TcI (nine cases) as well as TcII, TcIII, TcIV and TcV were identified, the latter accounting for 72.7% of the sample. In T. sordida, TcII was detected in Brazil [28, 54], and TcV and TcVI in Argentina [55], and most of the insects were captured in domestic cycle. For the following nine species with a sample size < 10 (T. carrioni, T. gerstaeckeri, T. matogrossensis, T. nigromaculata, T. nitida, T. pseudomaculata, T. rubrovaria, T. sanguisuga, and T. venosa), TcI was a major DTU (28/37, 75.7%), TcII was identified in one T. matogrossensis [28], TcIII was the only DTU identified in T. rubrovaria from one study in the State of Rio Grande do Sul in Brazil [7], and TcIV was identified in one T. sanguisuga in the US. The hybrid DTUs (TcV and TcVI) were not identified [24]. Mammal infections Domestic cycles. Considering domestic cycles and their mammalian hosts, most of the identifications made were from samples isolated from humans (1902 identifications, 84.0%). Those in dogs reached 220, from nine countries. Sixteen identifications were from cats from Argentina [56, 57], 91 others were from small rodents in six countries that readily nest in peridomestic structures (e.g., in sheds and piles of building materials), including 52 reports in Rattus rattus from Venezuela. There was also a single study of DTU identification in 24 goats in Chile [58]. In North America and Central America together (n = 112), there were 76 identifications in humans, accounting for 4.0% of the total number of human identifications, ten in dogs, and 26 in small rodents; mostly TcI was found (85.7%). The other DTUs were TcII in small rodents from the US [31] and TcIV from three human cases in Guatemala [37] and in the US two rodents [31] and six dogs [35]. In South America, all DTUs except Tcbat were identified in humans and dogs, and their distribution was rather similar, except for the hybrid DTUs TcV and TcVI (Fig 4). TcV was more abundant than TcVI in humans and conversely in dogs. In both cases, the DTUs TcI, and the hybrid DTUs TcV or TcVI were the most frequently identified, reaching at least 25% each. For goats in Chile, TcI and hybrid strains were found to be abundant, but TcII also reached 29.2% [58]. For small rodents, TcI predominated. However, TcII and TcIV were, together with TcI, recently identified in the peridomestic area of a house surrounded by forest in Louisiana [31]. Moreover, TcV was found in R. rattus caught in peridomiciles of dwellings in the region of Chiquitania in Bolivia where T. sordida is the main vector species [59]. 10.1371/journal.pntd.0004792.g004Fig 4 Comparative distribution of the T. cruzi DTUs in humans and dogs. Sylvatic cycles. In the wild environment, the identifications of T. cruzi strains isolated from mammals involved nine orders and 106 species for which very few characterizations were available for each one. Currently, TcI appears to be the most frequent DTU (58.3%) in the overall samples. For the orders, Artiodactyla [47, 60] and Pilosa [10, 24, 61, 62], which included no more than five identifications each, all were TcI. For the order Xernathra the two strains identified were TcIII or TcIV [63] (Table 3). For the other orders, several specific trends are detailed below. 10.1371/journal.pntd.0004792.t003Table 3 Inventory of DTUs of T. cruzi identified in 960 wild mammals belonging to nine orders. Mammal orders DTU of T. cruzi                 Tcbat TcI TcII TcII/TcV/TcVI TcIII TcIII/TcIV TcIV TcV TcVI Total Artiodactyla   3               3 Carnivora   46 1 15 4 8 36     110 Chiroptera 59 57 21   4   4     145 Cingulata   2 1   78   1 1   83 Didelphimorphia 262 1   7 3   3 2 278 Pilosa   5               5 Primate   43 10   1   10     64 Rodentia   91 37   10 1   24 20 183 Xernathra 2 2 Total 59 509 71 15 104 14 51 28 22 873 Examining the DTU distribution in the main orders and species (Table 3), it is worth noting that TcI reached 94.2% in the order Didelphimorphia, while TcIII reached a similar percentage in Cingulata (94.0%). The 278 identifications in Didelphimorphia were from 12 countries in North and South America, and from 20 species. In this order, only a few of the other DTUs were identified: TcII in Chile [58], TcIII in Brazil [7, 63, 64] and Paraguay [65], TcIII or TcIV in Mexico [34] and TcV and TcVI in Bolivia [24, 59] and Chile [58]. Species from the order Cingulata were sampled in six countries (Bolivia, Brazil, Colombia, Paraguay, United States, and Venezuela) where TcIII was the main DTU, and in the US out of three samples, two were TcIV and one TcI. Two other identifications in Paraguay were TcII and TcV. In the order Chiroptera, 145 identifications were reported in 23 species, all from South America. Tcbat and TcI were similarly identified (around 40% each), while TcII identified in Brazil, Colombia and Surinam was less frequent (14.5%), and TcIII and TcIV were very rare (2.7% each) [66, 67]. In the order Carnivora, TcI (37.1%) and TcIV (58.1%) were principally identified, but of the 36 TcIV samples, 35 were from Procyon lotor captured in the US, also infected with TcI (two cases) and one Nasua nasua from Brazil, both belonging to the same Procyonidae family. Four TcIII were identified in Conepatus chinga in Argentina. In the order Rodentia, the DTU distribution was quite different with high percentages of TcII (20.0%), TcIII (5.5%), TcV (13.1%) and TcVI (10.1%) in addition to TcI (49.7%). However, the majority of the sample was from Chile (61.7%) and Brazil (27.9%); TcV and TcVI were abundant only in Chile and not in Brazil where TcI was 78.4% and TcIII 17.6%. For the wild primates, a total of 51 identifications were made in 15 monkey species mostly sampled in Brazil (82.8%); TcI (67.2%), TcII (15.6%), TcIV (15.6%) and TcIII in one specimen were the four DTUs identified, TcI (Brazil, Colombia, Ecuador, France and Venezuela) and TcIV (Bolivia, Brazil, USA and Venezuela) were from different countries, while TcII was sampled only in Brazil. Discussion For many years, the characterization of T. cruzi strains was mostly conducted with specific goals in limited geographical areas and consequently with a limited number of strains. The current compilation, based on the consensus nomenclature of six DTUs, reached an accumulated number of 6,343 identifications. However, T. cruzi genotyping is associated with many biases and trapping methods, and several caveats must be considered, such as (i) unequal distributions of the research groups in the eco-epidemiology of T. cruzi in different countries, resulting in nonhomogeneous information; (ii) selection of some DTUs during the culture step; (iii) differential parasitemia levels in hosts, facilitating the isolation by hemoculture or xenodiagnosis, or facilitating the direct detection of some DTUs over others; (iv) markers’ differing ability to detect the different DTUs; (v) overrepresentation of humans in the overall sample; (vi) scarcity of mammals that are difficult to trap; (vi) difficulties discriminating closely related DTUs; and (vii) use of a nonstandardized set of reference strains. Despite of this nonexhaustive list of biases, the data reported herein constitute the most complete picture of the DTU distribution in the endemic area of Chagas disease. The purpose of this review is not to discuss the current nomenclature of T. cruzi in six DTUs. Indeed, there is an increasing number of new genetic analyses of T. cruzi strains, especially from sylvatic cycles, which show that it is increasingly difficult to obtain a relevant genetic structure that divides into six statistically supported clusters with the most in vogue genetic markers, microsatellites and nuclear sequence polymorphisms [68–70]. Moreover, at the mitochondrial level, we recently assessed that three robust clusters that we named mtTcI, mtTcII and mtTcIII actually exist [8]. The mtTcI cluster includes only strains belonging to the TcI DTU, the mtTcII includes only those belonging to the TcII DTU and mtTcIII includes strains belonging to several DTUs: TcIII and TcIV (ancient hybrids of TcI/TcII), TcV and TcVI (recent hybrids TcII/TcIII) and even TcI (a result of mitochondrial introgression for some strains labeled TcI with nuclear markers). These last few years, a number of studies aiming to characterize T. cruzi strains have used the nomenclature of six DTUs, so we proposed to examine the eco-epidemiological features of these DTUs and highlight new knowledge that may challenge the current paradigm. Geographical distribution and origin of the DTUs Based on the available typing data, the first outstanding result is the predominance of TcI strains. This DTU, genetically diversified, is found throughout the geographic distribution of T. cruzi and in all cycles where it is always dominant. There are probably no ecological systems (i.e. geographical areas where the parasite evolves between mammalian hosts and vectors specific species) where TcI is absent. However, it appears that TcI strains do not develop well in some mammal species such as those within the order Cingulata since this order is rarely infected with TcI (Table 3). The ecological systems are usually complex networks of relationships involving many species of mammals and vectors, and strain diversity may be maintained because of differential interactions between the parasite’s hosts and genotypes. TcI is an old DTU that has evolved since 3–16 MYA as previously proposed [71], and its very high genetic diversity is consistent with a long-term evolution. Moreover, recombination between TcI strains appears to be more frequent than previously thought [2, 3, 72]. The recombination events (i.e. sex) generally increase the variability of the organisms and thus increase their resilience, allowing new areas to be conquered and especially new hosts that have probably played a key role in the large dispersion and adaptation of TcI. Another question is the geographical origin of TcI. A North-South clustering was recognized, even if some incongruence remains to be explained [73–75]. In an analysis of TcI, the Gran Chaco region was proposed as an origin, while human TcI may have a North/Central American origin [75–77]. It should be noted that if the current trend is to propose sub groups within TcI, the presence of subunits, evolving separately, must be previously evidenced which is not yet the case. Also, it has been proposed that marsupial species of the family Didelphidae family are the ancestral hosts of TcI [78] given that, among others, TcI predominates in these animals. Based on our recent analysis of COII and CytB gene sequences previously deposited in GenBank [8], we evaluated the haplotype and nucleotide diversities of TcI within the order Didelphimorphia, and we observed that these indices were comparable to those obtained for all the other orders of wild mammals combined. This assesses the larger genetic diversity in marsupials than in other animals, supporting a longer association. The remarkable expansion of TcI, which invaded most of environments, does not allow its origin to be determined from the picture of its geographical distribution alone. TcII is a DTU as old as TcI, but it has been sampled much more rarely. The strains belonging to this DTU carry mitochondrial genes (mtTcII mitochondrial cluster) whose sequences show substantial genetic divergence from TcI. Moreover, this DTU presents a much lower genetic diversity than TcI. For example, the haplotype diversity of COII and CytB genes are 0.39 and 0.48, while for mtTcI they are 0.81 and 0.58 respectively [8]. A similar level of differences is also observed for nucleotide diversity. The available data on the geographical distribution of TcII suggest that it is absent or extremely rare in some ecosystems (Central and North America). It seems that TcII strains would not have had the same expansion capacity as TcI among the wild cycles, and they probably found refuge mostly in certain wild mammals. TcII is already reported in different wild mammals of the Chiroptera, Cingulate, Didelphimorphia and Primate orders. However, its strong association with primates in the Atlantic Coastal Rainforest in Brazil should be noted [79]. In humans, it is relatively abundant, accounting for 20% of human strains, but it is highly abundant in Brazil (66% of human strains identifications) and rare in most other countries except Colombia (15%) and Chile (30%). For now, its geographical distribution is more consistent with a South American origin, and further south than north of the Amazon basin where this DTU is more abundant. TcIII and TcIV are DTUs that do not seem to be present throughout the entire endemic area. First, it is important to note that the genetic data do not clearly define these two groups separately. The genetic diversity of TcIII-TcIV is very large and the monophyly of each DTU is not really highlighted. Several studies showed that these strains are the result of ancient hybridization(s) between TcI and TcII strains, which suffer over time from genetic rearrangements, decreasing their level of heterozygosity at the expense of mosaic mitochondrial and nuclear genes [80]. Recombination events have probably occurred several times and this would have given a mtTcIII group composed of polyphyletic subgroups of strains. Therefore, the wild strains from the US, attributed to TcIV, seem to be a monophyletic subgroup differing from the others long ago [81], but whose closest ancestors have probably disappeared. There is little doubt that TcIII and TcIV DTUs have a sylvan origin, but these strains infect humans more than occasionally: the current database shows that TcIV is reported in 84 human cases in six countries (Brazil, Colombia, Ecuador, Guatemala, Peru and Venezuela), and 11 canine cases. Similarly, TcIII is reported in 26 human cases in Brazil and Paraguay. The two TcV and TcVI DTUs include strains derived from the hybridization of TcII and TcIII strains [9]. They are usually considered hybrids and they are heterozygous at several loci and SNPs (single nucleotide polymorphisms). In our database, a total of 21.3% of the determinations belong to these DTUs. Some of these strains have spread across large geographic areas through the clonal propagation mode [82]. Both DTUs are clearly associated with domestic cycles since only 10.5% of them are identified in hosts from wild cycles. They are identified in some Didelphimorphia and in different species of rodents but only in the Southern Cone countries and Bolivia. Previously, the Gran Chaco region was proposed as the original location of these DTUs, where they are very abundant and where the putative parents are also present [15], and this hypothesis fits well with the current observed distribution of these DTUs. Host specificity The universe of Hemiptera vectors of T. cruzi or potential vectors is huge since currently over 141–147 triatomine species are recognized, about 130 occur in the Americas, and it appears that all of these are able to transmit the parasite. Most of these species are involved in wild cycles with at least 100 species of mammals playing a role of host and/or reservoir. In the current data, only 37 species of vectors are included and for the majority of them, very few DTU determinations were made, even though these vectors are generally widely distributed. Similarly, the knowledge of the parasite genetic variants that infect mammals, except for humans, and to a lesser extent for Didelphidae, is very limited. In various regions, in a context of high anthropization and climate changes, it is urgent to study the impact of these environmental modifications on potential vectors and their hosts. Several studies of experimental infections of vectors with different strains of T. cruzi showed differences in susceptibility [83] and even suggested that the strains are pathogenic and induce more or less deleterious effects in bugs [84]. Few studies relate comparisons of DTUs in experimental infections in a single triatomine species. For T. infestans in which this was done, significant developmental differences in the vector were observed depending on the DTU it was infected with, and after experimental double infections: in 50% of cases, only one of the two DTUS was detected after a few days of infection [85, 86]. As a field observation, we can report the case of Triatoma sordida, a primary vector in the northeast of the city of Santa Cruz, in Bolivia, in which TcI was predominantly detected while in mammals of the same area, TcV was a major strain [59]. In wild mammal hosts, experimental infections of two important reservoirs in the US (placental and marsupial) showed DTU-mammal association [87]. Examples could be multiplied but we can already conclude that the vectors and even the wild mammal hosts can influence the distribution of DTUs. Whatever the host, there is a balance between parasite genotypes and hosts which probably depends on environmental conditions such as outside temperature for vectors or immune and nutritional status for mammals. The diversity of hosts, and environmental conditions certainly explain the maintenance of parasitic diversity and the emergence of new variants by natural selection. Therefore the distribution of DTUs reported here, although very informative, is only a temporary picture that will inevitably evolve over time, above all if drastic environmental changes occur such as deforestation, intensive farming, urbanization, and unexpected climatic upheavals. Supporting Information S1 Table Compiled list of Trypanosoma cruzi discrete typing units (DTU) previously reported in the literature. Each line corresponds to the identification of a single strain for which the geographic, host, and time origins are specified. The publication source is also presented. (XLSX) Click here for additional data file. S2 Table Summary of the collected determinations of T. cruzi DTUs per region, country, hosts, and ecotope originating from the compilations of the data (S1 Table). (XLSX) Click here for additional data file. ==== Refs References 1 Tibayrenc M , Kjellberg F , Ayala FJ . A clonal theory of parasitic protozoa: the population structures of Entamoeba, Giardia, Leishmania, Naegleria, Plasmodium, Trichomonas, and Trypanosoma and their medical and taxonomical consequences . Proceedings of the National Academy of Sciences USA . 1990 ;87 (7 ):2414 –8 . 2 Ocana-Mayorga S , Llewellyn MS , Costales JA , Miles MA , Grijalva MJ . Sex, subdivision, and domestic dispersal of Trypanosoma cruzi lineage I in southern Ecuador . PLoS Negl Trop Dis . 2010 ;4 (12 ):e915 10.1371/journal.pntd.0000915 21179502 3 Barnabé C , Buitrago R , Brémond P , Aliaga C , Salas R , Vidaurre P , et al Putative panmixia in restricted populations of Trypanosoma cruzi isolated from wild Triatoma infestans in Bolivia . PLoS One . 2013 ;8 (11 ):e82269 10.1371/journal.pone.0082269 24312410 4 Messenger LA , Miles MA . Evidence and importance of genetic exchange among field populations of Trypanosoma cruzi . Acta Trop . 2015 ;151 :150 –5 . 10.1016/j.actatropica.2015.05.007 26188331 5 Zingales B , Miles MA , Campbell DA , Tibayrenc M , Macedo AM , Teixeira MM , et al The revised Trypanosoma cruzi subspecific nomenclature: Rationale, epidemiological relevance and research applications . Infect Genet Evol . 2012 ;12 (2 ):240 –53 . 10.1016/j.meegid.2011.12.009 22226704 6 Zingales B , Andrade SG , Briones MR , Campbell DA , Chiari E , Fernandes O , et al A new consensus for Trypanosoma cruzi intraspecific nomenclature: second revision meeting recommends TcI to TcVI . Mem Inst Oswaldo Cruz . 2009 ;104 (7 ):1051 –4 . 20027478 7 Marcili A , Lima L , Cavazzana MJ , Junqueira ACV , Veludo HH , Da Silva FM , et al A new genotype of Trypanosoma cruzi associated with bats evidenced by phylogenetic analyses using SSU rDNA, cytochrome b and Histone H2B genes and genotyping based on ITS1 rDNA . Parasitology . 2009 ;136 (6 ):641 –55 . 10.1017/S0031182009005861 19368741 8 Barnabé C , Mobarec HI , Jurado MR , Cortez JA , Brenière SF . Reconsideration of the seven discrete typing units within the species, a new proposal of three reliable mitochondrial clades . Infect Genet Evol . 2016 ;39 :176 –86 . 10.1016/j.meegid.2016.01.029 26845347 9 Westenberger SJ , Barnabé C , Campbell DA , Sturm NR . Two hybridization events define the population structure of Trypanosoma cruzi . Genetics . 2005 ;171 (2 ):527 –43 . 15998728 10 Ramirez JD , Duque MC , Guhl F . Phylogenetic reconstruction based on Cytochrome b (Cytb) gene sequences reveals distinct genotypes within Colombian I populations . Acta Trop . 2011 ;119 (1 ):61 –5 . 10.1016/j.actatropica.2011.04.009 21524641 11 Sturm NR , Vargas NS , Westenberger SJ , Zingales B , Campbell DA . Evidence for multiple hybrid groups in Trypanosoma cruzi . Int J Parasitol . 2003 ;33 (3 ):269 –79 . 12670512 12 Tomasini N , Diosque P . Evolution of: clarifying hybridisations, mitochondrial introgressions and phylogenetic relationships between major lineages . Mem Inst Oswaldo Cruz . 2015 ;110 (3 ):403 –13 . 10.1590/0074-02760140401 25807469 13 de Freitas JM , Augusto-Pinto L , Pimenta JR , Bastos-Rodrigues L , Goncalves VF , Teixeira SM , et al Ancestral Genomes, Sex, and the Population Structure of . PLoS Pathog . 2006 ;2 (3 ):e24 16609729 14 Brenière SF , Aliaga C , Waleckx E , Buitrago R , Salas R , Barnabé C , et al Genetic characterization of DTUs in wild Triatoma infestans from Bolivia: predominance of TcI . PLoS Negl Trop Dis . 2012 ;6 (5 ):e1650 10.1371/journal.pntd.0001650 22685616 15 Perez E , Monje M , Chang B , Buitrago R , Parrado R , Barnabé C , et al Predominance of hybrid discrete typing units of in domestic Triatoma infestans from the Bolivian Gran Chaco region . Infect Genet Evol . 2013 ;13 :116 –23 . 10.1016/j.meegid.2012.09.014 23047136 16 Cavazzana M Jr., Marcili A , Lima L , da Silva FM , Junqueira AC , Veludo HH , et al Phylogeographical, ecological and biological patterns shown by nuclear (ssrRNA and gGAPDH) and mitochondrial (Cyt b) genes of trypanosomes of the subgenus Schizotrypanum parasitic in Brazilian bats . Int J Parasitol . 2010 ;40 (3 ):345 –55 . 10.1016/j.ijpara.2009.08.015 19766649 17 Ramirez JD , Tapia-Calle G , Munoz-Cruz G , Poveda C , Rendon LM , Hincapie E , et al Trypanosome species in neo-tropical bats: biological, evolutionary and epidemiological implications . Infect Genet Evol . 2014 ;22 :250 –6 . 10.1016/j.meegid.2013.06.022 23831017 18 Ramirez JD , Hernandez C , Montilla M , Zambrano P , Florez AC , Parra E , et al First report of human infection attributed to TcBat genotype . Zoonoses Public Health . 2014 ;61 (7 ):477 –9 . 10.1111/zph.12094 25285940 19 Brenière SF , Bosseno MF , Telleria J , Bastrenta B , Yacsik N , Noireau F , et al Different behavior of two major clones: transmission and circulation in young Bolivian patients . Exp Parasitol . 1998 ;89 (3 ):285 –95 . 9676706 20 Devillers H , Lobry JR , Menu F . An agent-based model for predicting the prevalence of I and II in their host and vector populations . J Theor Biol . 2008 ;255 (3 ):307 –15 . 10.1016/j.jtbi.2008.08.023 18805428 21 Bosseno MFY , Vargas N. , Brenière F. , S. F. Selection of clonal genotypes (clonet 20 and 39) isolated from Bolivian triatomines following subculture in liquid medium . Memorias Do Instituto Oswaldo Cruz . 2000 ;95 (5 ):601 –7 . 10998206 22 Tibayrenc M , Barnabé C , Telleria J . Reticulate Evolution in: Medical and Epidemiological Implications In: Telleria J , Tibayrenc M , editors. American trypanosomiasis: Chagas disease One hundred years of research . Burlington : Elsevier ; 2010 p. 475 –88 . 23 Miles MA , Llewellyn MS , Lewis MD , Yeo M , Baleela R , Fitzpatrick S , et al The molecular epidemiology and phylogeography of and parallel research on Leishmania: looking back and to the future . Parasitology . 2009 ;136 (12 ):1509 –28 . 10.1017/S0031182009990977 19691868 24 Barnabé C , Brisse S , Tibayrenc M . Population structure and genetic typing of, the agent of Chagas disease: a multilocus enzyme electrophoresis approach . Parasitology . 2000 ;120 :513 –26 . 10840981 25 Brisse S , Verhoef J , Tibayrenc M . Characterisation of large and small subunit rRNA and mini-exon genes further supports the distinction of six Trypanosoma cruzi lineages . Int J Parasitol . 2001 ;31 (11 ):1218 –26 . 11513891 26 Lima L , Espinosa-Alvarez O , Pinto CM , Cavazzana M Jr., Pavan AC , Carranza JC , et al New insights into the evolution of the Trypanosoma cruzi clade provided by a new trypanosome species tightly linked to Neotropical Pteronotus bats and related to an Australian lineage of trypanosomes . Parasit Vectors . 2015 ;8 :657 10.1186/s13071-015-1255-x 26701154 27 Pinto CM , Ocana-Mayorga S , Tapia EE , Lobos SE , Zurita AP , Aguirre-Villacis F , et al Bats, Trypanosomes, and Triatomines in Ecuador: New Insights into the Diversity, Transmission, and Origins of Trypanosoma cruzi and Chagas Disease . PLoS One . 2015 ;10 (10 ):e0139999 10.1371/journal.pone.0139999 26465748 28 Cominetti MC , Csordas BG , Cunha RC , Andreotti R . Geographical distribution of Trypanosoma cruzi in triatomine vectors in the State of Mato Grosso do Sul, Brazil . Rev Soc Bras Med Trop . 2014 ;47 (6 ):747 –55 . 10.1590/0037-8682-0234-2014 25626654 29 Ramos-Ligonio A , Torres-Montero J , Lopez-Monteon A , Dumonteil E . Extensive diversity of Trypanosoma cruzi discrete typing units circulating in Triatoma dimidiata from central Veracruz, Mexico . Infect Genet Evol . 2012 ;12 (7 ):1341 –3 . 10.1016/j.meegid.2012.04.024 22569098 30 Ibanez-Cervantes G , Martinez-Ibarra A , Nogueda-Torres B , Lopez-Orduna E , Alonso AL , Perea C , et al Identification by Q-PCR of Trypanosoma cruzi lineage and determination of blood meal sources in triatomine gut samples in Mexico . Parasitol Int . 2013 ;62 (1 ):36 –43 . 10.1016/j.parint.2012.09.003 22995149 31 Herrera CP , Licon MH , Nation CS , Jameson SB , Wesson DM . Genotype diversity of Trypanosoma cruzi in small rodents and Triatoma sanguisuga from a rural area in New Orleans, Louisiana . Parasit Vectors . 2015 ;8 :123 10.1186/s13071-015-0730-8 25890064 32 Dorn PL , Perniciaro L , Yabsley MJ , Roellig DM , Balsamo G , Diaz J , et al Autochthonous transmission of Trypanosoma cruzi, Louisiana . Emerg Infect Dis . 2007 ;13 (4 ):605 –7 . 17553277 33 Bosseno MF , Barnabé C , Magallon Gastelum E , Lozano Kasten F , Ramsey J , Espinoza B , et al Predominance of Trypanosoma cruzi Lineage I in Mexico . J Clin Microbiol . 2002 ;40 (2 ):627 –32 . 11825982 34 Bosseno MF , Barnabé C , Sierra MJ , Kengne P , Guerrero S , Lozano F , et al Wild ecotopes and food habits of Triatoma longipennis infected by Trypanosoma cruzi lineages I and II in Mexico . Am J Trop Med Hyg . 2009 ;80 (6 ):988 –91 . 19478263 35 Roellig DM , Savage MY , Fujita AW , Barnabé C , Tibayrenc M , Steurer FJ , et al Genetic variation and exchange in Trypanosoma cruzi isolates from the United States . PLoS One . 2013 ;8 (2 ):e56198 10.1371/journal.pone.0056198 23457528 36 Shender LA , Lewis MD , Rejmanek D , Mazet JA . Molecular Diversity of Trypanosoma cruzi Detected in the Vector Triatoma protracta from California, USA . PLoS Negl Trop Dis . 2016 ;10 (1 ):e0004291 10.1371/journal.pntd.0004291 26797311 37 Iwagami M , Higo H , Miura S , Yanagi T , Tada I , Kano S , et al Molecular phylogeny of Trypanosoma cruzi from Central America (Guatemala) and a comparison with South American strains . Parasitol Res . 2007 ;102 (1 ):129 –34 . 17828552 38 Higo H , Miura S , Horio M , Mimori T , Hamano S , Agatsuma T , et al Genotypic variation among lineages of Trypanosoma cruzi and its geographic aspects . Parasitol Int . 2004 ;53 (4 ):337 –44 . 15464443 39 Brisse S , Henriksson J , Barnabé C , Douzery EJ , Berkvens D , Serrano M , et al Evidence for genetic exchange and hybridization in Trypanosoma cruzi based on nucleotide sequences and molecular karyotype . Infect Genet Evol . 2003 ;2 (3 ):173 –83 . 12797979 40 Ramirez JD , Turriago B , Tapia-Calle G , Guhl F . Understanding the role of dogs (Canis lupus familiaris) in the transmission dynamics of Trypanosoma cruzi genotypes in Colombia . Vet Parasitol . 2013 ;196 (1–2 ):216 –9 . 10.1016/j.vetpar.2012.12.054 23351975 41 Sanchez LV , Bautista DC , Corredor AF , Herrera VM , Martinez LX , Villar JC , et al Temporal variation of Trypanosoma cruzi discrete typing units in asymptomatic Chagas disease patients . Microbes Infect . 2013 ;15 (10–11 ):745 –8 . 10.1016/j.micinf.2013.06.008 23811021 42 Garzon EA , Barnabé C , Cordova X , Bowen C , Paredes W , Gomez E , et al Trypanosoma cruzi isoenzyme variability in Ecuador: first observation of zymodeme III genotypes in chronic chagasic patients . Trans R Soc Trop Med Hyg . 2002 ;96 (4 ):378 –82 . 12497973 43 Brenière SF , Braquemond P , Solari A , Agnèse JF , Tibayrenc M . An isoenzyme study of naturally occurring clones of Trypanosoma cruzi isolated from both sides of the West Andes highland . Transactions of the Royal Society of Tropical Medicine and Hygiene . 1991 ;85 (1 ):62 –6 . 2068764 44 Guhl F , Auderheide A , Ramirez JD . From ancient to contemporary molecular eco-epidemiology of Chagas disease in the Americas . Int J Parasitol . 2014 ;44 (9 ):605 –12 . 10.1016/j.ijpara.2014.02.005 24675555 45 Walter A , Lozano-Kasten F , Bosseno MF , Ruvalcaba EG , Gutierrez MS , Luna CE , et al Peridomicilary habitat and risk factors for Triatoma infestation in a rural community of the Mexican occident . Am J Trop Med Hyg . 2007 ;76 (3 ):508 –15 . 17360876 46 Rojas-Cortez M , Pinazo MJ , Garcia L , Arteaga M , Uriona L , Gamboa S , et al Trypanosoma cruzi-infected Panstrongylus geniculatus and Rhodnius robustus adults invade households in the Tropics of Cochabamba region of Bolivia . Parasit Vectors . 2016 ;9 (1 ):158 .26984679 47 Carrasco HJ , Segovia M , Llewellyn MS , Morocoima A , Urdaneta-Morales S , Martinez C , et al Geographical distribution of Trypanosoma cruzi genotypes in Venezuela . PLoS Negl Trop Dis . 2012 ;6 (6 ):e1707 10.1371/journal.pntd.0001707 22745843 48 Ferreira RC , Briones MR . Phylogenetic evidence based on Trypanosoma cruzi nuclear gene sequences and information entropy suggest that inter-strain intragenic recombination is a basic mechanism underlying the allele diversity of hybrid strains . Infect Genet Evol . 2012 ;12 (5 ):1064 –71 . 10.1016/j.meegid.2012.03.010 22449773 49 Lima Vdos S , Xavier SC , Maldonado IF , Roque AL , Vicente AC , Jansen AM . Expanding the knowledge of the geographic distribution of Trypanosoma cruzi TcII and TcV/TcVI genotypes in the Brazilian Amazon . PLoS One . 2014 ;9 (12 ):e116137 10.1371/journal.pone.0116137 25551227 50 Valadares HM , Pimenta JR , de Freitas JM , Duffy T , Bartholomeu DC , Oliveira Rde P , et al Genetic profiling of Trypanosoma cruzi directly in infected tissues using nested PCR of polymorphic microsatellites . Int J Parasitol . 2008 ;38 (7 ):839 –50 . 18154957 51 Arenas M , Campos R , Coronado X , Ortiz S , Solari A . Trypanosoma cruzi genotypes of insect vectors and patients with Chagas of Chile studied by means of cytochrome b gene sequencing, minicircle hybridization, and nuclear gene polymorphisms . Vector Borne Zoonotic Dis . 2012 ;12 (3 ):196 –205 . 10.1089/vbz.2011.0683 22022808 52 Coronado X , Rozas M , Botto-Mahan C , Ortiz S , Cattan PE , Solari A . Molecular epidemiology of Chagas disease in the wild transmission cycle: the evaluation in the sylvatic vector Mepraia spinolai from an endemic area of Chile . Am J Trop Med Hyg . 2009 ;81 (4 ):656 –9 . 10.4269/ajtmh.2009.09-0053 19815882 53 Toledo A , Vergara F , Campos R , Botto-Mahan C , Ortiz S , Coronado X , et al Trypanosoma cruzi genotypes in Mepraia gajardoi from wild ecotopes in northern Chile . Am J Trop Med Hyg . 2013 ;88 (2 ):285 –8 . 10.4269/ajtmh.2012.12-0227 23249691 54 Abolis NG , Araujo SM , Toledo MJ , Fernandez MA , Gomes ML . Trypanosoma cruzi I-III in southern Brazil causing individual and mixed infections in humans, sylvatic reservoirs and triatomines . Acta Trop . 2011 ;120 (3 ):167 –72 . 10.1016/j.actatropica.2011.08.001 21855523 55 Maffey L , Cardinal MV , Ordonez-Krasnowski PC , Lanati LA , Lauricella MA , Schijman AG , et al Direct molecular identification of Trypanosoma cruzi discrete typing units in domestic and peridomestic Triatoma infestans and Triatoma sordida from the Argentine Chaco . Parasitology . 2012 ;139 (12 ):1570 –9 . 10.1017/S0031182012000856 23036510 56 Cardinal MV , Lauricella MA , Ceballos LA , Lanati L , Marcet PL , Levin MJ , et al Molecular epidemiology of domestic and sylvatic Trypanosoma cruzi infection in rural northwestern Argentina . Int J Parasitol . 2008 . 57 Enriquez GF , Cardinal MV , Orozco MM , Lanati L , Schijman AG , Gurtler RE . Discrete typing units of Trypanosoma cruzi identified in rural dogs and cats in the humid Argentinean Chaco . Parasitology . 2013 ;140 (3 ):303 –8 . 10.1017/S003118201200159X 23058180 58 Rozas M , Botto-Mahan C , Coronado X , Ortiz S , Cattan PE , Solari A . Coexistence of Trypanosoma cruzi genotypes in wild and periodomestic mammals in Chile . Am J Trop Med Hyg . 2007 ;77 (4 ):647 –53 . 17978065 59 Brenière SF , Morochi W , Bosseno MF , Ordonez J , Gutierrez T , Vargas F , et al Trypanosoma cruzi genotypes associated with domestic Triatoma sordida in Bolivia . Acta Trop . 1998 ;71 (3 ):269 –83 . 9879736 60 Herrera HM , Rocha FL , Lisboa CV , Rademaker V , Mourao GM , Jansen AM . Food web connections and the transmission cycles of Trypanosoma cruzi and Trypanosoma evansi (Kinetoplastida, Trypanosomatidae) in the Pantanal Region, Brazil . Trans R Soc Trop Med Hyg . 2011 ;105 (7 ):380 –7 . 10.1016/j.trstmh.2011.04.008 21600622 61 De Araujo VA , Boite MC , Cupolillo E , Jansen AM , Roque AL . Mixed infection in the anteater Tamandua tetradactyla (Mammalia: Pilosa) from Para State, Brazil: Trypanosoma cruzi, T. rangeli and Leishmania infantum . Parasitology . 2013 ;140 (4 ):455 –60 . 10.1017/S0031182012001886 23253893 62 Roque AL , Xavier SC , Gerhardt M , Silva MF , Lima VS , D'Andrea PS , et al Trypanosoma cruzi among wild and domestic mammals in different areas of the Abaetetuba municipality (Para State, Brazil), an endemic Chagas disease transmission area . Vet Parasitol . 2013 ;193 (1–3 ):71 –7 . 10.1016/j.vetpar.2012.11.028 23261089 63 Lisboa CV , Xavier SC , Herrera HM , Jansen AM . The ecology of the Trypanosoma cruzi transmission cycle: Dispersion of zymodeme 3 (Z3) in wild hosts from Brazilian biomes . Vet Parasitol . 2009 . 64 Marcili A , Lima L , Valente VC , Valente SA , Batista JS , Junqueira AC , et al Comparative phylogeography of Trypanosoma cruzi TCIIc: New hosts, association with terrestrial ecotopes, and spatial clustering . Infect Genet Evol . 2009 . 65 Llewellyn MS , Lewis MD , Acosta N , Yeo M , Carrasco HJ , Segovia M , et al Trypanosoma cruzi IIc: Phylogenetic and Phylogeographic Insights from Sequence and Microsatellite Analysis and Potential Impact on Emergent Chagas Disease . PLoS Negl Trop Dis . 2009 ;3 (9 ):e510 10.1371/journal.pntd.0000510 19721699 66 Fernandes O , Souto RP , Castro JA , Pereira JB , Fernandes NC , Junqueira AC , et al Brazilian isolates of Trypanosoma cruzi from humans and triatomines classified into two lineages using mini-exon and ribosomal RNA sequences . Am J Trop Med Hyg . 1998 ;58 (6 ):807 –11 . 9660469 67 Lopez-Cancino SA , Tun-Ku E , De la Cruz-Felix HK , Ibarra-Cerdena CN , Izeta-Alberdi A , Pech-May A , et al Landscape ecology of Trypanosoma cruzi in the southern Yucatan Peninsula . Acta Trop . 2015 ;151 :58 –72 . 10.1016/j.actatropica.2015.07.021 26219998 68 Barnabé C , De Meeus T , Noireau F , Bosseno MF , Monje EM , Renaud F , et al Trypanosoma cruzi discrete typing units (DTUs): microsatellite loci and population genetics of DTUs TcV and TcI in Bolivia and Peru . Infect Genet Evol . 2011 ;11 (7 ):1752 –60 . 10.1016/j.meegid.2011.07.011 21801854 69 Yeo M , Mauricio IL , Messenger LA , Lewis MD , Llewellyn MS , Acosta N , et al Multilocus sequence typing (MLST) for lineage assignment and high resolution diversity studies in Trypanosoma cruzi . PLoS Negl Trop Dis . 2011 ;5 (6 ):e1049 10.1371/journal.pntd.0001049 21713026 70 Lauthier JJ , Tomasini N , Barnabé C , Rumi MM , D'Amato AM , Ragone PG , et al Candidate targets for Multilocus Sequence Typing of Trypanosoma cruzi: Validation using parasite stocks from the Chaco Region and a set of reference strains . Infect Genet Evol . 2012 . 71 Lewis MD , Ma J , Yeo M , Carrasco HJ , Llewellyn MS , Miles MA . Genotyping of Trypanosoma cruzi: systematic selection of assays allowing rapid and accurate discrimination of all known lineages . Am J Trop Med Hyg . 2009 ;81 (6 ):1041 –9 . 10.4269/ajtmh.2009.09-0305 19996435 72 Carrasco HJ , Frame IA , Valente SA , Miles MA . Genetic exchange as a possible source of genomic diversity in sylvatic populations of Trypanosoma cruzi . Am J Trop Med Hyg . 1996 ;54 (4 ):418 –24 . 8615458 73 O'Connor O , Bosseno MF , Barnabé C , Douzery EJ , Brenière SF . Genetic clustering of Trypanosoma cruzi I lineage evidenced by intergenic miniexon gene sequencing . Infect Genet Evol . 2007 ;7 (5 ):587 –93 . 17553755 74 Cura CI , Mejia-Jaramillo AM , Duffy T , Burgos JM , Rodriguero M , Cardinal MV , et al Trypanosoma cruzi I genotypes in different geographical regions and transmission cycles based on a microsatellite motif of the intergenic spacer of spliced-leader genes . Int J Parasitol . 2010 ;40 :1599 –607 . 10.1016/j.ijpara.2010.06.006 20670628 75 Herrera CP , Barnabé C , Brenière SF . Complex evolutionary pathways of the intergenic region of the mini-exon gene in Trypanosoma cruzi TcI: A possible ancient origin in the Gran Chaco and lack of strict genetic structuration . Infect Genet Evol . 2013 ;16 :27 –37 . 10.1016/j.meegid.2012.12.028 23380053 76 Llewellyn MS , Miles MA , Carrasco HJ , Lewis MD , Yeo M , Vargas J , et al Genome-scale multilocus microsatellite typing of Trypanosoma cruzi discrete typing unit I reveals phylogeographic structure and specific genotypes linked to human infection . PLoS Pathog . 2009 ;5 (5 ):e1000410 10.1371/journal.ppat.1000410 19412340 77 Ramirez JD , Guhl F , Messenger LA , Lewis MD , Montilla M , Cucunuba Z , et al Contemporary cryptic sexuality in Trypanosoma cruzi . Mol Ecol . 2012 ;21 (17 ):4216 –26 . 10.1111/j.1365-294X.2012.05699.x 22774844 78 Yeo M , Acosta N , Llewellyn M , Sanchez H , Adamson S , Miles GA , et al Origins of Chagas disease: Didelphis species are natural hosts of Trypanosoma cruzi I and armadillos hosts of Trypanosoma cruzi II, including hybrids . Int J Parasitol . 2005 ;35 (2 ):225 –33 . 15710443 79 Lisboa CV , Monteiro RV , Martins AF , Xavier SC , Lima Vdos S , Jansen AM . Infection with Trypanosoma cruzi TcII and TcI in free-ranging population of lion tamarins (Leontopithecus spp): an 11-year follow-up . Mem Inst Oswaldo Cruz . 2015 ;110 (3 ):394 –402 . 10.1590/0074-02760140400 25946156 80 Lewis MD , Llewellyn MS , Yeo M , Acosta N , Gaunt MW , Miles MA . Recent, independent and anthropogenic origins of Trypanosoma cruzi hybrids . PLoS Negl Trop Dis . 2011 ;5 (10 ):e1363 10.1371/journal.pntd.0001363 22022633 81 Barnabé C , Yaeger R , Pung O , Tibayrenc M . Trypanosoma cruzi: a considerable phylogenetic divergence indicates that the agent of Chagas disease is indigenous to the native fauna of the United States . Exp Parasitol . 2001 ;99 (2 ):73 –9 . 11748960 82 Tibayrenc M , Ayala FJ . Isozyme variability in Trypanosoma cruzi, the agent of Chagas’ disease: genetical, taxonomical, and epidemiological significance . Evolution . 1988 ;42 (2 ):277 –92 .28567853 83 Campos R , Acuna-Retamar M , Botto-Mahan C , Ortiz S , Cattan PE , Solari A . Susceptibility of Mepraia spinolai and Triatoma infestans to different Trypanosoma cruzi strains from naturally infected rodent hosts . Acta Trop . 2007 ;104 (1 ):25 –9 . 17904090 84 Peterson JK , Graham AL , Dobson AP , Chavez OT . Rhodnius prolixus life history outcomes differ when infected with different Trypanosoma cruzi I strains . Am J Trop Med Hyg . 2015 . 85 Pinto AS , de Lana M , Bastrenta B , Barnabé C , Quesney V , Noel S , et al Compared vectorial transmissibility of pure and mixed clonal genotypes of Trypanosoma cruzi in Triatoma infestans . Parasitol Res . 1998 ;84 (5 ):348 –53 . 9610630 86 de Lana M , da Silveira Pinto A , Barnabé C , Quesney V , Noël S , Tibayrenc M . Trypanosoma cruzi: compared vectorial transmissibility of three major clonal genotypes by Triatoma infestans . Exp Parasitol . 1998 ;90 (1 ):20 –5 . 9709026 87 Roellig DM , McMillan K , Ellis AE , Vandeberg JL , Champagne DE , Yabsley MJ . Experimental infection of two South American reservoirs with four distinct strains of Trypanosoma cruzi . Parasitology . 2010 ;137 (6 ):959 –66 . 10.1017/S0031182009991995 20128943